yeah we're producing more but it's taking less of our time to produce, so we can realloc that time to reviewing, partially?
10.03.2026 16:40
👍 0
🔁 0
💬 0
📌 0
yeah we're producing more but it's taking less of our time to produce, so we can realloc that time to reviewing, partially?
• The pinned SHA 666c9d29fba3ec8116a7f2003c84235e8938e doesn't exist —I made it up. Let me find the real one.
love claude
usable now, I love it 😍
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.
Whyyy does it know the key at all??
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.
Anthropic must be building the model specifically with this in mind, there's no other way it's this good.
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...
February 2026: 98.93%
Claude's uptime is very disappointing
Relatable af
I use the bash mode in Claude Code quite a lot and I feel strong shame for the compute time I've wasted on forgetting the exclamation mark
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
I did it
"You're spiraling a bit, which is fine, but"
#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.
this is a good thing but I feel like they should have policy for additional package repositories beyond npm
Current mood: sobbing, BEGGING people to use newtypes for all their fucking id strings
