simonw

30 ms·

961.
by simonw 4mo ago
That was a problem when testing code was expensive to create and expensive to update. That's no longer the case. Plus, SQLite famously has 590x the test core compared to the implementation code, and they wrote that by hand! https:&#x2
962.
by simonw 4mo ago
Better headline: We deliberately gave GitHub's AI Agent permission to access both public and private repos and then tricked our configured agent into leaking private repos.
963.
by simonw 4mo ago
What do you mean by "steal it"? What are they doing with that code that's underhand?
964.
by simonw 4mo ago
It still solves "breaks two other things".
965.
by simonw 4mo ago
They solve "adding a feature now takes days and breaks two other things".
966.
by simonw 4mo ago
> You have an AI-generated codebase that works, but adding a feature now takes days and breaks two other things? Sounds like you forgot to have the agents use red/green TDD and build a robust test suite while they were shipping all
967.
by simonw 4mo ago
This looks like a really solid app. I like that it's 17 MB and uses the ContainerAPIClient library directly. 28 commits in 3 days, 5,015 lines of Swift, every commit "Co-Authored-By: Claude Fable 5". Also neat that it's
968.
by simonw 4mo ago
Good luck implementing and then maintaining a project of this size and complexity at ~100 lines of verified code per human developer per day.
969.
by simonw 4mo ago
> I think AI is powerful tool, but we still can't give DC expansion a pass. I really don't think we are. In the wider culture the idea that data centers use an absurd amount of water is baked in at this point. It frustrates me
970.
by simonw 4mo ago
Andy has a good track record for writing about this. He shares plenty of credible citations - more so than most other people commenting in this space. He also caught a major error in one of the most widely read books that helped kick off
971.
by simonw 4mo ago
This is a useful piece on that: https://andymasley.substack.com/p/the-ai-water-issue-is-fake
972.
by simonw 4mo ago
> This is the caching API we've always wanted Workers to have. Here's why it took us this long I was looking forward to the "why it took us this long" explanation but it wasn't explicitly spelled out. Any Cloudfl
973.
by simonw 4mo ago
I expect it's a model problem and not a harness problem, purely because some of the best harnesses (including OpenAI Codex itself) are open source and can be very easily tried against a new model.
974.
by simonw 4mo ago
Which harness software were you using for this 2 years ago? VS Code Copilot? Cursor?
975.
by simonw 4mo ago
This is a thinner TechCrunch rewrite of this Reuters story: https://finance.yahoo.com/technology/ai/articles/exclusive-z... The exact quote appears to be: > In retrospect, he said, the "trajectory of
976.
by simonw 4mo ago
This whole thing is a pretty fascinating insight into a whole other corner of the internet ecosystem: > For more than a decade, one of the most rewarding blueprints for making money online was to “start a blog." Then: > These hun
977.
by simonw 4mo ago
Yeah these skills match that list of names, looks like this one may have extracted all of the prompts: https://github.com/JimLiu/baoyu-design/tree/main/skills/baoy...
978.
by simonw 4mo ago
If you ask Claude Design itself to list the names of the skills available to it you get: Animated video Interactive prototype Make a deck Make a doc Make tweakable Claude API in prototypes Frontend design Wireframe Expor
979.
by simonw 4mo ago
I can't even tell if this repository is based on prompts extracted from Claude Design or if the author had an LLM create all of these prompts in it from scratch. The fact that they encourage and accept PRs indicates that this isn'
980.
by simonw 4mo ago
That was an article in The Information but it didn't read very well to me, I didn't get the impression the author was enough of a technical expert on how LLMs work to credibly evaluate the claim, which came from an insider rumor:
981.
by simonw 4mo ago
I didn't know you could do this with fetch() and a data: URI: fetch('data:;base64,1ZpLsgIxCEXnrM...==').then( r => r.body.pipeThrough(new DecompressionStream('deflate-raw')) ).then( s => new Resp
982.
by simonw 4mo ago
Disagree. Many people are really bad at clear communication. It doesn't matter how good an LLM gets, it will still be less effective to use it if you give it prompts with no single obvious meaning - you'll have to go a few extra
983.
by simonw 4mo ago
The bits about learning to communicate in clear and unambiguous way, having a skeptical ear and knowing when to double-check information that is being presented to them, and understanding how to use tools in a way that enhance their own lea
984.
by simonw 4mo ago
Yeah, the big advantage of playwright is that it's cross-browser. It's still my go-to tool for automated testing - I use the CDP trick for one-off automation tasks.
985.
by simonw 4mo ago
"AI really isn't a skill that needs to be taught" At this point that's like saying Microsoft Excel isn't a skill that needs to be taught. None of this stuff is easy to use, or obvious. If you want to get meaningful
986.
by simonw 4mo ago
I'm a big fan of Parakeet v3 - I run it using the MacWhisper app, it's a 494MB model and the quality is excellent.
987.
by simonw 4mo ago
That $200/month is already more like $4,000/month if you have to pay full API pricing - "enterprise" companies for example. That drops the equivalent to 10 months. (I'd be surprised if that local rig really can driv
988.
by simonw 4mo ago
I've been telling agents to run Chrome and talk to it directly via the CDP recently and finding that to work very well.
989.
by simonw 4mo ago
What programming language are you using for Design by Contract?
990.
by simonw 4mo ago
Welcome to volunteer-driven open source. (Update: you're a Debian developer so you're even more familiar with how that world works than I am.)

More ›