GPT 6 (Non-Astra) Underdelivers: Devs Report Prompt Misreads, UI Rewrites, and Long-Task Failures
vaishnavsm · reddit · 2026-09-27
A developer reports that GPT 6 (non-Astra) has been frustratingly worse than expected — the worst experience since the 5.3 era:
- Sol misinterprets prompts and rewrites UI unprompted (turning charts into card-based lists)
- Mistakes of a kind unseen at the frontier for 6 months, behaving like a passive-aggressive genie
- Latency unchanged, ruling out a thinking-budget issue
- Long-running tasks no longer trustworthy
The author's headcanon: OpenAI renamed 6 Sol to Astra, promoted Terra to Sol, and kept Luna — explaining the missing Terra. The thread asks whether others are seeing the same regression.
Related event: Developers Report GPT-6 Regression: Misreads Prompts, Unreliable Coding(2 posts)→
More from coding & agent
- Sydney Codex Community Buildathon set for Nov 14 with Cerebras-backed ultrafast compute track — paw_lean · 2026-11-11
- shadcn's improve hits 10k stars: let top models plan, cheap models execute code fixes — shadcn · 2026-11-11
- OpenAI open-sources Codex Security CLI and SDK for AI-driven vulnerability scanning and fixing — openai · 2026-11-11
- Claude Opus 5.5 runs an architecture search and finds AttnRes-Canon, a new top Transformer design — SonglinYang4 · 2026-11-11
- 5 research habits worth knowing: read papers backwards, reproduce first, wire coding agents via MCP — cneuralnetwork · 2026-11-11
- 'Adam can now do CAE': AI agent showcase in engineering simulation draws buzz — burhop · 2026-11-11