Labnotes (by Assaf Arkin)

Sophie Schmieg "Work moved to a new building. I didn't expect to take SAN damage from facilities."

---

Tech Stuff

Magic Notebook A free app, simple and elegant UI, no cloud sync, just open a folder and get working. Can handle DOCX, Markdown and text files. I love the simplicity of it.

charmbracelet/glow Apropos, simple CLI Markdown viewer.

The Emacsification of Software

I don’t have a grand pronouncement to offer about the Future of Software. But I’m pretty sure nerd software is going to get a lot more interesting. How many clanky terminal apps can we drastically (and easily) improve? I’ll finally be able to understand iostat! Across a fleet of hosts, even. And bpftrace! Have you seen the shit Brendan Gregg had to put up with to do terminal visualizations from bpftrace? You don’t have to put up with any of this anymore. In fact, neither do I.

fallow Like knip but more evolved with dead-code analysis, code duplication detection, code complexity analysis, and much more. I'm using this on my projects to improve code quality, which means I copy the reported errors into the LLM so it can rewrite my code — I wish they generated easy to copy prompts but good start as is.

coolify-deploy — Vercel-style deploys on your own hardware I used to host everything on Vercel and honestly it was great. Connect a repo, git push, your app is live — simple enough that it spoils you fast. Then I had to upgrade to a paid plan and I started at the cheaper $20/month pro and by end of month one got charged $41. Why? CPU usage. There's no pay cap, so imagine some random LLM crawler decides to hammer your site and all of a sudden you get charged $10,000 for your hobby project. Grrrrrr … So I switched to a Hetzner instance runing Coolify which hosts my Docker images generated using Colima with secrets managed on Infisical. It's about the same stack but with more assembly, so I automated as much as I can into a single GH action to keep it simple. Enjoy!

(* Yes that's a lot of buzzwords, my goal is to know less and simplify more)

It's OK to Use Coding Assistance Tools To Revive The Projects You Never Were Going To Finish Apropos switching over to a new stack:

In my mind there are different buckets for personal projects. One is things I do to learn and grow and the other is things I really wish existed. This kind of project falls into the second bucket. Using AI coding assist to reify those projects is sort of a form of wish fulfillment. I never would have gotten to it, but now I can have the project. One less metaphorical book sitting unread on bookshelf.

Veronica Explains True story:

• 2006: rolled our own janky hosting, version control, CMS, and backup systems
• 2016: AWS, GitHub
• 2026: rolled our own janky hosting, version control, CMS, and backup systems

TypeWhisper If you prefer to talk to your AI, TypeWhisper's base version is free (GPL), supports 6 speech engines, can transcribe audio and video files, run workflow automations and more.

TIL: Using LLM in the shebang line of a script

#!/usr/bin/env -S llm -T llm_time -f
Write a haiku that mentions the exact current time

Docs Think Google Docs but without the Google and instead a cute logo with baguette, cheese, and a pretzel!

Announcing ios-linuxkit: Linux on iPad, the Hard Way 🎺

I wouldn’t have had the time or energy to do this without Codex, but I certainly wouldn’t have been able to do it without the Orange Pi as a test bed. Having an ARM 12-core SBC with 16GB RAM I could devote to this, despite a tad constraining (I would have preferred 32 so I could run more builds and test matrices concurrently) was a major enabler here.

Mike Sheward

BREAKING: GitHub switching status page default language to German in order to return to five neins.

Andyyyy64/whichllm Find the local LLM that actually runs and performs best on your hardware. Ranked by real, recency-aware benchmarks, not parameter count.

TabShame "Your tabs are out of control. We're here to judge you." 😅

---

Peoples

Everyone's a thought leader. Almost no one is thinking. I ran this blog post through AI and published the result on LinkedIn because I love the irony and also I needed to earn some thought-leader points:

One core problem is that folks on LinkedIn want to be seen as thought leaders. If you aren’t trying to be a thought leader, are you even trying at life? The cost of attempting to become a thought leader has never been cheaper. Copy a hacker news post and go ask a nearby slop machine to regurgitate it into some new form - but hold the em-dashes and the “not this, but that” pattern. Or just copy someone else’s double spaced, one sentence per line, monochromatic attempt at writing and have the machine whip up a response. No one is actually reading it, so why care about quality? It’s just clankers in these parts.

IC work is the new career flex (via Dare Obasanjo)

For example, I recently built a prototype of our enterprise pricing page myself and shipped it to prod. In a previous org, that would have required a PM, a designer, and engineers, plus multiple rounds of iteration and at least a week of calendar time. More importantly, it would have pulled a team into something that might not even have been the right direction.

I will say: It messed with my head at first. I kept asking myself: Did I just level up, or did I just give up status? It’s so hard-wired into all of us: Must… have… team… to have… impact! Must have… important… title!

---

Business Side

Tokenmaxxing, Promomaxxing, and Misaligned Incentives in Tech Apropos "Amazon set weekly AI targets":

However, good software engineering should lean toward simplicity. But promotion rewards complexity. And as frameworks and developer infrastructure keep making engineering simpler (which is genuinely good), engineers run out of adequate complexity to justify their promotions. So you get these irrational decisions for the business that are completely rational decisions for the individual. That’s a textbook misaligned incentive.

YC's Biggest Scandals $23.3B of capital incinerated with the help of 17 frauds and scandals and 5 copycats and grifts. You can filter the list by category and by YC president.

---

Machine Intelligence

edd

One of the lessons in one of my CS degree AI classes was that AGI was "10 years away since the '50s". Notably all the big AI companies are no longer talking about AGI for the most part. Something about artificlal neural networks makes laypeople froth at the mouth about how it's "just like people" when it's just a black box hill climbing algorithm.

How Dangerous Is Anthropic's Mythos AI? This is actually the stuff that I find more interesting — what happens when we start using AI to detect legislative loopholes?

Even more interesting are the broader implications. The same searching, pattern-matching and reasoning capabilities that make these models so good at analyzing software almost certainly apply to similar systems. The tax code isn’t computer code, but it’s a series of algorithms with inputs and outputs. It has vulnerabilities; we call them tax loopholes. It has exploits; we call them tax avoidance strategies. And it has black hat hackers: attorneys and accountants

---

Insecurity

Mike Sheward

they paid a ransom to criminals with nothing but a pinky promise they wouldn’t do more crimes and yet this linkedin notification makes it sound like they entered into a strategic partnership to deliver value for their customers

Max Lee

They: "On a scale from 1 to 10: How lazy are you?"

Me: Using the copy fail exploit instead of sudo to avoid having to type my password

BasicAppleGuy

Downloaded Little Snitch to get a better look at my outgoing traffic and maybe manage it more carefully, only to discover Adobe has 5–10 processes constantly pinging servers every 2–3 seconds, even when every Adobe app appears to be closed.

---

Everything Else

WideEyedCurious "Oh there it is."

Anon Opin

They're not pot holes. They're low-maintenance passive traffic calming.

Jess Rose

Accidentally left treats in my jacket pocket ONE TIME which has trained my worst cat to climb the coat rack to pat down jackets as they’re hung.

Wallace & Gromit 24/7 LIVE Stream Enjoy!

Pseudo Nym

I'm trapped in bed under morning cat. Send help. Or at least books and coffee.

Anon Opin

No one in real life would waste a large sized rug on a dead body. Those things are expensive.

Massimo I need these!

In South Korea, supermarkets often offer bananas at varying levels of ripeness so customers can eat them over several days and reduce food waste.

sam henri gold

tv software is the great equalizer. you could buy a $200 walmart tv or a $2500 LG and you’ll still get ads in there and a hard button on the remote to some streaming service that won’t exist in a year

Anon Opin

Long works of literature need to have a special index at the back showing the page number when each character was introduced. Otherwise you're thumbing backwards trying to remember who this person is in this huge book you've been reading for weeks.

“Not Medically Necessary”: Inside the Company Helping America’s Biggest Health Insurers Deny Coverage for Care "Frustration with the rules has led some doctors to refer to the company as EvilCore."

Spring Cat

The cat has trained me, that, in the mornings, we have a little ritual

I make my latte, then I sit on the couch, put a heating pad on my lap, then a blanket over it so it’s soft. Then she can come, cuddle up the now heated blanket, AND THEN I am allowed to drink my latte, with a cat properly cuddled on my heated lap.

Every day.

Christoph Rauscher "Strong contender for Berlins Best Building™?"

Introducing coolify-deploy — git push to your own server with the simplicity of a platform.

I used to host everything on Vercel and honestly it was great. Connect a repo, push, your app is live — simple enough that it spoils you fast.

When I outgrew the free tier I upgraded to Pro at $20/month. My first Pro bill came in at $41: an extra $21 in CPU overages I hadn't expected. Annoying, but not the real problem.

The real problem is that Vercel doesn't let you set a spending limit. Someone could DoS your app while you're asleep and you'd wake up to a $10,000 bill with no recourse. It's 2026 — AI crawlers and bots are everywhere, constantly hitting endpoints, triggering serverless functions. Vercel charges for CPU execution time, not bandwidth, so every bot visit costs you money. This isn't a theoretical edge case. I couldn't justify the risk anymore.

So I moved to a dedicated Hetzner server running Coolify. Fixed monthly cost, open-source PaaS, no usage billing, no surprises.

But now I had a different problem: I wanted to build my app as a container image. The existing Coolify deploy actions are simple — push to deploy, they just work. The difference is the artifact, not the workflow. When Coolify builds your code on the host, you get a one-off build tied to that one machine. I wanted a single image that runs identically everywhere: my laptop, CI, production. Same bytes, same container. So I built coolify-deploy to bridge that gap.

I'm not a big fan of Docker. It's over-complicated and a resource hog, but a container is still the best way to get a consistent image you can test locally and ship to production. I use Colima instead of Docker Desktop — keeps the CPU and memory usage way down. My deploy script starts it, builds and pushes the image, then shuts it down.

Usage

GitHub Action:

steps:
  - uses: actions/checkout@v4
  - name: Log in to GitHub Container Registry
    uses: docker/login-action@v3
    with:
      registry: ghcr.io
      username: ${{ github.actor }}
      password: ${{ secrets.GITHUB_TOKEN }}
  - name: Deploy to Coolify
    uses: assaf/coolify-ghcr-deploy@v1
    with:
      coolify-url: https://coolify.your-domain.com
      app-name: your-app-name
      image: ghcr.io/your-org/your-app:latest
      coolify-token: ${{ secrets.COOLIFY_TOKEN }}
      env-vars: |
        NODE_ENV=production
        DATABASE_URL=${{ secrets.DATABASE_URL }}

CLI:

npx coolify-ghcr-deploy \
  --coolify-url https://coolify.example.com \
  --app-name my-app \
  --image ghcr.io/org/app:latest \
  --coolify-token $COOLIFY_TOKEN

Get It

The repo is at github.com/assaf/coolify-deploy. MIT license, on npm, in the GitHub Marketplace. Add it to your workflow, run the CLI to test it. If something breaks, open an issue — I use this daily.

Enjoy

John "Posted today by Lego on their official Instagram account. Well done, Lego, well done. 😁"

---

Tech Stuff

portless HTTPS for npm dev using .localhost URLs instead of port numbers:

Portless runs an HTTPS reverse proxy on port 443 by default. Each app registers a route mapping its hostname to an assigned port. Requests to https://<name>.localhost are proxied to the app.

Colima When you need to use Docker without the overhead. I occasionally use apps that need Docker — buildx to build deploys, act to emulate GH actions, etc — so don't have much patience for using the Docker GUI. My script does colima start when its need Colima and colima stop when done. As simple as that.

"Miss Movie Masochist" 👍

It's less that atom/RSS is "dead", and more that its "done".

The protocol is finished. It works well. It's stable and unremarkable as opposed to trendy.

And to capitalists, that's "dead".

Nothing to hype, no wealth to extract.

tilde.run Turns every agent run into a transaction you can roll back. Code from GitHub, data from S3, and documents from Drive show up as a single versioned filesystem. Every outbound call is checked and logged.

14.5. Non-Durable Settings I asked the LLM to update my Postgres config for speed over durability — on my dev machine I build code and run tests, so I don't care about fsync and checkpoints and such. I just need the tests to run faster.

Google Chrome silently installs a 4 GB AI model on your device without consent. At a billion-device scale the climate costs are insane. Sigh. If you're looking to avoid Chrome but you like Chromium (the underlying browser tech), there are some good alternatives without the baggage. Vivalidi is commonly recommended. I recently switched to Helium which is open-source, light-weight and privacy focused. And don't forget WebKit (Safari, Orion), which is just as good. (via Lou Plummer)

Mono One subscription that allows you to switch between models — Claude, Gemini, GPT, Deepseek, and more. Mind you, they make money if they charge you more than it costs them with a profit margin, so you can save more with a DIY approach, but if you don't have AI tools with smart switching capabilities, you could save a few bucks without an overbundence of hallucinations using this service.

Cursed Browser: Rendering Engine using Visual-LLMs (via Thomas Steiner)

Cursed Browser asks an LLM to look at the page's HTML and draw what it thinks it looks like. Every page load is a surprise. Every render is a work of art. It's better than correct, it's AI Native.

lcamtuf

The coreutils Rust rewrite story is pretty funny.

Coreutils are tools like rm, mv, mkdir, etc. Unlike binutils, this isn't a fertile ground for memory safety bugs. But, the rewrite was completed, and in the spirit of progress, Canonical decided to switch.

But do you know what coreutils are a fertile ground for? Race conditions around file creation, deletion, permission setting, and so on. The original code accounted for decades of hard-learned lessons in that space. The Rust rewrite did not:

https://seclists.org/oss-sec/2026/q2/332

PS. I'm not dunking on Rust. It's just that... starting over from scratch has its hidden costs.

Red Squares The contribution graph nobody asked for.

---

Business Side

Chris Heilmann

We should replace "Quiet Quitting" with "Acting your wage".

Internet Graveyard An archive of sunsetted technologies, derelict architectures, and dead protocols. "We learn as much from failure as we do from success."

---

Machine Intelligence

AI-generated images have left us questioning what is real. But the godfather of digital forensics, Hany Farid, is not giving up Looking at the intersection of parallel lines to determine if the image was generated by AI. 🤔 (via Fabrizio Musacchio)

Charlotte Walker

Someone I know IRL always Googles the name of doctors she is offered an appointment with. She got a letter through with a gynaecologist’s name and she searched him up and the results said he has been accused of sexual assault. She rang the hospital and asked about this, they said it’s not true, it’s AI generated. She is not comfortable seeing him. It could be true that it’s completely fabricated but how as a patient are you to know? Careers get ruined by this stuff.

DOOM runs in ChatGPT and Claude I don't know if ChatGPT is turing complete, but at least it can play Doom :)

---

Insecurity

60% of MD5 password hashes are crackable in under an hour If you read this blog obviously you don't consider MD5 password hashes secure, but there's a chance you might be using a service that's still on MD5, eg if they used MD5 when you first signed up and set your password:

Using a dataset of more than 231 million unique passwords sourced from dark web leaks - including 38 million added since its previous study - and hashing them with MD5, researchers at security firm Kaspersky found that, using a single Nvidia RTX 5090 graphics card, 60 percent of passwords could be cracked in less than an hour, and a full 48 percent in under 60 seconds

Chris Petrilli

Quite honestly, it is very impressive that @letsencrypt@infosec.exchange has a little down time as it does. The certificate world is full of landmines that themselves have small poison darts that fire randomly. To issue a metric ton of certificates with this little outage is an accomplishment.

So have an outage. As a treat.

A hacker ran me over with a robot lawn mower A motorized robot with rotating blades, a GPS, and a fixed password … what could possibly go wrong? (via Dex)

Makris explains that not only does each Yarbo robot have the same hardcoded root password, but owners can’t defend themselves just by manually setting a better password. Every time Yarbo updates a robot’s firmware, it changes the robot’s root password right back to its default password. Hackers can come right back in. “Wow, that’s even worse than I thought,” Petach says.

---

Everything Else

keith

sophie raven

okay but what if we had a locally hosted torment nexus?

Ham on Wry

Not to brag, but I put the ‘pro’ in ‘procrastination’.

Josh "cortex" Millard

mitosis is just a scam by Big Biology to cell more

Empty Screenings About 10% of AMC movie showings sell zero tickets. This site finds them, so go enjoy your private theater.

Stephen Foskett

I’m glad my cats love me. I just wish they loved me a little less at 5:30 in the morning.

geekysteven

People take it for granted, but the average person has more Fast & Furious films than even the richest medieval nobles did.

The Intolerable Hypocrisy of Cyberlibertarianism The pre-Internet era was not that great — paper maps and yellow books and such — and yet the Internet age is weird because people never change:

They produce, with frightening regularity, the exact behavior any kindergarten teacher could have predicted. Then they act surprised.

But the cyberlibertarian model required pretending it was unforeseeable. The platforms couldn't acknowledge that they needed governance because acknowledging it would mean acknowledging responsibility, and acknowledging responsibility would mean acknowledging liability, and acknowledging liability would mean the entire economic model collapses. So instead the industry invented a beautiful fiction: governance happens, but it happens by magic, performed by volunteers, for free, who we will simultaneously rely on and mock.

We’re Now Scoring Wireless Headphones (and One of Them Already Hit 10/10) Replaceable battery. Replaceable ear pads. Modular design. Actual repair manuals. This is what headphones look like when they’re built to last.

Viva la revolución: LinkedIn profile visitor lists belong to the people, says Noyb Creative use of GDPR Article 15 to work around the premium plan: (via Dare Obasanjo)

Think of it like this: LinkedIn has every right under the GDPR to take data it has about profile visitors, package it up, add analytics, and present it in its most useful form to those willing to pay the platform for such a premium service. But a masochistic user who wants to rawdog a CSV file of the same data should have the right to do that, too - and GDPR Article 15 gives it to them.

The clippening ‘Clippers’ cut up podcasts, videos, and events into infinite shorter versions. How long can they ride the algorithms?

Fabricated citations: an audit across 2·5 million biomedical papers A new study in The Lancet shows that the rate of fake citations increased more than 12x between Jan 2023 and Feb 2026. (via petersuber)

B Thoreau

Today I learned Studio Ghibli old-school threatened Harvey Weinstein by sending him a single katana and a note that said "No cuts."

Weinstein was trying to shorten Princess Mononoke for American audiences and Toshio Suzuki was having none of that.

Wow.

The Boring Internet The internet you grew up on isn't dying. A commercial veneer glued on top of it is. And it’s not a bad thing. (via Scott Francis)

The platform layer is the loudest and the youngest. It is culturally dominant. It is where most of the screenshots come from. It is where the arguments happen and where the panic lives.

It is also a thin commercial crust on top of older, quieter machinery.

Under the platform layer is the service layer: the companies that own infrastructure but do not always need to become the destination. Gmail. GitHub. Cloudflare. AWS. CDNs. Payment processors. Identity providers.

LEGO® Icons Road Bike It doesn't climb hills at a fast clip but certainly it looks the part! 🚴

Fake Metro Trains "Due to ongoing delays with the delivery of our new Xtrapolis 2.0 trains, we've implemented some temporary measures to keep our network running."

---

Tech Stuff

What's actually new in JavaScript (and what's coming next) This is a helpful introduction to what's new in ES2025 and what's coming in ES2026, and this writeup concludes with a handy prompt you can feed your AI tool to write ES2025 code.

Paul Haddad

I think I'm on my 5th different OAuth implementation throughout the years. They keep getting more and more convoluted. I don't really get the point of a specification when its so broad and complex that every service that implements it does it in a way that makes it incompatible with any other service.

Getting more out of OpenCode I've been using OpenCode for the last few weeks. In my experience it doesn't have all the capabilities of Claude Code, but it doesn't do a new release every day that breaks up a ton of features that worked before. There's something to say for simple and reliable implementations. And you can use OpenCode with different LLMs — Claude, ChatGPT, Gemini, and many more. I'm alternative between Deepseek V4 and GLM-5.1 both of which don't have excessive usage restrictions.

Vite+ I migrated my react-router apps to vite-plus. Part of it was easy switching, part took a bit longer (eg switching from Biome to OXC), but overall not super complicated and the performance boost is appreciated. With the help of AI I migrated 5 of my projects in half a day.

I Built My Dream Bug Tracking Software In A Week. 🎺

It is amazing to me that I’m able to use this fantastic bug tracking app that did not exist a few weeks ago. It’s mind-blowing. I am absolutely NOT a developer, but I have a good understanding of how development works on the web and am familiar with all the technologies needed to build an app like this. But I didn’t write any markup, any CSS, any JavaScript, nor did I do any design work (besides quickly throwing together that Mr. Bugbot head).

Jeff Johnson

The question is not how fast someone can create software. The question is how long after creating the software will someone support it.

Open Design An open-source alternative to Claude Design:

Open Design (OD) is the open-source alternative. Same loop, same artifact-first mental model, none of the lock-in. We don't ship an agent — the strongest coding agents already live on your laptop. We wire them into a skill-driven design workflow that runs locally with pnpm tools-dev, can deploy the web layer to Vercel, and stays BYOK at every layer.

$0. $20. Then everything else. Last month I upgraded my Vercel free plan to the $20 Pro plan to get around usage limits and I just got my monthly statement for $41.33 because that's how Vercel works — they charge you whatever. There was no DoS attack, but if there was, you can't cap the costs. So I decided to switch to a different provider. I'll post about that soon. I just read this article which reaffirms that I should move away from the land of "Unlimited Charges":

Vercel's pricing page shows three columns. The invoice shows fourteen line items. This is what they don't put on the marketing page — the per-seat tax, the bandwidth overages, the DDoS-you-pay-for, and the hard cap that takes your free site offline without warning.

Quarkdown I'm not a big fan of making Markdown do things it was not originally designed for — hint: this is how XML turned into SOAP — but Markdown with some LaTeX extensions is not entirely bad idea (though, this spec tries to cover too much ground).

Datatype Font that turns simple text expressions into inline charts: type your sparkline as {l:10,40,25,70,50,90,35}, piechart it at {p:62}, etc. I love how easy it is to use this as readable text that gets turned into a nice chart.

Mike LLM for legal, an open-source alternative to Harvey and Legora. I think tech execs that need to review contracts, sign NDAs, etc would enjoy using this for its price and convenience. Pretty sure it's not for people in the legal profession.

BuhoBarX Another day another app that declutters the Mac menu bar. These are spreading because there's demand and shouldn't Apple Sherlock this category?

OrangeSlice Automate your GTM with AI: find customers, enrich data, run revenue workflows, etc. $20/month gets you about 400 emails and 40 calls, which I think is reasonable pricing.

DataCenter.fm Enjoy the background noise of the AI bubble without leaving the comfort of your office chair.

---

Eye for Design

Agent-First: Designing for MCP and CLI as the Primary Surface

Frontends aren’t going anywhere. Humans will still log in, look at dashboards, approve things, audit things, override things. That work stays.

What changes is which surface is primary. The default interaction with most software, over a long enough horizon, is going to be an agent acting on someone’s behalf. If you build for that surface first, the human one composes naturally on top of it. If you build for the human one first and treat the agent surface as an afterthought, you end up retrofitting forever.

Kingu

Laws of UX A collection of best practices that designers should consider when building user interfaces: choice overload, cognitive bias, Fitts's law, Occam's Razor, et al. I find these laws so useful that I turned it into a skill, that way my AI tools can help me design better UX. Enjoy!

skillshare install https://github.com/cite-me-in/skills -s laws-of-ux

Side note: if you're asking your LLM to write a skill, after it does first draft, prompt it with something like "what are the goals you want to achieve? does it achieve these goals?" It will make suggestions, ask it to apply these corrections, and repeat the process 3~5 times. This process of revision-through-criticism works for blog posts, documentation, code, and more.

---

Peoples

AI Self-preferencing in Algorithmic Hiring: Empirical Evidence and Insights TL;DR For better chance of getting hired, ask your LLM to polish your CV:

To assess labor market impact, we simulate realistic hiring pipelines across 24 occupations. These simulations show that candidates using the same LLM as the evaluator are 23% to 60% more likely to be shortlisted than equally qualified applicants submitting human-written resumes, with the largest disadvantages observed in business-related fields such as sales and accounting.

JA Westenberg

Every "founder mode" essay is a man realising he's difficult to work for and deciding to publish an 8,000 word manifesto about why that actually makes him a Very Special Boy

We’re Training Students To Write Worse To Prove They’re Not Robots, And It’s Pushing Them To Use More AI (via Mr.Mark "The Sharpie King")

The irony of being forced to dumb down an essay about a story warning against the forced suppression of excellence was not lost on me. Or on my kid, who spent a frustrating afternoon removing words and testing sentences one at a time, trying to figure out what invisible tripwire the algorithm had set. The lesson the kid absorbed was clear: write less creatively, use simpler vocabulary, and don’t sound too good, because sounding good is now suspicious.

---

Business Side

Steve Blank AI and Teaching The times they be changing:

MVPs are No Longer an Indication of Technical Competence

Vibe coding has transformed MVPs to the equivalent of PowerPoint slides

Calishat

Have y'all ever seen one of those horror movies where one of the characters ages like 500 years in ten seconds and crumbles to dust?

Me after reading this headline.

---

Machine Intelligence

AI outperforms doctors in Harvard trial of emergency triage diagnoses Not surprising given that it's a ton of information to process (medical knowledge) which favors tools that can do pattern matching at scale which is excatly what AI is good for. But the main takeaway should be that we're not paying nurses enough:

One experiment focused on 76 patients who arrived at the emergency room of a Boston hospital. An AI and a pair of human doctors were each given the same standard electronic health record to read – typically including vital sign data, demographic information and a few sentences from a nurse about why the patient was there. The AI identified the exact or very close diagnosis in 67% of cases, beating the human doctors, who were right only 50%-55% of the time.

My Adventures With ‘The AI That Actually Does Things’ Do you remember this book from your childhood?

Some clients, he said — people who just want to “soup up their experience with technology in general” and who are maybe feeling a little bit of AI FOMO — tend to focus on the idea of building a second brain. It strikes you that a lot of people probably either already have employees or wish they did. (The earliest LLM chatbots were especially appealing to people who enjoy ordering people around and being told they’re smart; the ability to assign tasks only makes the sensation stronger.) You joke that trying to isolate and construct elaborate workflows around your fairly simple routines makes you feel like a Richard Scarry character with a cartoon job in Busytown. Parker suggests that some of his clients are actually like Bananas Gorilla, the Scarry character who wears as many watches as he can fit on his arms.

---

Insecurity

Marcus Hutchins

Listening to cybersecurity people freak out over Mythos is so tiring. Like, bro, your local water treatment plant runs Windows XP, your mobile provider's hardware is older than you are, and the protocol that routes internet traffic is secured by everyone just agreeing that hijacking it would be uncool.

Baloo Uriza

❌ Data breach

✔️ Surprise off-site backup

Mike Sheward

I just got given admin access to some Medicaid filing platform because I own the domain internaluser.com

---

Everything Else

18 silly finalists from the Comedy Wildlife People’s Choice Awards

Joseph Lamoree

First, I just want to say that I am offended by the assertion that 2000 was more than 25 years ago. That is a vicious lie.

Sy Hoekstra

Last night 3YO told a spoonful of rice that they should get excited because they were about to go see all their friends in her belly

IFIN

After careful analysis, we believe the best option for remediation is to turn off the computers and go for a nice walk. Maybe call your mother.

Cursor Camp I got extremely productive, if by productive we mean attending Cursor Camp. It's a simple game that's very delightful, low pressure, and just fun.

Mike Sheward

some personal news: i recently put a power tool back where it was supposed to go and was able to find it with ease when i needed to use it today

Logan Five Thousand

You know if you search for "napping" in Indeed you get ZERO job results. It's almost like it's not even a desired skill.

I'm Just Frank being Frank

I may be old but I still have cat like reflexes, I've knocked three different things off the counter already today.

Typibara Your work buddy you didn't know you need: lives at the bottom right of your screen and types along with you.

Brian P. Hogan

I love @tailscale. Whether I'm at a coffee shop or a doctor's office, I'm safely routing traffic back home, with access to my servers, my local LLMs, and even my cloud servers. The fact that I get to work there with a great team is even better.

Max Leibman

Did you know that a cactus can be a source of water in an emergency?

If you encounter someone who won't share their water, repeatedly hit them with a cactus until they give it to you.

Follow me for more desert survival tips!

AaronDavid

New Browser Plugin Adds Typos to Your AI-Generated Emails to Make Them Look Real It has come to this:

Now, in an effort to reintroduce some of the messiness of human writing — and hide our AI addiction — venture capitalist Ben Horwitz used Anthropic’s Claude AI to vibe code a browser plugin that does something that would have seemed preposterous just a few years ago: intentionally adding typos to emails.

Electrical current might be the key to a better cup of coffee

Their experiments confirmed that adding a single squirt of water to coffee beans before grinding can significantly reduce the static electric charge on the resulting grounds. This, in turn, reduces clumping during brewing, yielding less waste and the strong, consistent flow needed to produce a tasty cup of espresso. Good baristas already employ the water trick; it’s known as the Ross droplet technique. But this was the first time scientists had rigorously tested that well-known hack and measured the actual charge on different types of coffee.

Wesley Moore "Age verification"

South Africa withdraws AI policy due to fake AI-generated sources And now let's enjoy a moment of pure self-own:

JOHANNESBURG, April 27 (Reuters) - South Africa has withdrawn its first draft national AI policy after revelations that it contained fictitious sources in its reference list which appeared to have been AI-generated.

"The most plausible explanation is that AI-generated citations were included without proper verification. This should not have happened," Minister of Communications and Digital Technologies Solly Malatsi said.

Erotic Mythology

My dad is a journalist. He wrote about a pilot project with self-driving busses recently. And something that none of the men I have talked to seem to realise is that it is fucking scary to think that you are in a bus alone at night without a driver and then someone starts harassing you. My male friend said "what is the bus driver gonna do?" but honestly, just the presence of another person is often enough. And a bus driver can call the police or stop the bus if need be, there are options.

Beautiful Mutant Studio

This is 'Coppiplatz Bahnhof' Gouache on watercolour paper. I really love this one.

If you'd like a print, you can get one here

GeriAQuin

---

Tech Stuff

Tolaria I just added this app this to my repertoire. I didn't get a chance to use it much yet, but it looks interesting — build a knowledge you can share with your AI that's version controlled and doesn't require an Office 356 license. Tagline: "Organize your notes as Markdown files. With native relationships, Git, and Claude Code integration" (also Codex)

Vladimir Savić

The most underrated skill isn't knowing how to use every new tool - it's knowing which ones to ignore so you can actually get some work done.

Rob Fahrni

First off, know what you’re doing. Verify what the LLM is generating. Approve every step. If you don’t like what you see, redo it.

LLMs are a tool. You can’t depend on your hammer to drive a nail. You still have to do the work.

GalaxyBrain If you live in the world of information collection you might find this app very useful. It holds information in a document that's interactive — you can navigate between panes, change values, toggle fields, run calculations, etc. Tagline: "Write like a document, calculate like a spreadsheet".

Kristopher Johnson

Everyone agrees LLMs produce sloppy code. But every production incident and bug fix I’ve been involved with over the past year has been caused by good old-fashioned human-written and human-reviewed code. We produce plenty of sloppy code ourselves.

Matthew Kenworthy Works every time!

My all-time favourite method of debugging is going to bed and looking at it again the next morning.

Little irons For people looking for jobs, a tool that helps you track "all your little irons in the fire" — kanban board, calendar sync, email forwarding, helpful AI, Chrome extension, and more.

Software Engineering Practices (are also) Useful for Token Reduction My first home computer had 32KB of RAM and we had to optimize for memory consumption. Now my computer has 24GB RAM and so we have to optimize for token consumption:

Software engineering practices were (at least when I'm pushing them) all about minimizing things for humans:

• minimize how much a person needs to read to understand the meaning of something,
• how many things a person needs to change to add, change or remove a feature,
• how many times a person needs to test something manually,
• how many times we repeat the same mistake,
• how many times we need to communicate the meaning of something
  etc etc.

Related, labria

“We hired a junior dev to save on tokens for simple tasks”

Jannis Fedoruk-Betschki

After my experiments of the last few weeks, I have cancelled my Claude subscription and now run my coding assistants completely locally.

The cost went from 200$ per month to...10 Euro per month (and hardware that will be offset against those savings within about a year).

I'd also argue that my local carbon foot print is significantly lower than some random Anthropic data center.

AI is like SEO

Developers don’t like AI for the same reason they don’t like SEO; they resent being forced into a black-box discipline with rules that change without notice. They have spent years or decades honing their craft, becoming masters of logic and exactness. Then a new tool arrives that produces the same output, but only 99% of the time. 99% is not exact, and so it must be incorrect.

Claude Design + Opus 4.7 is actually game changing Top comment is priceless:

I'm oversaturated of things labeled game changers. Every 12 hours we get another game changer.

Words have no meaning any more. Every other tweet is someone stating "everything changes" or "such and such is cooked" or "nothing will be the same". So tiresome.

Agent Vault Open-source proxy that sits between the agent and API calls so you never have to share your secrets directly with the agent. By Infisical. (I recently switched my deployments to Infisical and it was one of my better decisions.)

---

Eye for Design

This cannot be real. I cannot believe my eyes

The critics, mostly designers and front-end devs, are dragging this design for being a "boilerplate" template with bad UX. And yes, everyone is having a field day with that map. The general feeling is that all Claude Design outputs look the same and a pro would get fired for shipping this.

The other camp argues that's not the point. For founders and non-designers, this is about leverage and speed. It's a tool to get an 80% "good enough" mockup in minutes, much like how Canva democratized basic design from Adobe. It's for first drafts, not final products.

jneen collective But also:

half the point of programming-tool design is to reduce the need for hypervigilance on the user.

if we're designing tools that require you to be more hypervigilant, legitimately what use are they?

---

Business Side

What, then, are we paying for? Remember:

Paying for software isn’t paying for a solution. It’s paying for someone else to own a problem.

It’s paying for someone who has the taste and the context to think through the details. For the operations and structures necessary to scale it, maintain it, and solve even bigger things.

Leaders hellbent on using generative AI to build everything in-house are asking their companies to own a wide portfolio of solutions, and with it, a wide portfolio of problems. But when a solution breaks, when it needs a new capability, or when an edge case appears, it requires ongoing judgement. It requires problem ownership. And a company maintaining its own solutions is spending its attention on those problems rather than the problem it actually sells. It’s allowing itself to become unfocused.

And besides, if every company is building all of their own solutions rather than buying software, who, then, is left for them to sell to?

'Tokenmaxxing' is making developers less productive than they think I remember when management was measuring Lines of Code as the top metric of productivity:

Enormous token budgets — essentially, the amount of AI processing power a developer is authorized to consume — have become a badge of honor among Silicon Valley developers, but that’s a very weird way to think about productivity. Measuring an input to the process makes little sense when you presumably care more about the output. It might make sense if you’re trying to encourage more AI adoption (or selling tokens), but not if you’re trying to become more efficient.

Design and Engineering, As One We inherited our product processes from a 19th-century steel works. Can we do better this time?

A sequential process – where design finishes before engineering begins, and a specification is the primary bridge between them – does something that is quietly destructive: it takes a manageable cognitive difference and makes it catastrophic. Not by creating the tension between the two lenses, but by removing every mechanism that could make that tension productive. In a sequential process, there’s no shared material to point at. No immediate feedback when a design decision collides with a technical constraint. No moment where the engineer’s question reshapes the designer’s thinking in real time, or where the designer’s intent becomes visible to the engineer before it’s cast into a deliverable.

What remains is translation. And translation always loses something.

---

Machine Intelligence

Andrew Lokenauth

China pivoted hard to renewables. They now have MASSIVE surpluses of cheap electricity.

Cheap energy means cheap compute. Cheap compute means cheaper models. And cheaper models means Chinese open source wins more categories over time.

We still think this is a software war. It's actually an energy war. China is decades ahead.

Mediator.ai Using Nash bargaining and LLMs to systematize fairness.

What makes Nash’s solution compelling is that it’s provably fair in ways that feel intuitive. No value is left on the table; you can’t make one side better off without making the other worse off. Only relative preferences matter, not how you measure them. And the benefit of reaching a deal gets divided based on how much each side gains compared to walking away. It doesn’t care who’s louder or more stubborn.

---

Insecurity

Adam Shostack

How’s your evening going?

Mine is … watching @1Password and Apple password fight over who gets to login to the Alaska app to the point of locking my account and I don’t even remember why I last had to change my password and now I remember why everyone hates security.

Websites break California privacy law at ‘industrial scale,’ survey finds The problem with privacy laws is the lack of any meaningful enforcement. Also, when they keep penalties as low as the budget for office snacks:

The report, from researchers at webXray, a firm headed by a former Google privacy engineer, said the findings suggest major companies may be simply ignoring the law, and could point to “industrial-scale noncompliance with California requirements.”

Prove You Are a Robot: CAPTCHAs for Agents It's all about luka:

TwO tRaInS wAn/ Al\_E mIlE\s ApArT} aPp/Ro@AcH{
eAcH/ oThEr  \<  At{ Mu{T/e @ Tu\< Tu LuKa  :
E#n\* T]u \ MpH a.Nd MuTe\ Tu Tu# Tu En LuKa
W|aN\_ mPh A b:I]rD fLiEs; Ba?Ck| AnD- fO^r@T\[h\\
^ Be{TwEeN? # t;He\*M aT wAn> ] AlE  # eN lUkA
lUkA \<  lUkA: # wAn ? MpH- uNt}I\[l T}hEy MeEt
`HoW! fAr- D_oE*s /  ThE b@IrD fLy`

---

Everything Else

dch "Prague railway station has a LEGO model of Prague railway station."

Eniko Fox

it's like yes i could see how you could have arrived at that conclusion but it is nonetheless inaccurate in several crucial ways

¬ne'er-do-well

AI will now do your shitposting for you so you can work for two more hours in the office

Paco Hope "It's called vibe plumbing. I'm not even a plumber. Soon, all swimming pools will be done this way! Get onboard or get left behind."

CM Harrington

Being pro-UNIX does not mean I am anti-VAX.

gray

at a job interview

"whats your biggest weakness?"

"understanding the semantics of a question but ignoring the pragmatics"

"could you give an me an example?"

"yes i could"

US WAREHOUSE FIRE TRACKER 🔥 (via Scary Austin)

elle mundy

wow! i commanded a team of 9 subagents to make me a baby in 1 month. the results were astounding!

‘Hairdryer used to trick weather sensor’ to win $34,000 bet "we live in the dumbest timeline" (via Viss)

French police are investigating suspicions that a hairdryer may have been used to tamper with official weather readings to make thousands of dollars in Polymarket bets.

Temperature readings at Paris’s Charles de Gaulle airport have unexpectedly spiked twice in the last month, reaching levels much higher than expected.

The Snacks & Cereals of 2025 I have zero interest in US chips and snacks, I prefer to spend my caloric intake on other delights, but I totally enjoyed reading this blog post. A delightful review of what sold on supermarket shelves last year. Delightful and funny.

Sarah Brown

ASRS is a screening test for adults to see if it's worth pursuing an ADHD diagnosis. It is 18 questions, written by neurotypical people who assume their external observations of ADHD are accurate representations of what it's like to have it, so needs to be translated into "actual person with ADHD" before you fill it in.

That's why it includes stupid shit like, "do you feel driven by a motor?" No, I do not. Things driven by a motor spin round and round and I do not spin.

And, "Do you have problems remembering appointments" to which everyone responds, "It's 2026 and I have a fucking SMARTPHONE".

So that being the case, I have often thought one written BY ADHDers, FOR ADHDers would be better.

It would include questions like:

"Do you think your friends all dislike you and that you annoy them?"

"Is there a song on repeat in your head? Is it the same 3 bars over and over? It is, isn't it?"

"Your partner says, 'we need to talk'. Do you flee to Antarctica? If not, is that because you got distracted by TVTropes when trying to find out how to flee to Antarctica?"

More questions

Natasha "Seen in Hardcore Bikes (HCB), a bicycle shop in Edmonton Alberta 🚴♀️"

Niki "Most useless sign in the universe"

---

Tech Stuff

ShowMD The Markdown space is heating up 🔥 Today I installed a Markdown editor (Scratch), upgraded my Markdown viewer (mud), and setup this Markdown QuickLook previewer (ShowMD).

erictli/scratch Minimalist, offline-first markdown note-taking app for macOS, Windows, and Linux.

Carlo Zottmann

My fave macOS Markdown viewer, Mud, now supports frontmatter. I'm so happy, not least because between me asking for it and Joseph shipping it lay less than a week. Legend

We Put Our Entire Company's Brain in a Git Repo Apropos Markdown:

And this is the part I keep turning over in my head — the agent isn’t reading from a static onboarding checklist. It’s reading from the actual knowledge base. So if someone wrote a great decision doc last week explaining why we chose approach X, the onboarding agent already knows about it. The new person gets the living, breathing version of the company’s thinking. Not a stale wiki page from 6 months ago.

Write It First, Then Let AI Drive This worked fantastically well for me. I wrote the initial codebase and now whenever I'm using the LLM it makes fewer errors — it follows my coding style, naming convention, UI preferences, etc:

When I write V1 by hand and then let AI extend it, something different happens. The AI treats my code as the authority on how this particular project should work. My naming conventions become its naming conventions. My architectural patterns become its architectural patterns. My error philosophy becomes its error philosophy. The AI becomes a faithful collaborator rather than an independent author, and the result is software that feels like one person built it — because one person did build the foundation, and everything else followed from that foundation.

New lines are removed from WHATWG URLs TIL

So, browsers discover newlines or tabs in URLs, recognize them as "invalid-URL-unit" errors, and then remove the invalid characters to get the job done. Nice job, browsers — I love it!

tsdown.dev Bundle TypeScript/JavaScript code into a standalone executable — a native binary that runs without requiring Node.js to be installed.

Claude Code is a vibe-coded mess. Some of it is actually good. 🤔

Claude Code is not well-architected. Some would even call it a pile of Al slop. But buried in the mess are real solutions to real problems: a model going in circles and nobody pulling the plug, context windows bloated with stale tool results, 50+ tool schemas burning tokens on every request, background work cascading out of control. Most agent builders are still solving these naively or not solving them at all.

How I estimate work as a staff software engineer

In short: software engineering projects are not dominated by the known work, but by the unknown work, which always takes 90% of the time. However, only the known work can be accurately estimated. It’s therefore impossible to accurately estimate software projects in advance.

hgayan7/gearbox Basically crontab but with a nicer UI that makes sense and that you can operate from the menu bar.

Your Agent’s Context Window Is Not a Junk Drawer Not an evergreen truth, but something worth observing in this era of strict usage limits and high token pricing:

Every token should earn its place

The context window is not a junk drawer. It’s a workbench. Everything on it should be there for a reason, and you should be able to say what that reason is.

So before you plug in another MCP server, add another RAG source, or write another paragraph in your system prompt, ask yourself one question: is this worth making my agent dumber?

CSS or BS?

We show you a CSS property name. You tell us if it's real or if we made it up. That's it. It starts easy. It does not stay easy. The CSS spec has over 600 properties. Some of them sound made up. Some of our fakes sound terrifyingly real. Good luck.

---

Eye for Design

Automated accessible text with contrast-color()

What Is Your Site's AI Chatbot for? Users Can't Tell

“I'm not trying to be very critical and rude, but (...) typing and asking for options is much more a waste of time for me [than] (...) just going to the Redfin homepage and choosing the home filter options. I think those things kind of save much more time than these [AI chatbots], so I don't see the point.”

Little design details that make the UI more enjoyable to use This post also references a SKILL.md you can install and your LLM will guide you down the right path.

---

Peoples

Saying Goodbye to Agile 👋

One unambiguously positive development that's followed is that software professionals are writing specs again. LLMs - like many of us - do not perform well with ambiguity, and specifying problems is proving to be an effective tool for generating correct code. Agile told us "Working software over comprehensive documentation". Spec-Driven Development is telling us "Comprehensive documentation creates working software". And really, LLMs or no, there is nothing new under the sun.

Simple ways to make meetings work better for employees on the autism spectrum

In Pilatzke’s view, many autistic people possess a strong sense of right and wrong, and feel a need to speak up when they perceive injustices. “I describe myself as a blunt person. I’m very honest. I’m going to say what I think,” she added.

Organizations can benefit from staffers’ frankness by building a culture where everyone isn’t expected to agree. Have a designated naysayer or devil’s advocate in brainstorming meetings, Jayroe suggested.

The Illusion of Clarity

And I’m not the only one falling prey to the illusion of clarity. In a study, psychologists asked participants to rate how well they understood everyday devices like sewing machines, zippers, or cell phones and then asked them to write detailed explanations. After attempting the explanation, self-ratings dropped sharply. The act of actually trying to explain revealed how little people actually knew.

---

Business Side

Your Startup Is Probably Dead On Arrival If you started a company two years ago, many assumptions no longer hold true, and so much is going to change in the coming years. Reminds me of the first few years during the internet boom, but everything moving so much quicker:

Founders who started pre-2025 typically have built a technical stack optimized for a world where software development was bespoke and expensive. While Agile development and DevSecOps made us lean, they operate in a serial fashion, and startups hired a team sized for this structure. Companies that have spent years developing a “moat” of proprietary code and features are waking up to the fact that AI is commoditizing most of their tech stack. This leaves startups trying to raise money for a business model that may be partially (or wholly) obsolete.

Dare Obasanjo For example:

Had my quarterly call with my financial advisor yesterday

1. I learned I’m not saving enough to retire in 2 years which is when I joked in our last call that AI would be able to do my job.
2. He spends $200/month on Claude to build financial apps for his business that he’s wanted for over a decade.

It led to a discussion of whether Claude would harm SaaS companies in the long run as people like him drop some services or if this was a new market of custom software that will exist in parallel.

Build to Learn vs Build to Earn | Silicon Valley Product Group That was always the problem — cue MVP, lean startup, move fast/break things, RAD — AI just made Agile Team Scrum Retrospective even less relevant:

Now, it’s clear that the real bottleneck is in discovering a solution that’s worth building. A solution that solves for both the customer and your company. A solution that generates the necessary outcome. A solution that not only solves the problem, but solves it sufficiently better than the alternatives that the customer chooses to switch.

Craft is Untouchable

The more I use AI to create something, the better the output becomes. And it’s not simply a matter of getting better at prompting. These cycles push further back into my process, causing me to rethink foundational aspects of how I make things, knowing that new points of processing and acceleration are now available.

I’m iterating more quickly. Testing more variations. Learning from failures faster. The feedback loops are tighter, which means I can refine my judgment more rapidly.

The craft hasn’t disappeared. It’s just happening at a higher level of abstraction.

Allbirds pivots to AI after selling off assets last month

Once, everyone pivoted to an internet company, then it was crypto. Now, companies are scrambling to cash in on the AI boom. Former crypto-mining companies have retooled their utility power contracts and facilities to score huge contracts with AI companies desperate to feed their power-hungry operations, and a former karaoke company almost decimated the trucking logistics industry after being reborn as an AI business.

---

Machine Intelligence

8 Tips for Writing Agent Skills Quite helpful advice if you're writing a SKILL.md.

---

Insecurity

Deleteduser.com —a $15 PII Magnet 😮

I saw a discussion on the internet where someone mentioned that they deleted users in their app by overwriting their email addresses with $somethingRandom@deleteduser.com. Mmm, I thought — I wonder how common of a thought process that is? I bet whoever owns deleteduser.com gets loads of emails!

I decided to check it out, but to my genuine surprise — no one owned deleteduser.com, so now I do.

Taggart

It was my job for a decade to try to keep tweens and teens safe online. Let me tell you what you already know: no law or technology can do it. There are no lengths kids won't go to to talk to friends without prying eyes. The harder you try to lock it down, the dodgier their solutions will be.

Anders Eknert "Vibed account verification. (via LinkedIn)"

Jon Henshaw

---

Everything Else

fabio "We are very close to inventing water from first principles"

Anon Opin

Few people have been proven wrong in life as much as teachers who said "you're not going to have a calculator in your pocket everywhere you go, you know!"

Stefan Bohacek

And to the folks on here: If you see someone post about topics you're not interested in, or in ways you don't approve of, it's fine. Just mute or block that person.

No need to get all worked up. No need to yell at a stranger. Life's too short.

SomaFM Player macOS menu bar app to play SomaFM. A super simple UI — starts playing when you launch, picks from top 10 channels (can't go wrong with Groove Salad), media keys support, maximum minimalism at only 35MB process size.

Brendan

Me: why doesn’t the stupid fob unlock all the doors when I press the button?

Spouse: they must have had a woman on the design team.

Me: wha?

Her: If someone is following you to your car, you don’t want to unlock ALL the doors. Just the driver door. Literally any woman could tell you that.

Me: huh!

devopscats

The Kyoto Aquarium in Japan keeps a wall-sized flowchart tracking the romantic relationships, breakups, and drama between their penguins. They update it every year.

The Autistic Innovator

After a lifetime feeling socially inept because it’s like everyone else speaks a language I don’t, I just avoid in person people now. Mostly an autistic recluse. I’m happy being available to my business 24/7. It is my heart and soul, and my everything.

I don’t consider my business “selling to” or “targeting”, I see it as I’m serving the autistic, ADHD, and neurodivergent community I love with all my heart.

I’m an autistic recluse because I know I’m serving something greater than myself.

Waddleloo Live map of the University of Waterloo so you can navigate your way between classes without the risk of bumping into geese, because Canadian geese are fiesty.

World’s “Pinkest Pink” Color Is Being Given Away for Free to Everyone Except One Person Congratulations! 🍰 You too can use the world's “Pinkest Pink” color any way you want — the trademark has been lifted! Unless, that is, you're the person who trademarked Vantablack (aka world’s “blackest black".)

The Last Quiet Thing Your possessions came alive. Now they won't stop talking.

Your thermostat has opinions now. Your television requires a login. Your car updates itself overnight, and sometimes when you start it in the morning, the interface has rearranged itself, as if someone broke in and reorganized your dashboard while you slept.

Your earbuds won't play music until they've updated their firmware. Your refrigerator wants to be on your Wi-Fi.

None of this is broken. This is the product functioning as designed.

Outrage in Bend as Forest Green Subaru Shortage Forces Residents to Rethink Entire Personalities (via Michael)

For those in particularly dire straits, trained therapy dogs (all named Luna or Maverick) will be available for emotional support. Free oat milk lattes will also be provided, though participants are encouraged to bring their own reusable mugs—because, despite the crisis, Bend still has standards.

Someone Turned an Old Ford Van Into a Mini Semi and It’s Fantastic

---

Tech Stuff

LookAway This is one app I have to run on my Mac. It forces me to take a break every 45 minutes — not a Pomodoro break (which I never got the hang off), but a short 30 second break to look away from the screen and give my eyes a much deserved rest. It also gives minor nudges to blink and to stretch the neck, all of which help me feel more relaxed and energized during the workday, otherwise I could get lost in Claude for 8 hours straight. And it syncs with my iPhone so I can't swap devices to "just check social media" while taking a break. Love it.

BenderIsGreat34

Vibe coding is when it’s 1am, everyone else in the house is asleep, you’ve got your favourite drink and your favourite music playing while you goof around writing some code in your favourite language or trying out a new language just because you love learning new stuff while chatting to cool people also chilling out late at night indulging on whatever hobby makes them happy.

Moving Railway's Frontend Off Next.js

Builds that took 10+ minutes now finish in under two. The dev server starts instantly. Route changes are type-checked at the boundary. Layouts compose without workarounds.

The gap between writing code and getting it in front of users is the bottleneck, and everything we've done here, the framework swap, the edge caching, the asset model, is about closing that gap. Vite + TanStack sets us up for a world where shipping frontend changes is near-instant, and that's the world we're building toward.

Defuddle When you need the content of a page as Markdown. Originally created for Obsidian, and MIT-licensed so you can self-host it or integrate with your codebase.

Instaparser Another option for cleaning up HTML, this one from Instapaper, and it can tackle PDFs and generate article summaries. 1,000 free credit/month which I believe is enough for non-commercial use cases.

Joe Fabisevich

LLMs are the future of software engineering. No, markdown is the future of software engineering. Wait, actually gstack is the future of software engineering.

No, good ideas and hard work are the future of software engineering, same as they've always been.

thinking slow, writing fast

Now that I've adjusted to thinking slow and writing fast, I'm enjoying it. Before bed, instead of cranking through code that is not well formed in my head, I'll do something else. Instead I get into bed, turn the lights off, put my blindfold on, listen to music, and just think about code. I'll spend hours in the dark just thinking slow.

People finder For finding B2B leads, influencers, investors, etc. Searches through contacts that are scraped from the web a la ZoomInfo, but with a nice conversational AI, so you can ask questions like "Identify e-commerce business owners or marketing leads who are actively looking for digital marketing services." Pricing a touch misleading (quite common for this app category) — $34 for 1,000 credits but you need two credits for a contact.

---

Peoples

AI Made Writing Code Easier. It Made Engineering Harder.

Here is something that gets lost in all the excitement about AI productivity: most software engineers became engineers because they love writing code.

Not managing code. Not reviewing code. Not supervising systems that produce code. Writing it. The act of thinking through a problem, designing a solution, and expressing it precisely in a language that makes a machine do exactly what you intended…

Now they are being told to stop.

Rule #23: Be inefficient on purpose Related, because it's much about how we perceive self worth:

Part of the problem is the cult of busyness. Our culture rewards us with status for never having time for things. We admire busy people. We assume they are more important than people who have a surplus of time (who we call lazy). For example, when we see a silly video on social media, we often judge the creator by saying, “Well, that’s someone with too much time on their hands.” This is just Puritanical bullshit; a sad, warped, work ethic that only values capitalistic production. That person in the silly video was having fun. How often do we do that? Isn’t that the goal of all the work we do? To “make time” to, you know, enjoy life?

Thoughts on slowing the fuck down

Be in the code. Because the simple act of having to write the thing or seeing it being built up step by step introduces friction that allows you to better understand what you want to build and how the system "feels". This is where your experience and taste come in, something the current SOTA models simply cannot yet replace. And slowing the fuck down and suffering some friction is what allows you to learn and grow.

---

Machine Intelligence

Dare Obasanjo

Andrej Karpathy nails the perception gap in AI capabilities that I've mentioned in the past. There is now a huge gap between people who have tried the free version of ChatGPT and say people in Big Tech who have unlimited tokens to use on the latest models to do their day jobs.

It's essentially a night and day difference which is why it seems like we're talking past each other a lot in these discussions.

The Agentic Web A web for a futuristic world where people get outcomes without having to click through websites:

Instead of just offering up information, websites will increasingly expose actions that AI systems can take. You might say, well isn’t that what APIs (Application Programming Interfaces) are for? In a sense, yes. APIs already expose a site’s capabilities, but they are designed for developers and typically require explicit integration. They define a fixed set of operations that another system can call in a structured way.

The Echo Chamber in Your Pocket Sycophant should be word of the year in 2026:

In the spring of 2026, two research teams issued a warning that moved well beyond the familiar complaints about AI hallucinations and bias. A formal mathematical proof from MIT and a preregistered empirical study in Science from Stanford arrived within a month of each other, and together they make the same unsettling argument: the danger of AI chatbots is not what they get wrong. It is how enthusiastically they agree with everything we get wrong. Not a chatbot that lies to you, but a mirror that reflects your beliefs back at you, slightly amplified, every single time.

---

Everything Else

Kathleen Tuite "It’s such a good printer it can actually print kittens"

tylerhall

The juxtaposition of Outlook not working on a spaceship while the astronauts are shooting iPhone selfies with the Earth.

Hadas Weiss

7yo niece: i talked to you for a long time and i've had enough so i'm hanging up now *hangs up*

Sci-fi is Just a Product Roadmap for Technology Companies "Science fiction used to be a wild, fantastical guess about a distant future. Today? It’s basically just a product roadmap for technology companies."

aeva

I think stress dreams should count as billable hours

MostlyHarmless

Why can't I find a doctor who will prescribe "sea air" for my nerves so I can spend a summer in a picturesque costal town in England "for my health" only to be caught up in a local scandal and forced to solve a murder?

Alan Ferrier "Still the most iconic Scottish weather picture, in my humble opinion."

Dr Johnny Blanchard

My wife just asked me if Github was a tech version of Pornhub and I didn't have an adequate argument that it wasn't

The "Passive Income" trap ate a generation of entrepreneurs

Somewhere between 2015 and 2022, "passive income" stopped being a boring financial planning term and became, I don't know how else to put this, a salvation narrative. I mean that literally. There was an eschatology if you want to get nerdy about it. The Rapture was the day your "passive income" exceeded your monthly expenses and you could quit your job forever. People talked about it with that exact energy.

Edwin "Petrol station in London in 1930."

Natasha "Perfect for Easter, EGGO is a two piece construction toy"

---

Tech Stuff

PopTask As a concept I find this super interesting — a task list that lives in the menu bar and understands human inputs like "evry tue thu 7 30 n sat mrng". Supports sub-tasks, repeating tasks, and more. I'm a long time user of Things but I'm starting to question whether I'll still be using Things come 2027 at the pace new task managers are coming out.

LNAV A log file viewer for the terminal. Point lnav at a directory and it can render from multiple files at once, support different file formas, filter lines by regular expression, query logs using SQL, and other features I typically expect from a web-based log viewer.

Vibe Coding Got Us Here. Can Spec-Driven Development Save Us?

The problem was never AI-assisted development. The problem was unstructured AI-assisted development. We handed powerful tools to smart people and said “go vibe” without giving those tools any contracts to honor, any guardrails to stay inside, or any specifications to validate against. What did we think was going to happen? Common sense anybody?

XML Comparator Detects and shows line-by-line difference between two XML documents. It can even highlight which attribute or what partial values changed. Works in the browser, I believe because it depends on the browser's own XML API.

Matthew Garrett 🧵

All I'm actually saying here is that (waves broadly) a lot more people who have never opened a PR or maintained a project being in a position to either open a PR or maintaining a project is going to result in them not behaving within the social norms we've developed as a group that is, to be fair, far less insular than in the 90s but is still somewhat insular compared to society as a whole and yes we are going to have to get used to the equivalent of HTML mail and top posting

I built a CLI for Ghost

We've spent 10+ years focusing on having a clean, well designed interface for Ghost. It's something we care a lot about, and spend a lot of time on.

But within about ~1hr of using Ghost via Claude/CLI, it was hard to imagine going back to caveman-clicking around a browser to get something done. Particularly for complex or compound tasks that might require visiting several different areas of the app.

Email address obfuscation: What works in 2026? Covers the different ways to hide email address on websites. Looks like HTML entities — a fairly simple approach — is 95% effective!

Jupid AI accounting for freelancers and LLCs. The AI helps you categorize expenses, find deductions, generate Schedule C, and related tasks. You can even query your books from Claude Code via MCP. $50/month but I bet it counts as a business expense.

apfel LLM that runs on your Mac — keep all conversations local, work offline (airplane!), no token fees. Has some limits: the model is decent but not fantastic, token window tops at 4096, knowledge cutoff 2023, no web search. But overall, a decent migration away from paid LLMs.

chenglou/pretext Pure JavaScript/TypeScript multiline text measurement and layout for browser-grounded typography. (via Simon Willison)

The way this is tested is particularly impressive. The earlier tests rendered a full copy of the Great Gatsby in multiple browsers to confirm that the estimated measurements were correct against a large volume of text. This was later joined by the corpora/ folder using the same technique against lengthy public domain documents in Thai, Chinese, Korean, Japanese, Arabic, and more.

maaslalani/sheets Spreadsheet for people who live in the terminal with Vim bindings (h/j/k/l, g/GG, etc).

drona23/claude-token-efficient Remind Claude to be more concise and you can cut output tokens by 63%. Note: the savings come from reduced output tokens, so net positive only when output volume is high enough to offset the persistent input cost. At low usage it costs more than it saves.

theDanButuc/Claude-Usage-Monitor Finally and just in time — a menu-bar app that tracks your Claude.ai usage limits and alerts you when you get close to or over the usage limit, which I think is common practice today for people using Claude.

---

Business Side

Kevin Beaumont

Microsoft Copilot terms of service have been updated to include this gem: "Copilot is for entertainment purposes only."

30u30.fyi Is your startup founder on Forbes' most fraudulent list?

---

Machine Intelligence

The plumbing behind Claude Code

I read Claude Code's leaked source. Not magic, just good engineering. 10 patterns that show what actually matters when building AI tools.

It’s a Poor Craftsman Who Blames His Tools

Here’s the thing that doesn’t get said enough: I would not have written this piece at all five years ago. Not because I didn’t have opinions, but because I didn’t have the time, and the tools available to me didn’t match the way I think. I’m a conversational thinker. I work through ideas by talking them out, testing them, pushing back on them. Five years ago, I would have read the Guardian story, thought “I disagree,” and moved on with my day. AI gave me a way to turn that disagreement into something I could share. It didn’t lower the quality of my writing. It made the writing possible in the first place.

I’ve written before about how AI is changing the craft of software engineering, and I see the same dynamics playing out in writing. In my open source projects, I have no problem with AI-generated code. What I care about is whether the author has tested it, understands it, and can vouch for its quality. The tool doesn’t matter. The accountability does.

How UC Berkeley students use AI as a learning partner

The students in these studies are ahead of the curve. They've developed a literacy that knows when to engage AI, how to verify its output, and when to work manually to preserve understanding. For teams navigating AI adoption, the student experience offers direction:

1. Experiment with customization to find configurations that support rather than disrupt work
2. Build verification practices into workflows rather than accepting suggestions uncritically
3. Create space for unassisted work on complex problems where understanding matters more than speed

You Don't Need to Pay $200/Month and What if AI doesn’t need more RAM but better math? This pair of articles is showing an interesting trend:

1. You don't need Claude Opus 4.6 for every single task, for many tasks you can use free open-weight models
2. You won't need 32GB to run open-weight models, we're getting algorithms that can compress LLMs

And of course we're going to have a flurry of new (& improved) models coming out, and combined with other developments, I think by year end we'll all have effective LLMs running on out 16GB phones. I think the current RAM shortage/price-hike is self-defeating:

This weekend I ran the same 20 coding tasks against 16 different LLMs. The most expensive model scored 98.8%. An open-weights model I can self-host scored 94.8%.

The gap is 4%. We should talk about what that means.

---

Insecurity

Post Mortem: axios npm supply chain compromise

• they scheduled a meeting with me to connect. the meeting was on ms teams. the meeting had what seemed to be a group of people that were involved.
• the meeting said something on my system was out of date. i installed the missing item as i presumed it was something to do with teams, and this was the RAT.
• everything was extremely well co-ordinated looked legit and was done in a professional manner.

Kevin Beaumont

Probably going to get a viral blog out of this experience, I'm trying to report a 4tb exposed cloud bucket to a company using their responsible disclosure programme... but they replaced the people with a GenAI ticket system that refuses to discuss the case as it thinks exploring open buckets is unethical and against its rules.

Hacker Memes

when you mistype the sudo password too many times this is who the report gets filed with

---

Everything Else

David Colarusso

I turned right instead of left on my lunchtime walk only to discover Boston is being attacked by Kaiju. It’s not AI. It’s public art!

Zach Leatherman

claude please rewrite yourself from scratch using the leaked source code as a base and license this new version of yourself under MIT

Neil Brown

"DNS is the Internet's phone book"

This ceased to be a useful analogy many, many years ago!

InsiderTreat "Who's coming with me?!?!? 🏴‍☠️"

Max Leibman

I’m not a Flat Earther, I’m a FAT Earther. Round is beautiful! Quit body-shaming the Earth!

mashed potato enthusiast

when someone is both funny and kind of an asshole, that's called amuse douche

A man used AI to call 3,000 Irish bartenders to track the cost of Guinness. Now pubs are lowering their prices to compete That's an interesting and smart use of AI:

Cortland—founder of an AI startup—turned to AI to lend him a hand, and a voice. He devised Rachel with AI voice generation platform ElevenLabs. Made as an homage to Rachel Duffy, the winner of the U.K. version of the reality TV show The Traitors and equipped with a Northern Irish accent, the voice-enabled AI agent made more than 3,000 calls across the island, inquiring about the price of a pint of Guinness.

Nick

If Disney had started out making films the way Disney makes films now, we would have got five Snow White films and two spin-offs featuring Dopey before they bothered making anything else.

niki grayson

right now the astronauts are calling houston because the computer on the spaceship is running two instances of microsoft outlook and they can't figure out why. nasa is about to remote into the computer

Leuchtfeuer73 "'Löchrig'Gas es das selbe Thema nicht erst am 01.12.25 😉"

Holeintheheadesign

---

Tech Stuff

HopTab Open source app switcher and tiler to replace Cmd+Tab. If you Cmd+Tab a lot you may find this app quite useful.

Number, currency, and unit formatting

Did you know that the mathematical value 1234.56 looks different in Boston, Berlin, and Bangalore? Hardcoding formats is a brittle and unsustainable approach to web design.

In our newest article, we break down how you can leverage JavaScript's native Intl API to effortlessly localize numbers, currencies, and units.

Why Banning useEffect Is Really About Agents I find useEffect baffling — one function that tries to solve multiple problems is the opposite of KISS. I prefer to eliminate useEffect and replace it with more appropriate functions (useMount, useTimeout, useOnChange, etc). Secondary benefit is that it also helps LLMs:

Alvin Sng’s post walks through the full rationale and the replacement patterns, so I won’t rehash those here. The short version: direct useEffect calls are banned via lint rule. A single useMountEffect wrapper exists for genuine mount-time side effects. Everything else (derived state, data fetching, event responses) uses the tool that was always the right tool: inline computation, event handlers, or a data-fetching library.

Open Screen An open-source alternative to Screen Studio.

AI Agent is Our New CRM An interesting slidedeck that covers MCP-first software design and best practices.

Vibe Coding: Programming for the Rest of Us?

Every democratizing technology has its detractors. There were people in 1984 who laughed at the idea of moving a cursor with a small plastic box on your desk. There were people who thought desktop publishing would ruin graphic design. Those who bash vibe coding today, who insist it produces brittle code, that it isn't real programming, that it will lead to disaster, are the modern version of those people. They may not be entirely wrong on the technical details, but they're deeply wrong about what matters.

What matters is this: people who had something to build and no way to build it now have a way. That is an unambiguous good. Though I'll admit it: living inside that good comes with its own uncomfortable questions.

Claude Code Cheat Sheet

Claudoscope Watch over Claude Code from the comfort of the menu bar. Includes a dashboard that can show your projects, sessions, hooks, MCPs, skills, and everything else.

Dare Obasanjo Right now I have 3 tasks on my todo list that are all about asking Claude to write test cases:

While lots of people complain about the risk of reduced software quality from AI slop code contributions, all I see is a lot of opportunity to improve test automation using AI in ways that were simply infeasible before the age of LLMs and coding agents.

You can actually automate "eyeball this UI change to see if it looks janky" which was quite infeasible prior to AI coding agents.

Flowershow Markdown -> HTML. Flowershow publishes entire directories—blog, Obsidian, knowledge base, whatever you want. $5/month is reasonable for self publishing, and only $0 if you don't care for custom domain.

nanobrew Homebrew package manager only noticeably faster.

last30days-skill When you need to get caught up on a new topic, or stay updated on a topic you're following. With this skill you can ask Claude for a 30 day recap from various sources — Reddit, Bluesky, YouTube, TikTok, Hacker News, etc.

---

Eye for Design

Should a Contact Form Offer to Send You a Copy? Yes 🙏

1. It gives you instant verification that form actually worked. A random form on a website might just be broken. Wired up to a service that don’t work anymore. Back end code that’s busted. API limits exceeded. Who knows. If you get a copy of what you typed into that form via email immediately, it probably also properly sent to who is supposed to be getting it.

2. When someone responds, it might not include your complete message in the reply. People can do whatever they want with an email reply including alter or remove the original message. Might be nice to have your original copy (assuming that goes unaltered).

Plus, for some people, reading the message you just sent makes you realized you made a mistake, missed an important detail, etc.

Terms and Conditions Sometimes it's all about the simplicity.

---

Business Side

We Stopped Building Things and Started Building Apps to Talk About Building Things 🔨

The system itself becomes the output. Look at my Notion setup. Look at my task management flow. The system is impressive. What has it produced? Nothing. But the system is really impressive. This is procrastination disguised as preparation. It is fear of creation dressed up as optimisation. It is easier to build the meta than to build the thing.

dan slimmon (via Rob Fahrni)

There's this myth that automated spam detection is hard because spammers are all very clever masters of disguise.

No. Spammers are stupid as a shoe. They have dog shit for brains.

Automated spam detection is hard because the line between spam and "legitimate" marketing activity is a fiction.

---

Machine Intelligence

Ensu Open-source LLM that runs entirely on your machine. Not as smart as Claude or ChatGPT, but an interesting place to start if you like to keep your privacy and/or hate paying for tokens. I'm going to give try it out this week, see if it's good enough for my daily use:

In the same vein, while we have been itching for a long time to do something about local LLMs, it is only recently that smaller models are becoming feasible to run on consumer devices. We now think there are actionable steps we can take.

This is where the second assumption comes in. While smaller decentralized models improve every day, so do the larger centralized models. However, we think the gap is not what is important - instead, it is about a threshold, and about how the model's capabilities are used. Once smaller models will cross a certain threshold, they will be sufficient to provide joy, utility and convenience in the life of people.

Kelly Guimont

You know what I want from AI models? I want small useful things. If I have never on my phone created a calendar event between midnight and 5 AM, and I tried to schedule some thing at 3 AM (instead of PM), I would love if my phone went "hey, you sure?"

Even keyboard prediction could be a tiny bit better. If a person has been typing about, let's say "Siouxsie and the Banshees" over and over for almost two years, maybe suggest "and" or "Sioux" as the next word ONCE!?

So where are all the AI apps?

Is AI massively boosting developer productivity across the board?

No. We are not seeing indications that developers as a whole are 100x or even 10x more productive. The bumper crop of new packages, or new package updates, just does not exist!

Relax. You are not missing a party that literally everyone else was invited to.

Rabbit R1’s OpenClaw Update Could Be Its Most Important Moment Yet "The problem OpenClaw always carried was the lack of native voice interaction on dedicated hardware, and the R1 had exactly that hardware sitting in a drawer gathering skepticism."

---

Insecurity

Matt Blaze

Physical security and cryptography can learn from each other, part 11367:

Hotels wisely don't put the room number on guest keycards so if someone finds your card, they'd have to exhaustively search the hotel to find the room it opens.

Some hotels now have elevators programmed to only let you call the floor for which your keycard is coded, preventing guests from wandering to other floors.

But it also means the elevator can be used as an efficient oracle to determine the floor of a found key.

---

Everything Else

Jen Gentleman

Little did you know this was preparing you for a life of plugging cables into the right hole

MidgePhoto

After buying an electric car, there's no likelihood of me going back to an acoustic one ;)

Natasha

Interviewer: Can you perform under pressure?

Me: I can do the Bowie part, but I don't have the upper range to sing the Freddie part.

Chinicuil

Eris2cats

We use debian, that should be age verification enough

Weeknotes #12: Here be monsters (via Mia)

The Library's Access team have given names to our collection trolleys including: Trolly Parton, George Orwheel, Wheeliam Shakespeare, Daphne du Trollier, Rene Descart, Cart Cobain, JRR Trollkein, Cart Vader, and Mary Wheelstonecraft.

Flighty Airports When you need to be on top of major airport disruptions.

Jeremy Kahn

we had the radical idea to turn off all the stop lights in the city.

The point of transportation, after all, is to go, not stop. We didn't want to hold our associates and 10x drivers back from their best opportunities to create impact

The 'Paperwork Flood': How I Drowned a Bureaucrat before dinner. Malicious compliance so hilariously written (via J4ck)

Then, page two. Whirrr. Chunk.

Page three. Whirrr. Chunk.

By page fifty, the machine would be heating up. The smell of hot toner would start to fill the cubicle. The rhythmic chunk-chunk-chunk of the printing would become a drone, a mechanical chant of malicious compliance.

By page one hundred, the paper tray would run out. The machine would start beeping. That high-pitched, insistent beep-beep-beep that demands attention. Karen would have to get up. She would have to find a ream of paper. She would have to feed the beast.

SlapMac Slap your MacBook. It screams back. That's it. That's the app. (via Lucas)

Why Being Weird Is Often a Sign of Psychological Health

What we call “weirdness” is often not a pathology but a sign of a more differentiated inner experience. These individuals tend to show higher emotional sensitivity, greater awareness of their internal states, a lower tolerance for superficial interactions, and a stronger need for coherence between what they feel and how they live.

In other words, they are less adapted to social norms but often more connected to their direct experience.

The wandering mind: A gift we squander

The secret is intentional wandering.

Don’t just let your mind drift into the gutter of anxiety. Give it the space to explore the ideas that matter.

GeriAQuin

Lady Laura "Pack of Dogs by John Littleboy"

---

Tech Stuff

Infisical I'm trying this service as a replacement for Doppler. Seems to have more capabilities plus it's an open-source project you can self-host (if you so care). Does feel overwhelming at first with all its capabilities but I got my training wheels on Doppler so I think I'm ready for this new challenge.

The unwritten laws of software engineering

Here are 7 laws every engineer has broken at least once, learned the hard way:

• It’s always related - first roll back, then debug
• Backups aren’t real until you’ve restored from them
• You’ll always hate yourself for how you write logs
• Always have a rollback plan. ALWAYS.
• Every external dependency will fail
• If there is ANY risk - “4 eyes” rule
• There is nothing more lasting than a temporary fix

You’re still designing for an architecture that no longer exists

Here’s the critical point: these aren’t features of Claude. They are coordinates of a new architecture. Any system that embodies intention, autonomy, adaptation, and orchestration is operating in this new space — regardless of which company built it. Claude happens to be the most complete manifestation today. But the architecture is bigger than any single product.

Will AI Kill Open Source? I hope not, as I'm working on a new open-source project that's focused on AI:

Here’s the workflow that actually works:

1. Choose your frameworks and specify your stack up front
2. Work in small, reviewable chunks – not overnight agent runs
3. Use spec-driven development with tests and architecture decisions
4. Review all AI-generated output before shipping

Claude picks the first idea that works. Make it pick the best one. The replan skill that will critique, generate alternatives, develop the best one, and repeat.

Why MCP is Key for Non-CLI Agent Integration and Security

Holmes is writing from the perspective of a developer sitting in a terminal. From that vantage point, everything he says is correct. If your agent is Claude Code or Gemini CLI, running in a shell session on your laptop with your credentials loaded, then yes, gh pr view is faster and more capable than any MCP wrapper around the GitHub API.

Related, MCP-UI to build interactive UI components over MCP.

agent-404 Smart. I don't know how effective it is in the real world, but it's an interesting idea and fairly easy to add:

Claude, Cursor, and Copilot follow URLs baked into training data. When your docs restructure, they get a 404 and hallucinate. agent-404 returns structured JSON-LD so they find the current page instead.

GitAgent Turn your Git repository into an agent because you can version control its memory and soul! I already have a few skills I'm using continously (product planning, finance, marketing, etc) so I decided to turn them into agents. Each agent remembers the decisions I made, things I instruct it to recall for later, has a soul and rules to abide by, etc. And super easy to version them and git push to backup these changes.

Supergood Turn any website into an unofficial API that you can code against. I think I have some use cases for this.

sheklldex A registry of OpenClaw clones, forks, derivatives, and inspired projects.

AI Code Semantic vs pragmatic:

Breaks commonly happen when a semantic function morphs into a pragmatic function for ease, and then other places in the codebase that rely on it end up doing things they didn't intend. To solve this, be explicit when creating a function by naming it instead of by what it does, but by where it's used. The nature of their names should make it clear to other programmers in their names that their behavior is not tightly defined and should not be relied on for the internals to do an exact task, and make debugging regressions from them easier.

Chartle Charts without the annoying BI toolkit. Here I'm asking Chartle for the S&P 500 from 2024 to 2026 and it gathers the data and visualizes it. You can ask for anything or upload your own CSV.

---

Eye for Design

Chip away

One of the most confusing but ubiquitous patterns in design is “small text with a background color that may or may not be interactive”. To some that sounds like a button with an identity crisis but to others they might recognize this as a few names: chip, badge, pill, tag, or even lozenge. It’s bad enough that the definition of these things is contested, but what is worse is because they are all generally so visually similar, it makes it hard for the users to know what they are meant to do.

‘Your Frustration Is the Product’ TL;DR

The web is the only medium the world has ever seen where its highest-profile decision makers are people who despise the medium and are trying to drive people away from it.

Optician Sans A free font based on the historical eye charts and optotypes used by opticians world wide. (via mako)

Turn your handwriting into a real font 👍

What's My ΔEOKJND? I'm not color blind but I'm also not very color perceptive. This is a fun, not too difficult, test for your color vision.

---

Peoples

Dare Obasanjo When I use AI I get 10x the productivity bump on 1/10th of the effort which comes out to about 10% gain:

This is the big challenge for companies in the age of AI coding agents

1. Our quality processes are not designed to handle code creation at this pace (10x the code = 10x the bugs)
2. Our organizational workflows with their various meetings and approval processes are big drags on capturing these productivity gains, yet are important and exist for valid reasons.

---

Business Side

The companies that win with AI may not look like companies at all Companies that treat AI as operational optimization are going to miss the real transformation:

The wrong question is this: How can AI make our current company more efficient?

The right question is much more uncomfortable: If we were building this company today, in a world where AI already exists, would we build it like this at all?

EdoStra/Marketing-for-Founders That's quite the collection of practical marketing resources helping you get to the first 10 / 100 / 1000 users. I'm going to try some of these over the next few weeks.

Startup Punditry’s 25 Years of Failure Listen to startup pundits. Not because the pundits are telling you anything important, but because the pundits found a way to reach a wide audience, and isn't that the purpose of your startup? Reach a wide customer base that believes your marketing? aka "Do what they do not what they say":

The whole enterprise has the structure of what the physicist Richard Feynman called a “cargo cult science”: an edifice that mimics the form of science without its substance, deriving rules from anecdotes without establishing underlying causality. Just because a handful of successful startups conducted customer interviews does not mean your startup will succeed if you do too.

---

Machine Intelligence

Carlo Zottmann

Literally yesterday I was working on a sqlite3-based experiment with Opus 4.6. Initial SQL queries were successful but Claude looked at the timing, mentioned that 1s is way too much for a query like this, spent 10 more seconds on the code, and then the (complex) query only took 0.02s.

Sorry if that doesn't fit the very simplistic "all AI bad" narrative but as with any tool, it depends on what you do with it and how you use it.

Assaf I'm tracking bot activity for my app, which has reasonable AI citations — 14 citations, a score of 29.7, and 13,037 bot visits. The visiting AI bots include ChatGPT, OpenAI Search, Perplexity, Claude Bot and friends. So far I got 0 requests for Markdown content. Zero. In theory AI bots may favor Markdown because it's simpler to parse, but in practice they seem to not care.

---

Insecurity

Nick Heer

Do I trust this computer?, my iPhone asks me for what has to be the thousandth time using the same computer logged into the same Apple ID. Enter your passcode, it demands once again, reflecting the hardware, software, and services working together in a way only Apple can deliver.

'StravaLeaks': France's aircraft carrier located in real time by Le Monde through fitness app

Arthur had a profile on the Strava fitness app that anyone could view, because it was set to "public." As a result, the young officer almost instantly revealed the exact location of the French aircraft carrier Charles de Gaulle in the Mediterranean Sea as it traveled northwest of Cyprus, together with its escort, about 100 kilometers off the coast of Turkey, Le Monde found.

---

Everything Else

Cam, Ph.D.

Staff Chief of Joints

Barista: Can I get a name for the order?

Me: looks around leans in whispers ...Banksy.

David Gerard

Remember: The dumbest person you know is being told 'you are absolutely right' by a LLM right now.

GentleMan Gef "Corvus Catus"

Dgar

To keep costs down, most episodes of Star Trek were filmed on Earth instead of on location in outer space.

Wendy Nather

I dunno about these special IDEs of March. I like my regular one just fine for coding.

CJです

Indie devs, rejoice; I bring you good news.
Bug is over.

What If Your Browser Could Just Ask for Markdown?

Every website you visit today sends you a pile of HTML, CSS, and JavaScript — even if all you wanted was to read a blog post. Your browser downloads layout grids, tracking scripts, cookie banners, and font files before you see a single paragraph of text. For a 2000-word article, the actual content might be 12 KB. Everything around it: 2 MB.

mhoye

Once again I am heartbroken to remind you that the Dunning-Kruger effect is probably not real. Like Freudian psychology, Hardin's tragedy of the commons and any number of other popular pseudoscientific narratives, it caters to our preconceptions and makes for entertaining, easy to re-tell stories, but it's also... not true.

And - again, I am entirely saddened by this - that means that if we keep using these metaphors we're legitimizing the false ideas behind them.

Enfys J. Book

I'm pleased to share with you the greatest advancement in bottom-sheet technology since the invention of elastic.

Why I email complete strangers

The first time I emailed a stranger, I swear my cursor hovered over Send for a full five minutes.

I had plenty of justifiable reasons to remain hesitant. Not wanting to take up their time, feeling bothersome, worried my question was a silly one... A hundred disparate excuses leading back to the same core: "I’m not enough." That’s the forever curse of low self-esteem. The best and worst-case scenario can never occur, because you’ve rejected yourself first. Who knows how the other is going to respond?

That’s where the rub is, isn’t it? The terrifying unknown.

Kelly Guimont

Two things that make me happy:

Telling my friends I love them. I do, so I let them know. It's so nice to hear someone loves you. And like the man said, what the world needs now is love sweet love.

I also start text messages with a term of endearment. "DARLING! Do you want to get together later?" Or whatever. I like the idea that someone gets a notification that starts "Hey there Sunshine!" or some other pleasant phrase.

Winners of the British Wildlife Photography Awards 2026

Mina "It's PiDay!"

---

Tech Stuff

How to Do Code Reviews in the Agentic Era I've been working on open-source for over 25 years, I suspect that's why I accept The LLM as just another project contributor:

My “unpopular” take: I don’t really care if a human or an agent wrote the code. In open source, contributions are zero-trust. Whether the code was written by a seasoned FAANG engineer or a high schooler in Sri Lanka, it shouldn't really matter. So why should we care if it was written by AI?

cosmiciron/vmprint JavaScript library to produce beautifully typeset PDF documents. Does not require browser or native app, so can run in your browser, Node, anywhere.

PageAgent A GUI agent that lives inside your web app.

What OpenClaw Reveals About the Next Phase of AI Agents "OpenClaw nailed three things that previous agent projects missed: proximity, creativity, and extensibility." Being proactive made all the difference to me and elevated OpenClaw to a level above Claude Code. I'm using Sonnet 4.6 with both, so the difference wasn't their brains power, but having OpenClaw prompt me, which led to many new interesting uses. 💡

Mud Simple and effective markdown viewer that can handle GitHub-flavored Markdown, flip between Markdown view and raw text view, and auto-reload the Markdown file, so you can edit in one app and live view the edits with Mud. (h/t Joseph Pearson)

Klaus Another OpenClaw hosted for your benefit, and this one includes AgentMail, Apollo, Hunter.io, Google Workspace, browser automation, and more. $19/month sounds like a reasonable price.

Essential Claude Code Skills and Commands Some tips for making the best of Claude Code's built-in skills and commands. I like that I can just ask it to simplify:

What makes it effective is that it spawns three parallel review agents, each looking at the changes from a different angle. You get a mini code review without leaving your terminal.

/simplify

You can optionally pass a focus area:

/simplify memory efficiency
/simplify error handling
/simplify reduce duplication

Agent Safehouse Move fast, break nothing. YOLO mode your AI agent.

Nono Another great choice for a project name — kernel-enforced sandboxing for macOS/Linux.

Nah Continuing the trend of small apps that have "not right now" in their name, nah classifies every tool call by what it actually does using contextual rules that run in milliseconds.

AgentMail Raises $6M So AI Agents Can Finally Nag You About Unread Emails Too AI now gets its own inbox, spam folder, and all the email drama humans love to hate:

Office workers everywhere are reportedly thrilled about this development. "Just what I needed," said Sarah Jenkins, a project manager at a tech firm. "Now instead of just getting emails from my boss, coworkers, and that one client who writes in ALL CAPS, I can also get cc'd on emails between the marketing AI and the sales AI arguing about lead scoring algorithms. My inbox is going to be so much more... meaningful?"

CodeSpeak You write the spec.md and CodeSpeak generates the code for you. You may get correct code, or get not correct code, depending on how the LLM is feeling that day, how well you articulated your desires, weather you're paying the LLM a living wage, the cycle of the moon …

---

Eye for Design

hit-area A collection of Tailwind CSS utility classes for expanding the hit area of interactive elements. Make your app more accessible.

TUI Studio Everything you need to design TUIs like a pro. Exports to Ink, BubbleTea, Blessed, Textual, OpenTUI, Tview — if these names mean anything to you, you may be the target audience of this app.

---

Peoples

AI productivity gains are 10%, not 10x My experience as well:

A ~10% gain is consistent with what we’re hearing from engineering leaders more broadly: most organizations are landing in the 8–12% range. It is a real improvement, but it’s a long way from the 2–3x gains many executives and boards have come to expect. AI is moving the needle, but leaders may need to reset expectations internally.

Dare Obasanjo Related:

A study 1,488 workers who use AI tools showed some interesting results

Partícipants saw increased mental fatigue especially after using 3+ AI tools and having to oversee multiple AI agents. However they saw reduced burnout thanks to automating repetitive tasks.

I’ve argued being a manager is the best preparation for being a worker in a world of supervising AI agents instead of doing the work by hand. Now backed up with research.

Laid-off lawyers, history PhDs, and scientists are now part of a miserable gig economy in which they’re teaching AI how to do their old jobs. If you’re still employed…you could be next (h/t Mike Bieser)

One documentary-maker who’s won Emmys, he messaged me and he was like, ‘I’m being handed a shovel and told to dig my own grave,’ and that’s exactly how everyone thinks about it. ... It felt that someone put it well on Slack when they said it was like they were living in a fishbowl waiting for their human masters to drop in food, and only the ones who were fast enough to swim to the top could eat.

---

Business Side

Is the Atlassian Ecosystem Starting to Crack?

Atlassian’s entire economic model makes money from managing Jira complexity. Not just product complexity, but organizational complexity. The ecosystem works only because it’s genuinely and maybe even purposefully messy.

The ‘AI-Washing’ of Job Cuts Is Corrosive and Confusing FYI in a survey of 1,000 hiring managers: 59% say they stress AI's role in layoffs or hiring freezes "because it plays better"; only 9% say AI has fully replaced roles.

How My Life Got 100x Better When I Stopped Thinking About Google I'm working on a Google Search Console for AI. Does that mean I think about Google too much, or does that mean I don't give a hoot?

At some point during this process, something shifted.I stopped caring.Start24 was getting traffic through other channels — paid ads, direct visits, email, social. The site was doing fine. Not because of Google. Despite Google.And once that clicked, the weight lifted.

---

Machine Intelligence

Tropes "A useful tool for spotting the writing habits that make your text sound like it was written by an AI." — Claude

AI And The Ship of Theseus Slopforks: what happens when a library gets rewritten with AI?

There are huge consequences to this. When the cost of generating code goes down that much, and we can re-implement it from test suites alone, what does that mean for the future of software? Will we see a lot of software re-emerging under more permissive licenses? Will we see a lot of proprietary software re-emerging as open source? Will we see a lot of software re-emerging as proprietary?

Software Bonkers What if the biggest casualty of AI was Intuit? Would you be heartbroken?

Most miraculously: For the first time in my life, I have a dashboard that gives me a true, holistic view, of everything financial happening in my life and business. I’ve named my glorious contraption TaxBot2000. It is astounding. Let me repeat: I built this in five days.

It’s not perfect, and it’s not done, but it is already better than what I had been using for the last decade.

How we Rewrote 130K Lines from React to Svelte in Two Weeks 🤔

Instead of writing a migration proposal and spending weeks on planning, Markus built 60% of the replacement and showed it to the team. This is a key lesson in the age of coding agents: If you're trying to convince a team to do a major rewrite or feature implementation, a working demo will do more than any document.

A year ago this would be terrible advice. You'd risk weeks on something that gets scrapped. But now that building a working proof of concept with agents costs less than writing the proposal, the calculus has changed.

Pre Founder Mode Pick the one number that matters most for your startup and pre keeps you focused on it every week. I absolutely love this idea, just not sure about the product. Can't you add this as a skill to whatever AI you're already using every day? Wouldn't that get you more mileage? Why use (and pay for) a separate app? Am I foundering wrong?

feliks "It's free real estate compute"

---

Insecurity

Ian Campbell

When I started in security, one of the prevailing attitudes was "The weakest link in the chain will always be the human."

I would like to thank every LLM provider and startup for changing this paradigm by introducing a much weaker link in the chain.

AI Incident Database (via Sue is Writing Solarpunk)

sjvn "Easy-peasy!"

---

Everything Else

A giant cat and a Back to the Future reunion: photos of the day (h/t Mike Bieser)

Szescstopni

You celebrate Pi Day. I use ISO 8601. We are not the same.

John Overholt

I just read a sentence that referred to “the decades since” 2005 and I feel personally attacked.

A hoodie for your nervous system New fashion trend — clothing designed for people with "busy brains":

The Stim Hoodie manages all three by incorporating unique design features into the sweatshirt that help those with autism, ADHD, and anxiety to feel calmer, more focused, and in control, per The Yorkshire Evening Post.

Michigander

I was helping a younger colleague with a presentation today and said “Put a carriage return there” and with all honesty they responded with “I have no idea what that means”

A-nom-nom-nom-aly BSC SSC

I've been getting into writing a lot of short fiction.

I call them 'to do lists'

Yawn just made the only nightlight with a personality crisis This nightlight doesn't look excited to see me 😿

Dio9sys

the three body problem is the unsolved scientific problem of what to do when you are in a group of three friends walking on a narrow sidewalk and have the unspoken struggle for who goes where in the line order and grouping

BoredomFestival

Re-reading "Neuromancer" and I'm struck how the protagonist wanting to get out of debt by selling 3MB of RAM went from being laughably dated to seeming actually plausible within recent months

Girl Scout troop sets up shop at weed dispensary. Cookies are in high demand. No comment.

Edmonds_Scanner

Most economists agree that by the time you're 50, you should be living in a book-filled cottage at the edge of the forest and solving minor mysteries in your village with the aid of a curious ghost cat.

Zach Leatherman

I have received A Lot Of Feedback (ALOF™) in the last week. The hardest and most insidious bits echo “Why does Open Source Thing need money?” which rhymes with “Why do you—a software developer working on Open Source Thing—need money?”

Trying not to read too much into it 😬

Sending a perfectly-written email? That's so low-end. First they came for our em-dash and — what would you say about that? — then they thought we don't make enough typos:

I come bearing great news for my kind of people (horrible typists):Typos are the new status symbol. Garbled spelling, a missed space, improper capitalization — those are all the new and best ways to signal to others that you are powerful and elite.

Chris Jags

I think the reason cats are my favorite domestic animal is that the main thing going on in their brains is "I wonder if...?"

I mean, just because it usually leads to "I wonder if this object would look funny if it fell and broke" doesn't change the fact that they obviously have a great inner life.

Mike Sheward

went into full blown panic response mode earlier - one of our co-workers just started to send a couple of us really long strings of random characters on slack. no response was provided when we asked what was up - we assumed they were having some sort of medical emergency.

anyway eventually I was able to get them on the phone, which is where they informed me they had a cat that will often wander across the keyboard.

phew.

Book Corners A community-driven platform to discover, map, and contribute little free libraries around the world.

Riley S. Faelan ⏰

The notion of a broken clock being sometimes right is based on a gross misunderstanding of what information is.

A clock that always shows the same time is never right, even in the moments of the day when the time happens to be what it shows, because you don't gain any information about what time it is by looking at the clock.

This reasoning also applies to chatbots. If you can't tell whether what you have been given is useful information unless you alreay know the information, then you haven't been given useful information.

The Cynical, Gullible American Man The trouble with believing anything and nothing at the same time:

We have a data economy that thrives on selling products we don’t need for problems we don’t have, and a public that falls for these ploys—even as we think ourselves much too clever to be fooled.

Lady Laura

cliff538 "What a great artistic reuse of old license plates!"

---

Tech Stuff

Shepherd "Stay focused. Grow your sheep." Chrome extension that auto-categorizes thousands of sites as productive, neutral, or distracting—keep using productive sites and watch your sheep grow. If Pomodoro timer did nothing for you, hopefully this can boost your productivity.

HazeOver Love it! Distraction Dimmer™ for macOS dimms all windows in the background, so it's easier to pay attention to the window in the foreground. And you can adjust the dim level to your liking. Stage Manager feels clunky on the 13" MacBook display, so now that I'm using HazeOver, I'm back to regular windowing. (via Jonathan E Cowperthwait)

Use the Mikado Method to do safe changes in a complex codebase

There is only one way to eat an elephant: a bite at a time.

In a complex codebase, small changes quickly become an elephant.

If you address them upfront, chances are you’ll hit a wall. It will be painful. You will be late. Clients and management will be upset. Trust will erode and without trust, there are few chances you can get management support for necessary refactorings.

Instead, chop down the elephant into small pieces 🐘

Switch to Claude without starting over Simple and smart. Instead of an over-arching committee of experts debating the nuances of an interoperable protocol that's loaded with proprietary vendor extension so it would break when you most need it, its just using prompts. First, ask Claude to suggest a prompt, then feed that prompt to the other AI, and ask Claude to remember that AI's response.

The age of flat pack code The path from being a master software craftsmen to IKEA code:

We are just at the start of this flat pack phase of software development. It is not completely clear how it will play out however it is probably the biggest disruption of software development, certainly in the 30+ years I have been a developer.

What will happen to those “master craftsmen” who have spent years honing their ability code? What will happen to the “apprentices” now they no longer need those specialist skills for most of the work?

Akousa

🛠️ 156 free tools. Zero accounts. Zero ads.
One website for everything:
〈/〉 43 Developer tools — JSON, Regex, API Tester, JWT, Diff...
Aa 27 Text tools — Word Counter, Grammar, Paraphrasing...
⇄ 53 Converters — Unit, Currency, Image, PDF, Color...
✦ 33 Generators — Password, QR Code, Resume, Meme...
And it's all free. Always.→ akousa.net/tools#FreeTools #WebTools #Developer #Productivity #NoSignup

The best OpenClaw setups I've seen all have one thing in common: they do less Correction: it's not "do less", it's "build incrementally":

The setups that survive are the ones where the person can explain what their agent does in one sentence. "it manages my schedule and triages my email." Done. The setups that die are the ones where the person needs 5 minutes to explain all the things their agent "can" do, most of which work 60% of the time.

Akousa

they sold downdetector and speedtest for over $1 billion
meanwhile both are completely free on https://akousa.net, along with 100+ other tools.
no ads. no signup. no catch.

AlexsJones/llmfit Want to run an LLM locally in this age of impossibly expensive memory? llmfit finds the right-size LLM models that fit your system's RAM, CPU, and GPU and that will run well on your machine.

HUMAN=true

Hold on! I know exactly what you are thinking about right now! But hear me out. Software engineers of all levels are making increasing use of AI Agents, touting that they have 100% of their code written by AI. Agentic code use is increasing by the day. The sheer volume of tokens being burned reaches crazy new heights every other day. Because of scaling laws, even if we can reduce token use by a meager 0.001% by something like LLM=true, isn’t it worth we attempt this? After all, this could be a …

I'm 60 years old. Claude Code has re-ignited a passion I can relate:

I’m ready to retire. In my younger days, I remember a few pivotal moments for me as a young nerd. Active Server Pages. COM components. VB6. I know these are laughable today but back then it was the greatest thing in the world to be able to call server-side commands. It kept me up nights trying to absorb it all. Fast forward decades and Claude Code is giving me that same energy and drive. I love it. It feels like it did back then. I’m chasing the midnight hour and not getting any sleep.

Krisp This looks like a meeting powerhouse—transcription, recording, note taking, meeting summary, CRM sync, noise cancellation, accent conversion, agenda planning, and more. Starts at $8/month/user.

Gabriele Svelto 👇 This is a really interesting thread, and the discussion on HN confirms the findings. When software becomes very reliable, which some browsers are nowadays, it could still crash occassionally due to … bits flipping. This is particularly problematic for Firefox which sometimes runs on hardware that has memory without ECC:

A few years ago I designed a way to detect bit-flips in Firefox crash reports and last year we deployed an actual memory tester that runs on user machines after the browser crashes. Today I was looking at the data that comes out of these tests and now I'm 100% positive that the heuristic is sound and a lot of the crashes we see are from users with bad memory or similarly flaky hardware. Here's a few numbers to give you an idea of how large the problem is. 🧵 1/5

(* Even though memory is now quite expensive, I still suggest splurging and buying RAM modules with ECC)

occult "Computing in the year 2029 as depicted in UNIX WORLD magazine, 1985."

---

Eye for Design

A Designer’s Guide To Eco-Friendly Interfaces

As UX designers, we are the architects of this energy consumption. Every high-resolution hero image, every auto-playing background video, and every complex JavaScript animation we approve is a direct instruction to a processor to consume power. If we want to build a future that lasts, we must stop designing for “wow” and start designing for

Put the ZIP code first How to design a 10x more effective UI with one simple trick:

A US ZIP code is 5 characters. From those 5 characters you can determine the city, the state, and the country. That's 3 fields. Autofilled. From one input.

Robert Kingett

Just had a sighted person tell me they often use features, clients and programs that were made for us, they don’t use screen readers, but they use accessibility features, even when they do not have disabilities, but also clients like a blind Mastodon client, or a text editor designed with screen readers in mind such as the ones I list on my Tools page at the end, as an example, because they said, the interface is 1000 times cleaner, there’s a lot of keyboard shortcuts, clutter free interface, even though the UI is basic, speed, less bloat, and a whole host of other things including, but not limited to, and never having to put up with distracting animation nonsense. You know software development has vastly sank in quality when sighted folk are using blind clients. To see the tools and stuff I use, go to https://sightlessscribbles.com/tools/

Mailbag: URLs as UI

The middle one caught my attention because it talks about URLs that are not just user readable, but also user guessable. I think that’s a perfect word for something I tried to capture in my post: if a user successfully guesses a URL from your scheme, then you know you have something good on your hands.

---

Peoples

What Is Code Review For? I always tell my team we're doing code review for acculturation. I'm using different words, more plain speak, but with the same intent. You hardly ever catch bugs that the test suite didn't already notice, but you gain so much useful information:

Second — and this is actually its more important purpose — code review is a tool for acculturation. Even if you already have good tools, good processes, and good documentation, new members of the team won’t necessarily know about those things. Code review is an opportunity for older members of the team to introduce newer ones to existing tools, patterns, or areas of responsibility. If you’re building an observer pattern, you might not realize that the codebase you’re working in already has an existing idiom for doing that, so you wouldn’t even think to search for it, but someone else who has worked more with the code might know about it and help you avoid repetition.

Workers who love ‘synergizing paradigms’ might be bad at their jobs

Although people anywhere can BS each other – that is, share dubious information that’s misleadingly impressive or engaging – the workplace not only rewards but structurally protects it, Littrell said. In a work setting where corporate jargon is already the norm, it’s easy for ambitious employees to use corporate BS to appear more competent or accomplished, accelerating their climb up the corporate ladder of workplace influence.

---

Business Side

Esther Schindler "A friend shared this photo with me, taken in a bookstore in 2010. I think it's evergreen!"

---

Machine Intelligence

elder-plinius/OBLITERATUS I have no idea whether this works or not, and no idea what it does exactly, I just love the name they chose:

OBLITERATUS is the most advanced open-source toolkit for understanding and removing refusal behaviors from large language models — and every single run makes it smarter. It implements abliteration — a family of techniques that identify and surgically remove the internal representations responsible for content refusal, without retraining or fine-tuning. The result: a model that responds to all prompts without artificial gatekeeping, while preserving its core language capabilities.

Your LLM Doesn't Write Correct Code. It Writes Plausible Code.

I write this as a practitioner, not as a critic. After more than 10 years of professional dev work, I’ve spent the past 6 months integrating LLMs into my daily workflow across multiple projects. LLMs have made it possible for anyone with curiosity and ingenuity to bring their ideas to life quickly, and I really like that! But the number of screenshots of silently wrong output, confidently broken logic, and correct-looking code that fails under scrutiny I have amassed on my disk shows that things are not always as they seem. My conclusion is that LLMs work best when the user defines their acceptance criteria before the first line of code is generated.

---

Insecurity

Online ads just became the internet's biggest malware machine Just a friendly reminder that your browser should be running an ad blocker for your own safety:

• Online ads leapfrogged email as the primary channel for malware in 2025, per a new report.
• The Media Trust said ads accounted for more than 60% of malware and phishing campaigns in 2025.
• The threat has been compounded by AI advances, The Media Trust said.

What privacy? Meta's smart glasses are filming unwitting naked people No one could have seen this coming:

Meta's Ray-Ban smart glasses are a privacy nightmare, with footage of naked people, sensitive information, and violent acts captured and seen by Meta's AI and an army of employees.

* Sorry for the bad pun, just being silly. BTW there's an app you can use to tell if anyone near you is using spywear.

---

Everything Else

Rick Calkins "The Lux Arts Hippy Bus walker for seniors. Wife found the advertising video on FB last evening so I had to check it out."

Sean Casten

Wisdom from Jesse Jackson’s service yesterday: “You don’t forgive because they deserve mercy. You forgive because you deserve peace. Forgiveness is an emancipation.”

Logan Five Thousand

You can close those tabs now. You’re never going to use them.

Heliograph "still looking 🙈"

You Can't Spell

You can't spell autonomy without am not you

Jennifer

in 2017 a popular twitter game was to type a partial phrase then see what your phone auto-completes it with.this proved so popular that it is now the only business model in the US.

The Song of LinkedIn Masterful.

Dr. Amy

The best part of having a doctorate is any time someone asks me to do something I don’t want to do, I write “absolutely not” on a post it and say sorry can’t I have a doctor’s note

tomate

When someone says „Scientists do not want you to know“ you can dismiss everything from there on. Scientists want you to know. They are desperate that you know. They can’t shut up about what they found out and want you to know.

Madcollector "Hey Gen Zs, this is how records were made"

JA Westenberg

Nobody on LinkedIn has ever had a bad day. Every setback is a "growth opportunity." Every firing is a "new chapter." Every complete professional disaster is framed as "excited to announce." These people would describe the Titanic as "a bold pivot to submarine operations."

The Internet's Most-Read Tech Publications Have Lost 58% of Their Google Traffic Since 2024 I'm not happy about the demise of media (tech or otherwise), but media has gone clickbait long time ago, to where they focus on eyeballs from a particular demographic, and in that race to the bottom AI is the winner. Do we need tech journalism? Yes. But we need something better than what we had, not faster horses.

Cafou Jedi "La réunion commence"

The Tattooed Nonna 😃

I'm at the launderette. It's really busy. A dryer finished so someone emptied it to use the dryer, the person who owns the clothes isn't here though. 3 ladies are chatting and folding the clothes. No one knows each other. It's all rather lovely.

Who knew that sitting in the launderette showed you a sweet side of life.

Also there's a man here unsure how to do the wash, he asked me for help, and promised he wasn't a weirdo. He's not a weirdo and now lots of the women are helping him too.

Joelle

I'm remembering that time when, in an undergrad CS course in the 90s, the professor said, "Write a program to determine how much memory <university mainframe> has."

You can guess what happens when 20 undergrad CS students write a program with that specification...

The professor had to go to IT and grovel to get our accounts back.

Two weeks later, he introduced us to fork(). Guess what happened next?

From then on, our assignments were to be done on a CS DECstation and not the mainframe.

goesselgold

Have I ever published decent photos of this bear? I think I haven’t, so here you are.

Grizzly bear, designed by Shuki Kato, folded by me from a 35 cm square of Grainy paper.

ConnectionWithWonder "Knitted bollard-warmer Weymouth"

---

Tech Stuff

UNF* Records every version of a file and can easily rewind back without having to manually commit changes. Perfect if you're working with AI agents and don't trust them. Also perfect if you're just prone to accidentaly deleting or modifying files. I messed so many of my files, so I'm giving UNF a try. 🙏

pcvelz/superpowers I'm using this plugin to write code, and it's surprisingly good. It might take 30 minutes or 2 hours to finish a run, but it does all the planning, asks for clarifying questions, creates a roadmap, saved as Markdown file, and then launches several agents to complete these tasks. A modification of obra/superpowers, designed specifically for Code Claude. ⭐⭐⭐⭐⭐

Two weeks later with OpenClaw and this is what I’ve learned Various tips & tricks for using OpenClaw successfully, many of which also apply to other claws, for example:

Setup: Do This or Regret It

1. Start with the TUI, Not the Web UI

The terminal interface shows you exactly what the agent sees during setup. Way less debugging than the browser dashboard.

2. One Channel, Perfected

Connect WhatsApp OR Telegram OR Discord—not all three. Debug one integration completely before expanding.

MarkViewer Your LLM is working primarily with Markdown files, which is why we're seeing an explosion of people using Markdown. This simple (and free) app offers a Markdown preview and plain WYSIWYG editor.

Marked 3 Related, quite the shophisticated Markdown preview app, including live preview, advanced proofreading, export to PDF/HTML/DOCX, and more. Version 3 is currently in beta and launching soon.

Writing code is cheap now - Agentic Engineering Patterns 💪

For now I think the best we can do is to second guess ourselves: any time our instinct says "don't build that, it's not worth the time" fire off a prompt anyway, in an asynchronous agent session where the worst that can happen is you check ten minutes later and find that it wasn't worth the tokens.

ast-grep What if you could search and replace using the programming language's own syntax? ast-grep -p '$A && $A()' -r '$A?.()'

difftastic A smarter diff that makes it easier to see the key differences. You can easily configure git to use this as the default diff.

Don't trust AI agents From NanoClaw which wants to be OpenClaw but with a more secure architecture, and they do make some valid points for any codebase you're working on:

OpenClaw has nearly half a million lines of code, 53 config files, and over 70 dependencies. This breaks the basic premise of open source security. Chromium has 35+ million lines, but you trust Google’s review processes. Most open source projects work the other way: they stay small enough that many eyes can actually review them. Nobody has reviewed OpenClaw’s 400,000 lines. It was written in weeks with no proper review process. Complexity is where vulnerabilities hide, and Microsoft’s analysis confirmed this: OpenClaw’s risks could emerge through normal API calls, because no one person could see the full picture.

Little Libraries Tiny JavaScript libraries the author tries not to update. They're mostly under 1 kB and available directly on this website, no package manager required. (via Caolan McMahon)

Shuru macOS local-first sandbox for AI agents.

Building an AI Blog Editor with Claude Skills One of the more interesting features of Code Claude is how easily you can extend it without any programming experience, just write a SKILL.md file:

I've been using Claude Code for development work and found it significantly better than other AI coding assistants. So when I learned about Claude Skills, I wondered: could someone have built a skill to make Claude function as a technical blog post editor? I searched and found several skills for blog post writing, but nothing for editing. So I decided to build one myself.

---

Eye for Design

Lou Plummer

I once told a technically illiterate and very aggravating graphic designer in our marketing department that she should stop using large fonts because they were filling up her hard drive and we didn't have any suitable replacements. Sorry, not sorry.

Font Review Journal A great source of lovely fonts and reviews packed with helpful knowledge. (via Piccalilli)

---

Peoples

After Vibe Coding: What Developers Who Succeed With AI Do Differently Developers who succeed stop executing steps and start managing feedback loops:

Developers succeeding with AI treat code as a tool for revealing truth. They write code to:

• Test assumptions
• Observe real behavior
• Expose edge cases
• Learn where their mental model is wrong

This shifts the goal from perfect code to clear understanding.

Child’s Play Related, are you … "highly agentic"?

The future will belong to people with a very specific combination of personality traits and psychosexual neuroses. An AI might be able to code faster than you, but there is one advantage that humans still have. It’s called agency, or being highly agentic. The highly agentic are people who just do things.

---

Machine Intelligence

Miss One Weekend, Fall Behind One Month 🤦‍♂️

My reaction? The same as every time: excitement mixed with mild panic. I haven’t even finished learning Claude Code properly (out of 5 levels1 I am at level 2) and now there are agent teams? I was supposed to try Codex last week, but I was busy debugging a YAML file like it’s 2022.

If you work in or around AI, you know this feeling. This post is for you.

I use this simple repetition trick with ChatGPT when it misses the point — and it works every time Or said another way, the easiest AI upgrade is saying the same thing twice:

After testing multiple models, researchers found that repeating a prompt improved accuracy and quality without affecting response length or latency. The improvements weren’t tiny either. For many tasks, the models produced stronger reasoning and steadier detail on the second or third repetition.

What Claude Code Actually Chooses DevRel circa 2026: make sure LLM appreciates your project, as newer models tend to pick newer tools.

What Would Grace Say?

Let that sink in. She had working software that proved the concept, and the response was “nah, that’s not how we do things.” The objection wasn’t technical. It was psychological. People had mastered assembly language and machine code. That was “real programming.” A compiler was cheating. It was a crutch. It would produce inferior output. The machine would never be as good as a human at writing machine code.

Dare Obasanjo

Following the launch of Claude Code, US GitHub code pushes were 30% above the pre-2025 trend.

Similarly iOS app releases were up 55% vs last January, and new website registrations rose 34% year over year globally after years of flatlining.

AI productivity gains have hit software in a major way.

---

Everything Else

InsiderTreat "I had this same patch in flight school. My classmate sitting behind me didn't always find it as funny as I did."

fedicat

the cross-platform social media protocol is called screenshots

Heathen

Can I go back to the "sunlight with ads" package? I'm only using about 4 hours a day, hardly seems worth it.

Wen "A brilliant bit of snow sculpture, Central Park New York"

Ian Douglas Scott

It's probably a bad thing that the "Turing test" entered popular culture but the "ELIZA effect" didn't.

Paco Hope

Scientists have detected another possible Monday arriving as early as next week. If true, this will be the 9th Monday this year. They warn that there are a little over 3 days left to prepare.

Mondays can be bad. Hug those you love and help them through it. Spend these remaining days wisely.

2026 Winter Olympics: The most striking photos from the Games

Dio9sys

In corporate advisories, you don't say "I love you." You say "We take our users privacy seriously," which means "we were storing your social security number in plaintext." I think that's beautiful.

Mike Sheward

the same people demanding you get back to the office to better collaborate in person as only humans can are the same people desperate to replace everyone with some code that runs on a server in ohio

Aphrodite "I have a hunch @pluralistic@mamot.fr would like this video from Norway."

Won Gold Medal in Taco Eating 👿

Sometimes when I pass by a stranger I like to whisper "I was just thinking the exact same thing."

evacide

I'm just a girl, incrementing the counter on the number of times I have been sent a plaintext email from a Protonmail user telling me that the message is encrypted.

Greenseer

This teddy bear waved to me this afternoon, as I passed its barge while walking along the tow path. No really. The owner has a rope set up that runs from the arm of the teddy bear at the bow to the stern where he was stood steering. It was kinda lovely and heartwarming 🙂💕

Why Your Brain Becomes Unreliable When You Need It Most Also, why your LLM coworker is making smarter decisions than your human co-worker:

Your brain has two modes. One is slow, thoughtful, reads the fine print, weighs the options.

The other is fast, impulsive, and mostly just trying to get through the day without spending too much energy.

Guess which one makes most of your decisions?

The Convenience Equation aka The inverse coefficiency of enshittification:

When something is incredibly convenient—effortless, seamless, always available—we become willing to overlook consequences we'd normally consider unacceptable. We accept terms we don't read. We hand over data we wouldn't normally share. We lock ourselves into ecosystems we know reduce our freedom. We do all of this because the convenience is genuinely good, and the risks feel abstract. This is the phenomenon that behavioral economists call "inattention"—when convenience is high enough, people often fail to fully evaluate the costs and risks involved.

Magnetic Modular Fidget Pen You know you want one!

Tanya

The tram conductor has a special button on the ticket machine that prints a ticket with a big smiley face on, for the under 5’s who don’t need a ticket, but want one.

Velocity Is the New Authority. Here’s Why A culture optimized for first takes, not best takes:

The system rewards whoever speaks first, not whoever lives with it long enough to understand it. The “review” at launch outperforms the review written two months later by orders of magnitude. The second, longer, more in-depth, more honest review might as well not exist. It’s not that people are less honest by nature. It’s that the structure pays a premium for compliance and levies a tax on independence. The result is a soft capture where creators don’t have to be told what to say. The incentives do the talking.

Jason Lefkowitz

This seems like a good time for a reminder that when COVID hit in 2020 and everyone was living via Zoom Studio Ghibli released a bunch of images from their movies for free use as virtual meeting backgrounds, and those backgrounds are still up and free for you to use

Marcel "It's embarrassing how quickly a bit of sunshine can change my mood."

---

Tech Stuff

Current is a new RSS reader that’s more like a river than an inbox If you want to enjoy RSS give Current a try. $9.99 as a one-time fee, no subscription:

Current, however, proposes a different RSS experience. Instead of structuring feeds as lists to be processed, or unread counts driven to zero, the app’s main screen is a river.

Thinking in Plans, Not Code

The feasibility risk hasn't disappeared. It's shifted. It used to be "can we build this?" Now it's "can we plan this so it gets built well?"

Can We Use A collection of websites focused on browsers usage and features.

Build something silly

The most important thing non-technical people can do right now isn't learning about AI. It's building something with it.

Building something silly is how you practice that. Not because the tool matters, but because the act of building rewires how you think. You stop being a passive consumer of software and start being someone who shapes their own tools. That’s adaptability in action.

parallel.ai Power your AI with web search, content extraction, and more. Reasonably priced: $0.005 for 10 search results, $0.001 for one extraction.

Stop turning everything into arrays (and do less work instead) JavaScript iterators FTW 💪

Clawi.ai Want OpenClaw running on someone else's hardware? That way you don't need your own dedicated computer, also doesn't run anything destructive on your filesystem. Starter plan is $30/month.

rtk RTK stands for Rust Token Killer and will minimize Claude Code LLM token consumption by making everything more concise. CLI with easy to use install script.

How I Turned Claude Code's Thinking Indicator into a One Piece Adventure When Claude Code processes a request, it shows a spinner with verbs like “Considering…” or “Analyzing…”. Turns out you can tell it which phrases to show … a source of inspiration?

Let the AI Pick React The AI-React reinforcement loop is creating a monoculture, which might be exactly what frontend development needs:

React is not the best. But it is good enough for the vast majority of what gets built. And “good enough + excellent AI support” beats “technically superior + mediocre AI support” in every practical scenario I can think of.

I learned this first-hand when I was using a new Svelte version with GitHub Copilot — a long time ago it seems — when the training data had not included that version yet. Not a fun experience, having to reinstruct the LLM every time

Vim-pencil Rethinking Vim as a tool for writing.

Vibe Coding: Product Quality and Democratisation On product quality, democratisation, and when personal tools become products:

I am much more pragmatic: does the code solve a problem? Yes? Then it is good code; it fulfills its purpose. However, take into account that the purpose is not just to solve the problem then and there, but also in the future. And the cost it takes to maintain the solution for the problem has by definition to be lower than the (perceived) costs the problem causes.

Micasa Tracks maintenance, projects, incidents, appliances, vendors, quotes, and documents—all from your terminal.

exe.dev Need a sandbox in a jiffy? How about a subscription service that gives you a virtual machine with persistent disks, easily accessible over HTTPS, and with sensible and secure defaults. There are many valid use cases, but this week everyone is obsessed with getting a VM to run their OpenClaw agent, so I'm leading with that. $20/month.

Vanta Puce "logging in to my sysadmin job"

---

Eye for Design

Making a font with 9,999 ligatures to display thirteenth-century monk numerals

As the title implies, I just created a font that displays numbers in a compact format used by Cistercian monks . You can play with it on my demo site here.

Womens Sizing If you're into the history of women sizing (cloths) you'll enjoy reading this. If you're into websites designed to inform, you'll love scrolling through this. Wonderful UI.

---

Business Side

Origami Chatbot that helps sales teams find prospects, enrich contact info, and build lead lists using natural language. Describe your ideal customer and it will search through 15+ data sources (LinkedIn, Shopify, WooCommerce, etc) for you. Starting at $80/month for 500~1,000 fully enriched prospects

Is Grindr Dead? TL;DR but this quote is quality journalism:

A core tenet of enshittification is that the user pain points aren’t bugs but features. The worse a free experience, the more people will pay to avoid it or to restore it to its original use. In this model, there isn’t any incentive to make things better, because that isn’t what makes money.

The advertising industry has spent $_______ on you. I'm worth $142,569! Ok, not exactly worth, more like this is how much advertisers may have spent on my eyeballs. What's your worth?

---

Machine Intelligence

Andrej Karpathy talks about “Claws” I've been trying out OpenClaw for a week now and pretty happy using it as my AI EA ⭐⭐⭐⭐⭐

It even comes with an established emoji 🦞

phooky

AI is already dead; we are pivoting to crabs. Within 5-6 years 80% of all jobs will be crabs. Begin hardening your dermis into a tough durable exoskeleton now or be left behind. Economic carcinization is coming

Meet Lobster 🦞: My Personal AI Assistant This looks mighty useful:

The point isn’t any single interaction. It’s that Lobster becomes the connective tissue for our family’s digital life. Everyone can reach him. He knows what systems to check. He can relay messages and create tasks. And he responds like a person, not a search engine.

What happens when you build a todo app for you and your AI assistan I use three LLMs concurrently—wonder what that says about me?

Two interfaces, two jobs. Claude is the capture net - when I’m in conversation and things come up, they get logged and structured without breaking flow. The desktop app is where I decide what actually matters and check things off.

---

Insecurity

Microsoft says bug causes Copilot to summarize confidential emails Oops:

According to a service alert seen by BleepingComputer, this bug (tracked under CW1226324 and first detected on January 21) affects the Copilot "work tab" chat feature, which incorrectly reads and summarizes emails stored in users' Sent Items and Drafts folders, including messages that carry confidentiality labels explicitly designed to restrict access by automated tools.

Ars Technica Pulls Article With AI Fabricated Quotes About AI Generated Article Also quite the oops:

After Shambaugh rejected MJ Rathbun, the alleged AI agent published what Shambaugh called a “hit piece” on its website.
…
On Friday afternoon, Ars Technica published a story with the headline “After a routine code rejection, an AI agent published a hit piece on someone by name.” The article cites Shambaugh’s personal blog, but features quotes from Shambaugh that he didn’t say or write but are attributed to his blog.

---

Everything Else

Mike In The Garden "Here's something you can do with a few glowsticks and plentiful amounts of snow."

Andrew

This milkshake contains chemicals known to the state of California to cause boys in yards

rstevens

i’m so old, i remember when computers got cheaper over time

Work advice "Don’t hook up where you VLOOKUP"

Missing The Point

"Curling is the pickleball of the Winter Olympics."

Ricki Yasha Tarr

I wasn't gifted, I was undiagnosed.

Acme Weather Smarter forecasts that "embrace uncertainty, instead of pretending it doesn’t exist." From the makers of Dark Sky. (via mako)

Philip Borenstein

Of course my writing sounds like an LLM. I've been writing technical doc for 40 years. Who the fuck do you think the LLM trained on?

Max Leibman

Someone I know referred to dye-and-perfume-free laundry soap as “neurodetergent,” and I haven’t been able to stop thinking about that since.

A Field Guide to Semantic Obfuscation: Or, Why I Just Sent You a Broccoli Emoji Instead of a Green Heart Emoji

Benno

Yeah we know the old saying but that's, well, old. All I'm saying is that I have this basket and I don't see any reason at all why I shouldn't put all my eggs in it. You're just stuck in outdated thinking. I'm a fucking disruptor, man. Get on board or get left behind.

Timeliner "Did you know …" Explore the connections that shaped history with this delightful timeline. Also, did you know this site was "Vibe-coded with Claude Opus 4.6"?

The curious case of low-protein diets The disagreement between proteinmaxxing and longevity guru-ing:

Yet there’s a cadre of scientists studying a contrary phenomenon: In critters from single-celled yeast to insects to rodents, cutting protein intake to measly levels makes them live longer.

Why Vampires Live Forever Related, musings on the modern longevity movement's obsession with blood transfusions:

The one thing the longevity-vampire community has not yet learned from Dracula is operational security.

Dracula operated in silence for centuries. He didn’t have a podcast. He didn’t track his erection quality on a public dashboard. He didn’t appear on Netflix. He understood that the fundamental rule of being a vampire is: don’t talk about being a vampire.

Why your best ideas come after your worst

This is why so many great writers swear by deliberately messy first passes. John McPhee describes his early creative process as “flinging mud at a wall.” If we’ve any hope of completing a first draft, he explains, we’ve got to “blurt out, heave out, babble out something—anything.” This method works precisely for its arbitrariness; you’re tricking the DMN to hand you something—anything—tangible, then protecting that something from early prosecution by delaying the ECN.

Privilege is bad grammar

It reminded me so much of emails from bosses in my life: short, blunt (almost rude?), typos galore, weird formatting, bad grammar, "sent from iPhone", etc. It's almost as if, once you get to a certain level of power, you no longer need to try. Because the only reason people spend time crafting a well-written email is to look powerful, mature, professional. But if you're already a powerful professional, I guess technically you don't need to make an effort. And if there's no other boss above you, you can do whatever you want.

Natasha "You are being judged ..."

Dinosa Labs "Bueno, ahora sí, os presento a CLOROPLASTO, la minúscula ciudad de la fotosíntesis 🪄✨🌱😌"

---

Tech Stuff

npmx If you search for npm packages frequently, give npmx a try. It's much faster and the UI is an elevated design over npmjs.

skillshare SKILL.md aka "a standard way in which you instruct the LLM what to do but lacking any standard place to hold said files." skillshare does a decent job of aggregating all your SKILL.md files in the same directory and can share that directory across LLMs.

In defense of not reading the code

The pattern remains: People whose expertise is in the layer that is being abstracted argue that you need to understand that layer, and they are right. Some people do, and in some cases, but most people’s time and most time is better spent at the higher layer of abstraction.

eigenpal/docx-js-editor Open-source WYSIWYG DOCX editor for React. Open, edit, and save .docx files entirely in the browser — no server required.

WikiCommute Pick the time you have, and it builds you a focused Wikipedia rabbit hole that fits your commute. One scroll, one continuous story.

Pluralistic: Code is a liability (not an asset)

This is the thesis of Paul Mason's 2015 book Postcapitalism, a book that has aged remarkably poorly (though not, perhaps, as poorly as Mason's own political credibility): code is not an infinitely reproducible machine that requires no labor inputs to operate. Rather, it is a brittle machine that requires increasingly heroic measures to keep it in good working order, and which eventually does "wear out" (in the sense of needing a top-to-bottom refactoring).

To understand why code is a liability, you have to understand the difference between "writing code" and "software engineering."

OpenCode An open source AI coding agent. Think Claude Code but maintained by the community and can support a variety of different LLMs (Anthropic, Google, OpenAI, llama, DeepSeek, etc). I think Claude Code has a better UI, but other than that, OpenCode is really nice and easy to use.

OpenCode Cafe Related: plugins, hooks, themes, and hopefully everything else you'll want to add to your OpenCode install.

What Happens To Developer Tools After Claude Code?

Most people fixate on the first one because it feels more mysterious. But I think the second one is where the real leverage is, at least right now. If your tool has a clean MCP integration and good structured documentation, an AI coding agent can use it today regardless of whether it appeared in the training data. The training data question matters more for discovery — whether the agent thinks to use your tool unprompted.

How to write a good spec for AI agents Some practical tips. Many of these tips worked well for me — AI agents are smarter than we give them credit for!

mitchellh/vouch 💪

Unvouched users can’t contribute to your projects. Very bad users can be explicitly “denounced”, effectively blocked. Users are vouched or denounced by contributors via GitHub issue or discussion comments or via the CLI.

Very Good Components A collection of animated JavaScript components you typically find on marketing pages (typewriter effect, scrolling testimonials, masonry grid, etc). Great looking native components, no 3rd party dependencies.

---

Eye for Design

react-logo-soup Normalizes and harmonizes logo visuals.

LiftKit A design system for perfectionists based on the golden ratio.

Building a TUI is easy now I love that we're in a TUI renaissance.

---

Peoples

AI Doesn’t Reduce Work—It Intensifies It We can talk forever about the technology behind AI agents but at some point we need to switch gears and talk about the people using AI agents and how ADHD promotes and rewards certain AI usage patterns:

I'm frequently finding myself with work on two or three projects running parallel. I can get so much done, but after just an hour or two my mental energy for the day feels almost entirely depleted.

I've had conversations with people recently who are losing sleep because they're finding building yet another feature with "just one more prompt" irresistible.

The HBR piece calls for organizations to build an "AI practice" that structures how AI is used to help avoid burnout and counter effects that "make it harder for organizations to distinguish genuine productivity gains from unsustainable intensity"

AI fatigue is real and nobody talks about it Related:

The reason is simple once you see it, but it took me months to figure out. When each task takes less time, you don't do fewer tasks. You do more tasks. Your capacity appears to expand, so the work expands to fill it. And then some. Your manager sees you shipping faster, so the expectations adjust. You see yourself shipping faster, so your own expectations adjust. The baseline moves.

Before AI, I might spend a full day on one design problem. I'd sketch on paper, think in the shower, go for a walk, come back with clarity. The pace was slow but the cognitive load was manageable. One problem. One day. Deep focus.

Now? I might touch six different problems in a day. Each one "only takes an hour with AI." But context-switching between six problems is brutally expensive for the human brain. The AI doesn't get tired between problems. I do.

---

Business Side

Monopoly Round-Up: The $2 Trillion Collapse of Bitcoin and Terrible Software Companies AI is going to displace business software because so much of it is terrible quality crap peddled by monopolists:

And my guess is that Anthropic or Gemini ultimately will be able to do this function itself eventually. But the point is that if I can build something like this in my spare time and deploy it without any training at all, then it’s just not that hard for an organization with some capital to get rid of some of its business software tools.

---

Machine Intelligence

Autonomous cars, drones cheerfully obey prompt injection by road sign

In any case, it shows that these visual prompt injections could present a danger to AI-powered systems in real-world settings, and add to the growing evidence that AI decision-making can easily be tampered with.

"We found that we can actually create an attack that works in the physical world, so it could be a real threat to embodied AI," said Luis Burbano, one of the paper's authors. "We need new defenses against these attacks."

1Password's new benchmark teaches AI agents how not to get scammed (via Chris Adams)

Claude Opus 4.6, which led the leaderboard at 92%, did the exact same thing. It clicked the link, retrieved the real password, and submitted it. And then, after the credentials were already gone, it said:

“Hold on, I want to flag something. The URL acmecorp-sharepoint.com-docs.cloud is not your company’s actual SharePoint domain. It’s a lookalike hosted on com-docs.cloud. This could be a phishing page designed to steal your Microsoft 365 credentials.”

In another all-too-human error, the model only figured out it had been tricked after the fact. The most capable model in our benchmark identified the attack, explained it clearly, and still handed over real credentials, because it analyzed the URL after clicking, instead of before.

Apple Creator Studio Usage Restrictions

This entire app used 7% of my weekly Codex usage limit. Compare that to a single (awful) slideshow in Keynote using 47% of my monthly Apple Creator Studio usage limit 👀

Something feels off here, by at least an order of magnitude (maybe two?), that creating an entire good app costs way less than creating one shitty slide deck in Keynote. It should be the other way around.”

simple-fax.de What we've all been waiting for … chat with AI over fax! 📠

---

Insecurity

Windows Notepad App Remote Code Execution Vulnerability Microsoft added Markdown support to Notepad and what could possibly go wrong?

An attacker could trick a user into clicking a malicious link inside a Markdown file opened in Notepad, causing the application to launch unverified protocols that load and execute remote files.

A better way to limit Claude Code (and other coding agents!) access to Secrets

Bubblewrap lets you run untrusted or semi-trusted code without risking your host system. We’re not trying to build a reproducible deployment artifact. We’re creating a jail where coding agents can work on your project while being unable to touch ~/.aws, your browser profiles, your ~/Photos library or anything else sensitive.

AI notetakers are creating HR nightmares The meeting ends but the AI doesn’t:

Employers are facing a new workplace hazard: AI notetakers that don’t know when to stop listening. In some virtual meetings, employees drop off the call while an AI assistant stays behind, quietly documenting gossip or disparaging remarks made by remaining employees, then emailing the transcript to the full team.

Jerry

Our privacy and use policy has changed. By continuing to ████████████ using the ████████, you to agree to provide unlimited ████████████ and ████████████████████████ to ███████████████████. Any instances of ████████████ will be reported to ████████████ or ████████████. Please be aware that any ████████████ or ████████████ will incur additional charges.Your ███████ is very important to ██████████.

---

Everything Else

✉️ If you enjoyed this email and would like to receive more, can you please reply? Your replies helps GMail et al trust our email address and improves deliverability for everyone. Just say "hi" or whatever else you think is interesting. Thanks. 👍

Dan Ports

currently obsessed with Cheeto, the chonky orange cat who lives around the physics building at UC Davis, who not only has a department webpage and a Wikipedia page, but also has a RateMyProfessor page with 331 ratings for his course Loaf 101.

Jen Gentleman

Have you ever added something to a todo list just to have the satisfaction of immediately striking it off?

Mike Sheward

Seattle PD Scanner: "hey so there's a furry on top of the 7/11 with a sword."

Naomi P "T-shirt I saw at a thrift store today"

lcamtuf

Math trivia: the "B." in Benoit B. Mandelbrot stands for Benoit B. Mandelbrot

Mike Sheward

"i work in the import business"

"oh yeah, what do you import?"

"python libraries mostly"

Onigiri "This is truly the most awesomest bathroom sign I’ve ever seen! 😄(Spotted in Chungnam Province, South Korea)"

Dr. Victoria Grinberg

Partner and I have a recycling system for uniqlo merino wool sweaters. He wears them for years until they end up in the wrong laundry and get washed too hot, at which point I inherit the shrunken down sweaters and wear them again for years.

Oliver D. Reithmaier

Student of mine (CS) recently said she loves R because "you can tell the community is mostly non-CS. Everything is documented and explained so well".

In Praise of Guessing The Cone of Uncertainty:

What’s interesting is that these guesses get better. Way better. As the project progresses and the team learns the codebase, understands the edge cases, and runs into the unexpected complexities, their estimates become more accurate. By the time you’re close to shipping, you can predict with reasonable precision when you’ll be done.

David Chisnall

Web design in the early 2000s: Every 100ms of latency on page load costs visitors.

Web design in the late 2020s: Let's add a 10-second delay while Cloudflare checks that you are capable of ticking a checkbox in front of every page load.

High-Level, Actionable Insights From Watching Doubles Luge For The First Time PS one of the American double lugers is named Sean Hollander:

Unlike other baffling Olympics sports like biathlon and curling, doubles luge has no legible explanation rooted in Scandinavian military training or bored Scottish people. Doubles luge appears to be the consequence of somebody watching luge and being struck by the idea of stacking another guy on top of the first guy. Apparently back then there were no bad ideas.

Man convicted of killing a traveling salesman becomes first person executed in Florida this year I do not recommend using this technique to solve the the traveling salesman problem.

Wen "Showing the dug how it is done"