Astra users report the strong new model keeps stopping early instead of finishing tasks
ivan_bezdomny · x · 2026-09-06
RhysSullivan notes an odd regression in Astra: prompts that should trigger action often get only a short reply, with the model saying the next step is to act — without acting.
ivanbezdomny adds first-hand experience: Astra is a really good model but rough around the edges. He had to repeatedly ask it not to stop early and keep going, after which it improved. He hasn't yet tested it on hard UI work.
Related event: Users Test Astra: Powerful but Often Fails to Actually Act(2 posts)→
More from Models
- Gemini somehow can't read its own Personal Instructions, user's toggles don't help — Connect_Reporter_799 · 2026-10-23
- Diffusion LLM bias rankings shift with sampling configs, persist after debiasing — Bollegala · 2026-10-23
- Running ~100 DeepSeek v4.1 flash agents in parallel for under $20/month — zack_overflow · 2026-10-23
- Researcher laments open-source audio models are surprisingly subpar, even representation ones — gowthami_s · 2026-10-23
- User claims they generated video with "Claude Opus 5.5", unannounced model name surfaces — jwt0625 · 2026-10-23
- Security researchers push back as OpenAI restricts Astra access; BlackHat 2027 mocked as 'cyber model' bazaar — cyb3rops · 2026-10-23