GPT-6 Sol feels like a major downgrade from GPT-5.6 Sol, developer says
A developer says GPT-6 Sol makes frequent mistakes in domain modeling, data discovery and software architecture that GPT-5.6 Sol did not, citing the model's own line about starting at the top of the code listing.
GPT-6 Sol feels like a massive downgrade from GPT-5.6 Sol. 5.6 was genuinely great for intensive domain modeling, data discovery, and software architecture. It could understand the bigger picture without needing constant guidance. With Sol 6, I keep running into mistakes that I never got from 5.6. When the model itself says: > “I started at the top of the code listing, not the top of the model.” ...that pretty much sums up my experience with Sol 6. LOL😭 I really dk what happened between 5.6 and 6.