Opus 5.5 goes way too hard.
I have it working on a MacOS finances app for close to 24 hours now almost non-stop and it's got me thinking "what else can I throw at it before my weekly limit lapses."
@AnthropicAI cooked.
I am tweeting into an empty room without my Claude subscription Max 20x (that just renewed yesterday) to test Opus 5.5 with my fresh, unused, weekly reset...
Unrelated to the interesting post below... Just like the world "delve" has fallen in infamy, and the em-dash is now banned of human lexicon...
Is the suggestion of "Nova" the new "I vibe coded this app"? I keep finding "Nova" being the suggestion of (particularly @AnthropicAI ) AI modelsโฆ
Am I seeing things? Anyone else?
Well, I didn't have this in my bingo cards for the day.
Used Claude from 2 supported countries in that list, nowhere else. What could be going wrong?
@claudeai
@laozhang_ai the empty room is this twitter account. Thank you for your answer ๐
My claude max subscription was suspended a few hours ago for what I believe is a mistake... it renewed yesterday, I got a banked reset that I didn't get the chance to use!
Well, I didn't have this in my bingo cards for the day.
Used Claude from 2 supported countries in that list, nowhere else. What could be going wrong?
@claudeai
My AI agents used to end every session the same way: "Done! Everything works!"
Nothing worked.
So I wrote them rituals. A session can't close until it proves what it checked and admits what it skipped. A plan can't ship without a kill condition. A fix doesn't count until a different model fails to break it.
Customers kept asking for them IRL. Open-sourced today. 15 skills, Claude Code and Codex, MIT.
Link below.