gpt-6POST
Developer says OpenAI's dots can't fix poor taste in the underlying model
The developer argues OpenAI's new dots interface doesn't improve the model: instruction following and taste remain poor because the problem is the Astra/Sol base models.
David Shapiro (L/0)
@DaveShapiArchived 2026.10.03Originally posted 2026.10.02
David Shapiro (L/0) @DaveShapiOriginally posted 2026.10.02
not sure what the fuss is about for OpenAI "dots" 1) It has no better taste than ordinary chat 2) it does not really follow instructions (it just does the same stupid shit Astra/Sol does, but with a user friendly interface) It really just proves that if the underlying model is garbage, no amount of layering on agents will fix that. Anthropic models have far superior taste, and no amount of specificity will fix just how god awful OpenAI's taste is.
x.com/daveshapi/status/2105992399457538138Archived 2026.10.03 10:04