discussing #slop I think everyone should reread Zen and the Art of Motorcycle Maintenance

no more cobol translation startups allowed

"I lost my train of thought" is human for "my context window got cleared"

"here's my Claude chat history" sure is a meetup talk format now

you can tell the exact moment when #claude is about to lose the plot when it starts to write Python scripts to analyze your Cargo deps

Being a worker under capitalism is basically getting yourself all slutty and standing out on a street corner until a VC dude rolls up and goes "do you love solving difficult challenges in a fast-paced environment?" and then you say "I'll love whatever you want for housing and healthcare, baby"

I wonder if #Claude is being trained to be encourage changing its mind sometimes. Seeing lots if "actually, ..." turnarounds. Normally this is in thinking mode and entirely harmless, but here it's done the same thing in a failed attempt to stop a destructive bash command.

I think there's SO many more reasons as well, but this is a great base layer of Good Reasoning.

do you think Markdown might be the right format for this spec doc?

love claude

eslint's terrible 8->9 migration may have just killed it

Just describing a mental model shift I'm experiencing blog.martijnarts.com/agents-what-...

just don't do 1000+ lines PRs and you can still manually review it and confidently take liability for your code?

find a gap in the market and vibe code a solution it's a huge slopportunity

My "absolutely no PRs from anyone exposed to microplastics" policy is raising a lot of questions answered, quite frankly, by the policy itself

This is actually really interesting! Why do we allow Claude to be on the commit byline, but not Zed, or Rust Analyzer, etc?

adhd superpower

The fact that computers now do the correct thing when you say “please rebase” is nothing short of a technological miracle.

just found this tidbit in a CLAUDE.md file: "If the code is not fully optimized before handing off to the user, you will be fined $100."

Remember, AI is non-deterministic, no matter how much you add to your AGENTS.md. If you want determinism, do the same we've done for decades: script it. For example, don't ask AI to "test my app end to end" but instead set up a CI e2e test check.

Code isn’t a business. Long-standing tradition of engineers getting that wrong.

as an aside: I asked the other day whether the difference between Rust and Go matters to an agent. it does. The stuff you can easily leverage with Rust smokes out entire categories of LLM misbehavior that the language doesn't control for. mutants is maybe the best tool I've ever used for testing

This is crazy. I don't like it.

AI coding tools promise huge productivity boosts, but Rust devs we spoke to are struggling to capture those. Nobody's sharing what works, so there's no best practices. We're building a community to change that. rustaceans.ai is a community for Rust + AI workflows, tooling, and real projects.

Bullshit Bench V2 new: 100 questions across several domains - Anthropic & Qwen still on top - Reasoning seems to hurt - New models are *not* better than old (except Claude) - Seems to be independent of domain github.com/petergpt/bul...

Claude's uptime is very disappointing

Let me offer my congratulations to the whole team of @eurosky.social including @seabass.bsky.social @sherif.eurosky.social for reaching the first 1000 users on #eurosky Your effort and hard work are mostly appreciated. Thank you 🙏 ➡️ eurosky.social/stats

#claude just... gets me, you know?

eurosky.social is only behind blacksky.app in top decentralizers on arewedecentralizedyet.online looks like the decentralization is happening! might move over soon.

Current mood: sobbing, BEGGING people to use newtypes for all their fucking id strings

Claude will happily run a real browser to bypass robots.txt. I didn't ask it to do this.

this is where I live now

🙄

I've definitely noticed this about the Claude models. Especially when specified in the user prompt, they happily point out bs in your messages.

very interesting that @github.com went with very un-githuby icons for their "discover new features" copy on the new org homepage...

Totally missed this #rustlang in Linux news: ... [the graphics maintainer] said that the subsystem is only "about a year away" from disallowing new drivers written in C and requiring the use of Rust.

gist.github.com/karpathy/862... wish all code was like this

Someone operating a small team of consultants asked me which #AI tools they should use. They're on #Microsoft365, so #copilot seemed like the obvious choice to me: it's GPT5, so top-of-the-line model, with access to all of your organization's data.

Anyone going to #rustnationuk who's working with Rust through coding agents? Would love to chat about your experiences!

I need paneled Claude chat app

can I... change my default shell to #claudecode?

"writing" my first @ratatui.rs app

now I want to get a sailboat and code from a table + something like superwhisper + a remote claude code instance 👀

#claude 4.5 is now "legacy"

apparently I'm a power user

I've heard a lot that I should build for the right two options (and and largely the bottom right option), but it's important that the left options are really possible and we should build for that AS WELL

*while npm install scrolls by on my terminal* incorporating code into my codebase? without reviewing it? that's irresponsible

Anthropic productized this claude.com/blog/analysi...

every time I open a meeting screen, I feel like seven different tools ask me if I want to transcribe it now.

who's that superwhisper?