Posts by simon@fedi.simonwillison.net
(DIR) Post #AvKk2vEOvBYP0sQNWK by simon@fedi.simonwillison.net
2025-06-20T20:37:47Z
0 likes, 0 repeats
SnitchBench was fun enough already, turns out we need to add MurderBench to the collection of dystopian benchmarks that we run these models through https://simonwillison.net/2025/May/31/snitchbench-with-llm/
(DIR) Post #AvMdC8hlsviW2oWvkO by simon@fedi.simonwillison.net
2025-06-21T18:34:50Z
2 likes, 1 repeats
Microsoft released a new terminal text editor! It's called Microsoft Edit, it's open source, it's tiny (about 250KB Rust binary) and it works cross-platform. They built it for Windows 11 - I've been trying it out on my Mac and it's a nice alternative to Vim or nano - my notes so far: https://simonwillison.net/2025/Jun/21/edit-is-now-open-source/
(DIR) Post #AvMdC9zt53no3I0uci by simon@fedi.simonwillison.net
2025-06-21T18:36:29Z
0 likes, 0 repeats
I ended up running their Linux binary on my Mac via Docker - here's an incantation that will run the container I built (assuming you have Docker installed):docker run --platform linux/arm64 \ -it --rm \ -v $(pwd):/workspace \ ghcr.i''o/simonw/alpine-edit
(DIR) Post #AvMdCAttiobar0DdvU by simon@fedi.simonwillison.net
2025-06-21T18:37:34Z
0 likes, 0 repeats
This is my first time publishing a Docker image to the GitHub Container Registry (with a lot of help from Claude 4 Sonnet) - I wrote up a TIL of how I did it and everything I learned along the way https://til.simonwillison.net/github/container-registry
(DIR) Post #AvnTrT2DZkn3pjG3yy by simon@fedi.simonwillison.net
2025-07-04T17:39:40Z
2 likes, 0 repeats
The is diabolical... a Python object that hallucinates method implementations on demand any time you call them, using my LLM Python library https://github.com/awwaiid/gremllm
(DIR) Post #AvqxRb2okmJBFEzxJY by simon@fedi.simonwillison.net
2025-07-06T09:45:09Z
0 likes, 0 repeats
Just found out that George Orwell's Nineteen Eighty-Four, written around 1948, describes a "versificator" that composes the words for songs "entirely by mechanical means" https://simonwillison.net/2025/Jul/6/nineteen-eighty-four/
(DIR) Post #AvqxRj5ErezyB4C0CO by simon@fedi.simonwillison.net
2025-07-06T09:47:04Z
0 likes, 0 repeats
George Orwell: In my book I invented the Versificator as a cautionary taleGen AI company: At long last, we have created the Versificator from classic sci-fi novel Nineteen Eighty-Four!Hat tip: https://twitter.com/AlexBlechman/status/1457842724128833538
(DIR) Post #AwU7E8z4cZk5600wZU by simon@fedi.simonwillison.net
2025-07-23T17:19:21Z
0 likes, 1 repeats
I wrote up some notes on Google Security's new OSS Rebuild project, which increases supply chain security for popular packages on PyPI, NPM and Crates through offering independent build attestationshttps://simonwillison.net/2025/Jul/23/oss-rebuild/
(DIR) Post #AwU7ENyYslFHaktFM8 by simon@fedi.simonwillison.net
2025-07-23T17:22:19Z
0 likes, 0 repeats
OSS Rebuild doesn't (yet) have a web UI... but it turns out their data is in a public Google cloud bucket, which means you can host your own web app in a separate cloud bucket and use fetch() to access their data!So I had Claude Code build and deploy a vibe-coded search UI
(DIR) Post #AxL0U0zxs2OOf3kM8O by simon@fedi.simonwillison.net
2025-08-19T19:47:22Z
0 likes, 0 repeats
One of the quickest ways to start playing with a good local LLM on macOS (if you have ~12GB of free disk space and RAM) - using llama-server and gpt-oss-20b:brew install llama.cppllama-server -hf ggml-org/gpt-oss-20b-GGUF \ --ctx-size 0 --jinja -ub 2048 -b 2048 -ngl 99 -fahttps://simonwillison.net/2025/Aug/19/gpt-oss-with-llama-cpp/
(DIR) Post #Ay4RRhz91zUzzOgUmO by simon@fedi.simonwillison.net
2025-09-10T17:28:35Z
0 likes, 0 repeats
Some notes on the new Claude API web fetch tool, which I think can be used safely despite the risk of prompt injection exfiltration attacks if you're really careful with the allowed_domains parameter https://simonwillison.net/2025/Sep/10/claude-web-fetch-tool/
(DIR) Post #AysMS41hFQZA7whn3Q by simon@fedi.simonwillison.net
2025-10-02T15:13:55Z
1 likes, 2 repeats
Joshua Rogers, using AI tooling responsibly and professionally, reported 22+ genuine issues in curl that are now being addressedThis is especially notable because curl has had huge problems with floods of garbage slop AI "security issues" in the past that were nothing of the sort - my notes here: https://simonwillison.net/2025/Oct/2/curl/https://mastodon.social/@bagder/115241241075258997
(DIR) Post #AywVWBsLgOqTqiDP28 by simon@fedi.simonwillison.net
2025-10-06T19:50:01Z
1 likes, 0 repeats
I got the new GPT-5 pro API model to "Generate me an SVG of a pelican riding a bicycle". This pelican took 6m8s to generate and cost me $1.10! https://simonwillison.net/2025/Oct/6/gpt-5-pro/
(DIR) Post #AzRVfqxuIyLTnEBFAm by simon@fedi.simonwillison.net
2025-10-21T18:46:37Z
0 likes, 0 repeats
Wrote up my first impressions of ChatGPT Atlas, OpenAI's new browser - I remain unconvinced by the entire category of "browser agents", the security and privacy challenges still feel insurmountable to me https://simonwillison.net/2025/Oct/21/introducing-chatgpt-atlas/
(DIR) Post #AzRVfsCpgxsXdoAg4m by simon@fedi.simonwillison.net
2025-10-21T18:49:34Z
0 likes, 0 repeats
It's neat to see them encourage developers to add ARIA tags to pages though, an "agent" can be thought of as effectively another form of assistive technology
(DIR) Post #AzY99hdyaYf0jOqFSy by simon@fedi.simonwillison.net
2025-10-24T23:06:21Z
1 likes, 0 repeats
It's neat how if you ask Claude Code questions about itself it can answer them, because it knows how to fetch a Markdown index of its own online documentation and then navigate to the right placeI wish more LLM tools would implement the same pattern! https://simonwillison.net/2025/Oct/24/claude-code-docs-map/
(DIR) Post #B1FSA8bsnU0AyfLKFc by simon@fedi.simonwillison.net
2025-12-14T18:04:46Z
1 likes, 0 repeats
@dvshkn absolutely! It turns out the engineering practices of top-tier software development teams - automated tests, clean modular design, up-to-date documentation, linters, CI etc - are exactly the things you need to get the best results out of coding agent
(DIR) Post #B1I1BMth33NGI2uppw by simon@fedi.simonwillison.net
2025-12-16T00:37:25Z
0 likes, 0 repeats
I ported a Python library implementing a full HTML5 parser to JavaScript using GPT-5.2 and Codex CLI in 4.5 hours, and decorated for Christmas and watched Knives Out while I was doing it https://simonwillison.net/2025/Dec/15/porting-justhtml/
(DIR) Post #B1I1BVRfGdIknis4tk by simon@fedi.simonwillison.net
2025-12-16T00:39:50Z
0 likes, 0 repeats
I'm sharing this because it's quite a stark demonstration of what's possible using the latest frontier LLMs at the tail end of 2025Is it a GOOD idea to churn out a 9,000 line direct port while watching Daniel Craig chew on the scenery? That's a whole other set of questions!
(DIR) Post #B1NVu8a75bvLKwowFc by simon@fedi.simonwillison.net
2025-12-18T14:57:27Z
0 likes, 0 repeats
I see a lot of complaints about untested AI slop in pull requests. Submitting those is a dereliction of duty as a software engineer: Your job is to deliver code you have proven to work https://simonwillison.net/2025/Dec/18/code-proven-to-work/