π“Ÿ host1.bitchan.one

𓀐 @Evi=EQ8R

⏱ 4d seenβ›“ 0πŸ‘ 0β§– 1

filter [-]

languages:onother

In you guys' experience, which is "better" (faster, higher quality, etc) : One shotting with the larger slower more intelligent model, OR Research + plan with slow smart model, implement with fast dumber model, audit/revise and fix implementation with larger smarter model? I like the second one because often times the the audit step tends to find gaps beyond just the original plan, and even beyond the spec, but it does take more steps and is slower overall. I guess one shot (no formal research and plan step) + audit could also be an option. Or one shot with the smaller dumber model then audit with the larger slower model. >tl;dr : What's your vibecoding workflow with local models⏱ 4d↩