Post #1986388
2026-04-23 15:01 UTC
@nathan @martin @ChadF I think so much of it comes down to usage patterns and how each one of us interacts with it. I've basically seen zero difference in CC behavior over the entire few months I've been using it. It would be very interesting to see how other's work differs from mine and compare.
Replies (2)
-
@dave@podcastindex.social 2026-04-23 15:03
@nathan @martin @ChadF Local models will absolutely get to the level of hosted frontier models. They are very close already. 3 years from now I don't think any of us will be using the hosted models for coding.
-
@nathan@xoxo.zone 2026-04-23 16:05
@dave @martin @ChadF It's covered in the video, but here's findings shared by AMD's AI director: > Quantitative analysis of 17,871 thinking blocks and 234,760 tool calls across 6,852 Claude Code session files reveals that the rollout of thinking content redaction (`redact-thinking-2026-02-12`) correlates precisely with a measured quality regression in complex, long-session engineering workflows. https://github.com/anthropics/claude-code/issues/42796