"CLAUDE.md is over the 40.0k-char limit (45.7k chars)" is why I think the death of SaaS is overblown.
Like, Anthropic, why make this my responsibility? I haven't been writing things to the CLAUDE.md file. I haven't given Claude Code its strategy for what to document there. Its little AI butt has been making the decisions and abstracting that away from me, but then it falls down and now you're asking me to decide what to prune?
It's like a five year old who eats too many blueberries and then starts crying about how miserable they are. I mean, play stupid games, kid. But now it's my problem...
Anyway, building software is fun. Maintaining software, less so, yet still really important. Dare I say, even more important.
And when you have AI designing and implementing and maintaining software, you often get warnings like the above: "I did something wrong, please tell me what to do, even though you have no idea what's going on under the hood."
And, yes, we just ask Claude Code to figure it out and it generally does, but we lose a bunch of trust in the tool in te process.
Anyone who has designed software used by more than one person knows that 90% of great software boils down to UX and edge cases.
Taste, as the people say today.
Claude.ai, Claude Code, Claude Design and others are what happens when the best AI company in the world decides to leave taste up to the models.
A dazzling array of half-built features and bewildering interface decisions that we put up with because, well, Claude is magical.
Your software is not Claude. You shouldn't leave the experience to the AI.
Anyway. Anthropic. The Apple of AI. IYKYK