Theo digs into the Opus 5.5 nerf reports
After a claim that Opus 5.5 changed on November 3, Theo looked and says what moved was the reasoning summarizer, not the model's capability. Two very different kinds of regression, and only one of them matters to your code.
Receipts
Theo says the Opus 5.5 nerf reports point at the reasoning summarizer
Theo dug into claims that Opus 5.5 was quietly nerfed and says Anthropic appears to be rolling out a reasoning summarizer change per account, with no sign of degraded capability for him.
No sign of degraded capability for me.
Levels used Claude Code to design furniture and order the parts
Pieter Levels says Claude Code can design furniture part by part and hand back a bill of materials to order and assemble yourself. He says @theannalux, who is neither a furniture designer nor a coder, designed a table from scratch working with Claude Code in the terminal, which he insists beats the app.
How they actually work
David K Piano: tell agents what you want, not what to do
David K Piano posted that telling your agents what you want beats telling them what to do, and flagged it himself as possibly painfully obvious. Kent C. Dodds amplified it with his own framing about talking with the agent rather than at it.
You should be talking with your agent, not at your agent.