gpt-6POST
Developer says Astra codes worse than Opus 5.5
The author claims GPT models keep getting worse at coding, that using Sol or Astra means spending more time and tokens fixing code, and that Astra feels dumber than Opus 5.5.
Haz
@diegohazArchived 11 days agoOriginally posted 2026.10.18
Haz @diegohazOriginally posted 2026.10.18
I posted this 2.5 months ago, and it still holds true. GPT models have only gotten worse at coding. At this point, whenever I consider using Sol or Astra for a coding task, I know I'll waste more time and tokens fixing the code or even rewriting everything from scratch. And compared with Opus 5.5, Astra feels even dumber. I really hope OpenAI can turn things around in the near future.
This post quotes another post: x.com/diegohaz/status/2083894348572172339
x.com/diegohaz/status/2111797826292203741Archived 2026.10.19 10:10