Anthropic engineers discuss agentic behavior in Sonnet 4.5 and Opus 4.5
Anthropic's Applied AI team identified 'context anxiety' in Sonnet 4.5, where the model prematurely ended tasks as it approached its context limit. The team implemented context resets to mitigate this behavior, which was subsequently resolved in the release of Opus 4.5.
Featured · Gagan Bhat, Isabella Kai He
1 source
Daily brief
Get tomorrow's AI brief in your inbox
More stories today
- Discovered Materials raises $9M to find chip-cooling materials with AI
- Your agent didn't hallucinate; it exceeded its authority
- User runs 1M context on 17GB model using 24GB VRAM
- MOSS-TTS-Nano runs realtime multilingual speech on CPU
- The AI Slop Backlash Is Actually Having an Impact