What are side projects you have been putting off for a long time? For me it's writing a poker solver using counterfactual regret minimization

Apple SSO and "hide my mail" are my favourite Apple features

Excited for a future where all APIs provide a corresponding MCP server

Have you configured any hooks or do you just manually tell claude to format all files/format them yourself?

Is there anything that hits as hard as refactoring your code into using perfectly abstract generic traits? 😮‍💨

If you don't mog your coworkers with the way you navigate code/windows yngmi

The best thing about AI coding tools is that people now actually do the tedious but marginally usefull things. Creating unit tests, writing doc comments, etc

Currently experimenting with doing task tracking directly inside the repository instead of in jira/github/linear. This synergizes very well with claude code. I like it so far. It does of course not completely replace project management tools, but may work for very small teams.

When you read a take so dumb, all your fears around the future of the software job market are cured

Do you have scheduled work breaks or just work until you snap out of it?

The lion does evals on synthetic data

Me, my notebook and these against the world

Every project needs a great CI (clanker integration)

What the actual fuck did deepseek cook up with this model

Currently spending my entire day trying to convert a huge html file using tables for layouts into a markdown/semantic html file. Not sure if OCR wouldn't have been faster

One time claude pissed me off so hard that I threw one of gel pens on the floor. The case cracked and the whole room was splattered with ink. The blue spots on my wall remind me that patience is a virtue

mfw "You're absolutely right"

Mistral is the most cooked AI lab out there. Today i told it to summarize a file's contents and it completely hallucinated the entire content based on the file name. That's like not finding the haystack on top of the needle

If i worked at anthoripic I would simply change the tokenizer to treat "You're absolutely right!" as a single token and then leave