2025: The year in LLMs

Blog posts, photos, and micro updates

Key Points\ The public perception of Large Language Models (LLMs) is shifting from "superintelligence" to standard software infrastructure. While the novel...

Books, writings, projects, AI tips, and personal reflections from 2025.

2025年里ai coding有一波非常大的能力增长，随着模型能力/agent工具/IDE整个上下游能力的变化，我的使用策略也在不断调整，所以写一篇blog记录下今年我的工具箱变化。视角很不专业，纯粹是一个使用者角度体验（belike 使用评测）。

Talking fast and swearing more since 2004.

The community keeps framing MCP and Skills as competitors. They're not. They solve completely different problems.

Talking fast and swearing more since 2004.

Standout articles I read this year

In this post, I look back on all the stuff I did in 2025

Blog about things

MCP became the default way to connect AI models to tools faster than security could keep up. What's actually involved in deploying MCP servers safely.

There is a messy reality of giving AI agents tools to work with. This is particularly true given that the Model Control Protocol (MCP) ha...

AI became a teammate in 2025. From startups back to academia, advisory, and a summer of full-time AI experimentation.

A Perplexity vs Claude vs Mistral vs Gemini vs ChatGPT competitive intelligence face-off

2025 was a blockbuster year for both AI and critical minerals! Pretty much every other week either we got news of a new discovery or governmental initiative, or an AI company releasing a new model …

2025 was a turning point for many developers in their relationship with AI, with the emergence of agents — and that's probably what we'll remember most.

I’ve accumulated hundreds of Claude Code sessions and kept losing track of where I solved specific problems. The built-in /resume shows recent sessions, but anything older than a few days? Gone. So I built CCS (Claude Code Search) to fix that. TLDR Link to heading brew install agentic-utils/tap/ccs ccs Type to search, Enter to resume. That’s it. The problem Link to heading Claude Code stores conversations in ~/.claude/projects/ as JSONL files. Each session gets its own file:

As I start writing this year in review, the first thing that comes to my mind is this holiday memory I have (and many other nerds have 🤣) from my youth: Going to see Lord of the Rings in theatres.

Source: https://simonwillison.net/2025/Dec/31/the-year-in-llms/ 2025 in LLMs (Simon Willison) — clustered 1) Capability shift: “reasoning” becomes the new default RLVR/inference-scaling drives 2025 progress: not necessarily bigger base models, but longer RL runs and “reasoning modes/dials.” The practical unlock isn’t puzzles—it’s tool-using planning, especially for research/search and debugging gnarly code. 2) Agents become…

Link collection

Doug Belshaw has a thoughtful post about the financial/work advice often being given to young folks today: The financial guidance given to ...