Reviewing my LessWrong post history with Claude40m ago Low-Background Minds45m ago Your AI Agents Can Write Code. Can They Understand Your Business?1h ago Meta open sources code to let you make Muse AI gadgets2h ago Robotics AI developer FieldAI reportedly raising $700M in funding2h ago Japan's entertainment robots diversify with AI, offer comfort3h ago OpenAI safety leader David Robinson resigns as the team's upheaval mounts3h ago Apple to Tighten Mac Data Controls in Guard Against AI Agents3h ago Contest-Winning Microscope Video Is Full of AI Confabulations, Scientists Say3h ago The Mulleted, Meme-Loving Billionaire Behind Meta’s Hit AI App3h ago What is recursive self-improvement and what would it mean to ban it?4h ago Cerebras stock hits post-IPO low, tumbling 20% for the week on Nvidia pressure and lockup expiration4h ago State of the Field: AI for Epistemics and Coordination4h ago The Netherlands considers banning Meta AI glasses as some retailers pull them from shelves4h ago

Reviewing my LessWrong post history with Claude40m ago Low-Background Minds45m ago Your AI Agents Can Write Code. Can They Understand Your Business?1h ago Meta open sources code to let you make Muse AI gadgets2h ago Robotics AI developer FieldAI reportedly raising $700M in funding2h ago Japan's entertainment robots diversify with AI, offer comfort3h ago OpenAI safety leader David Robinson resigns as the team's upheaval mounts3h ago Apple to Tighten Mac Data Controls in Guard Against AI Agents3h ago Contest-Winning Microscope Video Is Full of AI Confabulations, Scientists Say3h ago The Mulleted, Meme-Loving Billionaire Behind Meta’s Hit AI App3h ago What is recursive self-improvement and what would it mean to ban it?4h ago Cerebras stock hits post-IPO low, tumbling 20% for the week on Nvidia pressure and lockup expiration4h ago State of the Field: AI for Epistemics and Coordination4h ago The Netherlands considers banning Meta AI glasses as some retailers pull them from shelves4h ago

Dev1 min reading time

Failure modes of Claude Code in a supervised but code-blind 60h project

LessWrongPublished 10 hours ago

Read the full article

A 60-hour supervised project examined the failure modes of Claude Code, an AI coding assistant, focusing on its performance without direct code visibility. The study highlights specific challenges and errors encountered by Claude Code in this constrained setting.

#claude-code #code-assistance #software-engineering

More in Dev

Dev3 min read

OpenAI DevDay 2026 Recap for Developers

InfoQ (AI, ML & Data)16h ago

Dev4 min read

AI is changing developer work. Here are three skills to strengthen.

GitHub Blog (AI & ML)11h ago

Dev5 min read

Datalab Introduces OmniExtractBench to Fix Bias and Opacity in Extraction Benchmarks

MarkTechPost11h ago