It helps to enforce thoroughness with directives in AGENTS.md

I don't know why, but only chinese models like kimi or deepseek give me consistent results

Seriously. About 3 weeks ago they turned down the intelligence of Opus by several orders of magnitude. I had a whole setup configured where it knew what files to read for context on whichever project i was working on. All of a sudden its not reading half of what i say to it. I now spend half my time...

Try Cursor auto mode brah. It’s pretty pretty pretty good. Also, token efficiency is awesome. I don’t even hit my limits on my $20 a month plan.

Yeah. Even between yesterday and today Claude seems lazier.

Yes it’s more lazy. I use sub-agents as my way around it. But have to also add a maker/checker pair to them to make sure they do the work.

Don’t worry it’ll only get worse after the ipo.

Claude's lack of effort is a symptom of a larger issue: institutional complacency, perpetuated by enablers like Bezos and Musk.

Wow it's getting more and more human by the day.