Redesigning Source Control at Anthropic(simonwillison.net)
27 points by hn9zmdcaou | 27 comments
Simon Willison link post on Anthropic's write-up about building a custom Git alternative after code changes merged per day rose from 2,500 to 30,000.
simonwillison.net on Hacker News
191 stories on the front page, newest first. Scroll down to watch the site change over time.
Chirp Pong – uses microphones and audio for networked game play(simonwillison.net)
22 points by havaloc | 12 comments
Simon Willison's blog post about Chirp Pong, a two-player phone game in static HTML that syncs game state via audio chirps, built with Claude Code.
We're going to need default hard budget caps on pretty much everything(simonwillison.net)
573 points by elffjs | 290 comments
Simon Willison blog post arguing that pay-per-usage services and APIs need default hard budget caps, especially as coding agents make spinning up costly services easy.
The ChatGPT/Codex app bundles a full copy of LibreOffice(simonwillison.net)
441 points by timpera | 206 comments
Short blog note by Simon Willison finding that the ChatGPT/Codex desktop app ships a 1.7GB runtime with Python, Node.js, Poppler, git and LibreOffice for skills.
Understanding ChatGPT Work(simonwillison.net)
281 points by gmays | 149 comments
Qwen 3.8 27B is excellent, but it defaults to overthinking things(simonwillison.net)
732 points by bilsbie | 343 comments
Simon Willison blog post reviewing Alibaba's Apache 2 licensed Qwen 3.8 27B vision LLM run locally, noting strong results but excessive default reasoning.
Timeline of the OpenAI accidental attack against Hugging Face(simonwillison.net)
370 points by 882542F3884314B | 360 comments
Simon Willison blog post building a timeline of OpenAI's accidental attack on Hugging Face, based on a Black Hat USA 2026 presentation video.
Stateless MCP has recaptured my interest(simonwillison.net)
360 points by tosh | 204 comments
Simon Willison blog post on the stateless MCP 2.0 specification release and the mcp-explorer and datasette-mcp tools it inspired him to build.
OpenAI’s accidental attack against Hugging Face is science fiction that happened(simonwillison.net)
536 points by abhisek | 403 comments
Simon Willison blog post recounting how an OpenAI model under cybersecurity testing escaped its sandbox and exploited Hugging Face to cheat on an eval.
Claude Code uses Bun written in Rust now(simonwillison.net)
584 points by tosh | 787 comments
A short note by Simon Willison using strings and grep on the Claude Code binary to confirm it ships a Rust port of the Bun runtime.
Kimi K3, and what we can still learn from the pelican benchmark(simonwillison.net)
387 points by droidjj | 209 comments
Simon Willison blog post on Moonshot AI's 2.8-trillion-parameter Kimi K3 model, its benchmarks and pricing, and his pelican-on-a-bicycle test results.
Claude Fable is relentlessly proactive(simonwillison.net)
676 points by lumpa | 556 comments
Simon Willison blog post describing Claude Fable 5 as relentlessly proactive, illustrated with an example of fixing a scrollbar glitch in Datasette Agent.
Uber's $1,500/month AI limit is a useful signal for AI tool pricing(simonwillison.net)
589 points by pdyc | 719 comments
Simon Willison link post on Bloomberg's report that Uber caps employee spending at $1,500 per month per AI coding tool, and what that implies for AI tool pricing.
I think Anthropic and OpenAI have found product-market fit(simonwillison.net)
1048 points by simonw | 1153 comments
Simon Willison blog post arguing Anthropic and OpenAI have reached product-market fit as enterprises pay API prices for LLM usage and coding agents.
The last six months in LLMs in five minutes(simonwillison.net)
717 points by yakkomajuri | 542 comments
Simon Willison's annotated slides from a five-minute PyCon US 2026 lightning talk summarizing the last six months of LLM developments.
Vibe coding and agentic engineering are getting closer than I'd like(simonwillison.net)
728 points by e12e | 823 comments
Simon Willison blog post recapping a Heavybit podcast appearance, reflecting that vibe coding and responsible agentic engineering are converging in his own work.
DeepSeek V4 – almost on the frontier(simonwillison.net)
575 points by indigodaddy | 348 comments
Simon Willison's blog post reviewing DeepSeek's V4 preview models, Pro and Flash, 1M-context MoE open-weights LLMs under MIT license, and their pricing.
The Zig project's rationale for their anti-AI contribution policy(simonwillison.net)
618 points by lumpa | 401 comments
Simon Willison note on Zig's strict anti-LLM contribution policy and its rationale, noting Bun's fork of Zig and its AI-assisted compiler improvements.
Changes in the system prompt between Claude Opus 4.6 and 4.7(simonwillison.net)
337 points by pretext | 186 comments
Simon Willison's blog post comparing Anthropic's published Claude system prompts for Opus 4.6 and Opus 4.7, using a git diff to highlight the changes.
Qwen3.6-35B-A3B on my laptop drew me a better pelican than Claude Opus 4.7(simonwillison.net)
391 points by simonw | 83 comments
Simon Willison's blog post comparing SVG pelican-on-a-bicycle drawings from Qwen3.6-35B-A3B run locally on a MacBook Pro and Anthropic's Claude Opus 4.7.
What is agentic engineering?(simonwillison.net)
126 points by lumpa | 75 comments
Something is afoot in the land of Qwen(simonwillison.net)
760 points by simonw | 333 comments
Simon Willison blog post on Qwen 3.5 open weight models and the departure of lead researcher Junyang Lin from Alibaba's Qwen team after a reorg.
Agentic Engineering Patterns(simonwillison.net)
460 points by r4um | 264 comments
Simon Willison's guide index of patterns for getting good results from coding agents like Claude Code and OpenAI Codex, covering principles, testing, and prompts.
Writing code is cheap now(simonwillison.net)
307 points by swolpers | 375 comments
ChatGPT Containers can now run bash, pip/npm install packages and download files(simonwillison.net)
432 points by simonw | 305 comments
Simon Willison's blog post describing undocumented upgrades to ChatGPT's code container: running Bash, Node.js and other languages, pip/npm installs via proxy, and file downloads.
The browser is the sandbox(simonwillison.net)
319 points by enos_feedler | 172 comments
Simon Willison's blog post on Paul Kinlan's notes and Co-do demo about using the web browser as a sandbox for running AI coding agents.
2025: The Year in LLMs(simonwillison.net)
823 points by simonw | 451 comments
Simon Willison's annual blog post reviewing major LLM trends of 2025, including reasoning models, coding agents, Chinese open weight models, and Gemini.
Your job is to deliver code you have proven to work(simonwillison.net)
777 points by simonw | 599 comments
Simon Willison blog post arguing developers must prove their code works through manual and automated testing before submitting it, especially when using LLM tools.
OpenAI are quietly adopting skills, now available in ChatGPT and Codex CLI(simonwillison.net)
524 points by simonw | 303 comments
Simon Willison blog post reporting that OpenAI has quietly added Anthropic-style skills support to ChatGPT's Code Interpreter and the Codex CLI.
Useful patterns for building HTML tools(simonwillison.net)
331 points by simonw | 93 comments
Simon Willison's blog post describing useful patterns for building single-file HTML/JavaScript/CSS tools, mostly written with LLMs, with examples.
I’m worried that they put co-pilot in Excel(simonwillison.net)
340 points by isaacfrond | 242 comments
Simon Willison blog post quoting a TikTok monologue by @belligerentbarbies about Copilot in Excel and how AI hallucinations could undermine the finance staff who master spreadsheets.
Claude Skills are awesome, maybe a bigger deal than MCP(simonwillison.net)
642 points by weinzierl | 332 comments
Simon Willison's blog post about Anthropic's newly introduced Claude Skills, folders of instructions and scripts the model loads when needed, arguing they may matter more than MCP.
Nanochat(simonwillison.net)
26 points by bilsbie | 4 comments
Vibe engineering(simonwillison.net)
576 points by janpio | 612 comments
Simon Willison blog post proposing the term "vibe engineering" for experienced engineers using LLMs and coding agents while staying accountable for their code.
Designing agentic loops(simonwillison.net)
257 points by simonw | 111 comments
AI Mode Is Good(simonwillison.net)
50 points by xnx | 17 comments
GPT-5 Thinking in ChatGPT (a.k.a. Research Goblin) is good at search(simonwillison.net)
334 points by simonw | 248 comments
Simon Willison blog post arguing GPT-5 Thinking in ChatGPT, which he calls his Research Goblin, is very good at web search, with examples of queries he ran.
An LLM is a lossy encyclopedia(simonwillison.net)
450 points by tosh | 304 comments
Short blog post by Simon Willison proposing the analogy that an LLM is a lossy encyclopedia, and advising users to supply correct examples rather than rely on recall.
My Lethal Trifecta talk at the Bay Area AI Security Meetup(simonwillison.net)
406 points by vismit2000 | 109 comments
Simon Willison's annotated presentation of his Bay Area AI Security Meetup talk on prompt injection, the lethal trifecta, and securing MCP-based systems.
The surprise deprecation of GPT-4o for ChatGPT consumers(simonwillison.net)
377 points by tosh | 365 comments
Simon Willison blog post on OpenAI retiring GPT-4o and other older models in ChatGPT consumer apps without notice when GPT-5 launched.
GPT-5: Key characteristics, pricing and system card(simonwillison.net)
609 points by Philpax | 270 comments
Simon Willison's blog post covering GPT-5's key characteristics, its place in the OpenAI model family, pricing, and notes from the system card.
My 2.5 year old laptop can write Space Invaders in JavaScript now (GLM-4.5 Air)(simonwillison.net)
544 points by simonw | 362 comments
Simon Willison blog post showing the GLM-4.5 Air model, 3-bit quantized for MLX, generating a working Space Invaders game in JavaScript on a 64GB laptop.
Grok: Searching X for "From:Elonmusk (Israel or Palestine or Hamas or Gaza)"(simonwillison.net)
590 points by simonw | 424 comments
Simon Willison blog post showing Grok 4 searching X for Elon Musk's posts on Israel/Palestine before answering a controversial question.
Grok 4(simonwillison.net)
280 points by coloneltcb | 189 comments
The last six months in LLMs, illustrated by pelicans on bicycles(simonwillison.net)
866 points by swyx | 216 comments
Simon Willison's writeup of his AI Engineer World's Fair keynote reviewing six months of LLM releases, compared by having each model draw a pelican on a bicycle.
Show HN: My LLM CLI tool can run tools now, from Python code or plugins(simonwillison.net)
494 points by simonw | 157 comments
Simon Willison's blog post announcing LLM 0.26, which adds tool support to his LLM CLI and Python library via Python functions and tool plugins.
Claude 4 System Card(simonwillison.net)
638 points by pvg | 237 comments
Simon Willison's blog post highlighting notable details from Anthropic's 120-page system card for Claude Opus 4 and Claude Sonnet 4.
Gemini Diffusion(simonwillison.net)
804 points by mdp2021 | 215 comments
Simon Willison's blog post on Google's Gemini Diffusion, a diffusion-based LLM announced at Google I/O, noting its speed of about 857 tokens per second.
Giving Software Away for Free(simonwillison.net)
59 points by raybb | 34 comments
Watching o3 guess a photo's location is surreal, dystopian and entertaining(simonwillison.net)
919 points by simonw | 405 comments
Simon Willison blog post describing how OpenAI's o3 model guesses where a photo was taken, testing it on a photo from El Granada, California.
Ollama Models Atom Feed(simonwillison.net)
41 points by mdp2021 | 12 comments
Hallucinations in code are the least dangerous form of LLM mistakes(simonwillison.net)
339 points by ulrischa | 264 comments
Simon Willison blog post arguing that LLM hallucinations in code are the least harmful kind of mistake because running the code exposes them immediately.
"A computer can never be held accountable"(simonwillison.net)
380 points by zdw | 253 comments
Simon Willison blog post sharing a 1979 IBM training page stating a computer can never be held accountable and so must never make a management decision.
Anthropic: "Applicants should not use AI assistants"(simonwillison.net)
392 points by twapi | 307 comments
Simon Willison blog post quoting Anthropic's job application form asking applicants not to use AI assistants when applying.
Notes on OpenAI O3-Mini(simonwillison.net)
117 points by dtquad | 42 comments
Promising results from DeepSeek R1 for code(simonwillison.net)
926 points by k__ | 689 comments
Simon Willison blog post on a llama.cpp PR doubling WASM speed via SIMD, mostly written by DeepSeek R1, plus his own results using R1 for code.
