Dev Says Claude 'Forgot Everything' After Two-Day Break, Pines for a Consistent API
HanchungLee · x · 2026-10-26
Developer @ivanbezdomny reports that after two days away from Claude and Astra, the models started making wild mistakes they hadn't made in weeks, "as if they lost all context." He sees this as a reason companies want sovereign AI, and wishes for an API that stays consistent and does what you ask. A reply asks whether the degradation is measurable via objective evaluation or just a psychological effect of stepping away from the recsys-like experience.
Related event: Developer Complains Claude 'Forgets' Context After Two Days Idle(3 posts)→
More from Models
- I Asked 5 Text-Only LLMs to Draw Objects They've Never Seen — gabriberton · 2026-10-26
- Japan to build AI cyberattack simulation team as AI defense goes mainstream — evijit · 2026-10-26
- Anthropic engineer: Claude Code cache hit rates are now high, savings live — teortaxesTex · 2026-10-26
- Fine-tuned 2015 CNN beats GPT-6 and Claude on skin lesion diagnosis, 62% vs 42% — lmoroney · 2026-10-26
- "Never seen Claude this happy": community speculates Opus 4.5 is imminent — adonis_singh · 2026-10-26
- 'Absolute humiliation': China open-weights model outscoring OpenAI flagship draws mockery — alejandroll10 · 2026-10-26