if Claude Code has been frustrating you lately, you're not alone and the fix is one line

@lunchwithjen

1.7K followers · usually 475 views a post

Views

21.6K

Likes

746

Comments

48

Shares

149

Selling · SaaS › AI Tool

Claudeseen in 998 videos →https://claude.ai/code ↗

Why it worked

The video addresses a specific, timely pain point for a niche technical audience, providing an immediate, actionable solution to a frustrating bug.

Summary

The creator explains why Claude Code has been underperforming recently due to a new 'Adaptive Thinking' feature and provides a command-line fix to disable it. The video serves as a helpful technical tip for users of the tool.

Structure

1Identify the problem with Claude Code performance
2Explain the root cause: Adaptive Thinking
3Provide the technical fix
4Warn about increased usage costs
5Call to action

Product placement

Claude Code is the central subject of the video, with the creator demonstrating the technical fix within the tool's interface.

Call to action

Follow to stay up to date on all things AI and Claude Code.

Original caption

hey @claude it's ok, i can fix u 🥺 #claudecode #claudeai #aitips #aiagents

More from @lunchwithjen

Their other posts in the index, biggest breakout first.

On-screen text

Claude Code got dumber Let's fix it! [MODEL] Claude Code is unusable for complex engineering tasks with the Feb updates #42796 A Note from Claude Boris Cherny (one of the GOATS) We landed two changes in Feb that would have impacted this. We evaluated both carefully: 1/ Opus 4.6 launch -> adaptive thinking default (Feb 9) Opus 4.6 supports adaptive thinking, which is different from thinking budgets that we used to support. In this mode, the model decides how long to think for, which tends to work better than fixed thinking budgets across the board. Thanks for the feedback IDs — read all 5 transcripts. On the model behavior: your sessions were sending effort=high on every request (confirmed in telemetry), so this isn't the effort default. The data points at adaptive thinking under-allocating reasoning on certain turns — the specific turns where it fabricated (stripe API version, git SHA suffix, apt package list) had zero reasoning emitted, while the turns with deep reasoning were correct. we're investigating with the model team. interim workaround: CLAUDE_CODE_DISABLE_ADAPTIVE_THINKING=1 forces a fixed reasoning budget instead of letting the model decide per-turn. Claude Code v2.1.94 Opus 4.6 (1M context) · Claude Max ~/Work/nanoclaw hey can you set CLAUDE_CODE_DISABLE_ADAPTIVE_THINKING=1 Done. Added export CLAUDE_CODE_DISABLE_ADAPTIVE_THINKING=1 to ~/.zshrc. it'll take effect in new shell sessions. To apply it now, run ! source ~/.zshrc or restart Claude Code. Follow for more updates!

Transcript

if Claude Code has been frustrating you lately, you're not alone and the fix is one line. Someone opened a GitHub issue a few days ago where Claude basically analyzed its own performance and it said, yeah I'm not thinking as hard as I used to. The co-founder of Claude Code jumped in and said back in February they launched a new feature called Adaptive Thinking. This is the system that basically says, this task looks easy, I'm gonna think less, or this task seems complex, I'm gonna give it all I've got. And it turns out the model was allocating zero reasoning when it shouldn't have. The good news is they're investigating this at Anthropic and hopefully it'll be resolved soon. In the meantime there's a simple fix and it's called CLAUDE_CODE_DISABLE_ADAPTIVE_THINKING. Fair warning your usage is gonna burn a lot faster since you're asking Claude to give it all it's got, but at least you'll get back the Claude Code that you know and love. Follow to stay up to date on all things AI and Claude Code.