Subscription fatigue has just moved from streaming providers to AI providers.

Throttling TTS in parallel with a lockfile is dopamine maxxing

If one happens to be looking for a Discord alternative that doesn't have ID verification in place - qave.chat could be what they are looking for.

anyone needing a chat app to try in light of the discord stuff. qave.chat should help!

Back to rounded-none

I might try Effect again, feels like it's beginning to adopt the tools I use in typescript anyway and the extra defensiveness is nice.

The issue I feel with writing too many test and big features per hour is maintenance and iteration of this feature will be painful. Too many test is smell for testing implementation and not intent Big feature per hour is smell for new features that are good enough but continue to grow the entropy of system complexity. AI can 10x the volume of code, good programs though are usually re-written 10 times, each time the intent is clearer and simpler. I would love to see this type of LOC but for removals too, removing 4K LOC is a signal that I will trust a bit better.

I wish the token inference was 4x faster. My dev loop is bottlenecked right now with waiting for token output and I’m not cracked enough to be multi managing them.

With GPT Codex being 40% faster OpenCode supporting it, I'm switching from Opus 4.5. It feels pretty much the same now and until Jarred makes Claude Code workable or Sonnet 5 comes out. This is my new default.

Instead of talking to users, I over-engineered the not-found page. qave.chat/lol

CSS gets squircles and fancy corner radius tricks. The vibe shifts to no radius. Every time.

Devs

Would you trust a nurse to perform your surgery with a surgical robot?

I'm convinced non-technicals claiming more ownership over the stack is an extreme case of copium.