Anthropic may kill plan mode in Claude Code
Thariq says the models do not need plan mode anymore and is thinking of handing the shift+tab hotkey to effort levels instead. Matt Pocock wants it gone but kept in userland, and Kent C. Dodds says he has not used it in months.
The big disagreement
Thariq floats killing plan mode, Pocock says make it userland
Thariq said the team is thinking of killing plan mode and reusing the shift+tab hotkey to adjust effort levels, because he does not think the models need it anymore. Matt Pocock agreed it should go but argued planning belongs in userland, since how people plan is custom to the task, org and person.
Don't mandate a 'mode' and users AND maintainers will be happier
Receipts
Anthropic says it made claude.ai 3x faster in two weeks
Anthropic's ClaudeDevs team published how it made claude.ai 3x faster in two weeks, using Claude itself to measure, debug and improve performance, with the prompts and methods included.
this is a new type of post we're trying where we share in-depth about how we do work using specific prompts and techniques you should be able to replicate
Theo says no lab currently has both small and large models worth using
Theo posted a three line verdict: Anthropic has no small models worth using, OpenAI has no large models worth using, and Google has no models worth using. He gave no benchmarks or tasks behind the ranking. Nobody in our window has published numbers either way.
Google has no models that are worth using right now.