Allie K. Miller's early Astra review: stunning computer use, still fails wordplay
alliekmiller · x · 2026-09-04
Allie K. Miller shares a 7-point early review of GPT-6 Astra for business users:
- Computer and browser use are incredible — she believes most stable computer workflows can be at least partially delegated to AI, and challenges people to go mouse-free for a day.
- Still fails her complex wordplay challenge; no model has passed it.
- Doesn't frustrate her, unlike Claude's increasingly convoluted technical replies — she no longer says Claude is the better writer, and Codex's share of her AI work keeps rising.
- Codex is the better daily driver.
- Not worried about tokens, unlike the Fable family (not worried about Opus either).
- 3D/animation: others were delighted; she'd rather see accurate worlds from reference points than from-scratch generation.
- Would still use Fable 5.1 for "mega architect" builds.
She also thanks an OpenAI engineer for shipping her request: Codex voice notifications for approval prompts.
More from coding & agent
- Prime Intellect's RL stack adopts NIXL, cutting 800B-model weight transfer 9x from 86s to under 4s — xeophon · 2026-09-04
- Alex Zhang defines the agent harness: code around an LLM, an agent once placed in an environment — CShorten30 · 2026-09-04
- OpenAI's Heidel: compaction in Responses API lets GPT-6 Astra max out ARC-AGI-3 — stevenheidel · 2026-09-04
- Non-engineer at Earendil shares how he customized the Pi coding agent for daily work — solyarisoftware · 2026-09-04
- GPT-6 Astra builds explorable 3D city scenes in Unity, 1.9x faster than GPT-5.6 Sol — ChrisGPT · 2026-09-04
- Dev builds a working macOS 27 simulator with Astra in 75 minutes, cloud sync included — pvncher · 2026-09-04