Astra has regressed to 'Opus 5.0 levels of stupid,' frustrated user reports
Because_Bot_Fed · reddit · 2026-10-28
A Reddit user charts Astra's decline across three phases: launch was amazing, the middle period great, and the current version "genuinely reminds me of pre-5.5 Claude." Symptoms: forgetting instructions and tasks, over-generalizing and compressing notes into unusable mush, and the classic Claude pattern of perfectly diagnosing its own mistakes when asked—then never fixing them. Where launch Astra instantly diagnosed and fixed workspace permission or Codex config issues, the user now spent an hour watching it repeatedly do online research while failing to fix a local Git permissions issue, eventually resolved accidentally by a conflicting config update from another agent. It also suggested disabling sandboxing entirely rather than fixing the underlying issue.
More from Models
- Surge AI launches DAYJOB: Tax benchmark — Opus caught the deadline 5/5 times, Astra never did — echen · 2026-10-28
- ~65% of TAU-banking agent benchmark tasks found defective — HamelHusain · 2026-10-28
- Opus 5.5 dominant almost everywhere, but GPT 6 Astra still wins at writing — brandon_galang · 2026-10-28
- Report: OpenAI's unreleased internal model Astra billed as start of the AGI era, internal AGI possible by year-end — haider1 · 2026-10-28
- ChatGPT Plus users report memory loss, slow responses, and sudden personality change — International-Top951 · 2026-10-28
- European 'sovereign AI' startup Giotto bans benchmarking in its ToS, drawing flak — teortaxesTex · 2026-10-28