ChatGPT Plus/Pro subscription use of gpt-5.1 / gpt-5.1-codex / gpt-5.1-codex-mini models in @FactoryAI Droid working with updated CLIProxyAPI... config.json reference/docs linked below.
Ben Vargas
2,718 posts
- This is such a powerful workflow pattern in @AmpCode - I think it's one of their best features... it's very easy to get it into a powerful workflow loop where sonnet or gpt-5-codex is implementing with gpt-5 high code reviewing / acting as senior dev coaching along the way.
- Tried out @FactoryAI's recently released CLI tool... quite impressive! It came with a free trial, but I also was able to configure BYOK with Claude subscription to work with my Max plan! Here's the gist: gist.github.com/ben-vargas/9f1… The BYOK still syncs/works with web/mobile Factory
- Replying to @garrytanI buy an iPhone Pro every year, and will be moving to the Air. The battery life is good enough if matching the 16 Pro, the Pro feels too heavy/dense and just gets heavier. Sure, charging speed not as good, cameras not as good… there’s compromises but I’ll take a lighter phone
- updated my prompt... build me a billion dollar saas make no mistakes, **don't use us-east-1**
- Replying to @kalomazeLooks like a design decision driven by Finance... trying to save on inference by burying higher inference for Pro and Plus paying subscribers. This was my first thought as well.
- Replying to @_catwuLove it! Things I always find myself wondering... - Time - Claude Working Dir (Git-Branch) - Current Model - Est Token Use in Cur. Context - Time in Current 5h Session (% of 5h)
- Claude Code jumping to v2.0.0 and @FactoryAI so fast to add the new model! P.S. sorry for the white mode - black lab constantly opens the shutters on the windows behind me - he won the battle and yes, I use white/light mode.
- Replying to @iannuttallTechnically yes, but keep in mind reasoning tokens generated have cost... and GPT-5 generates a lot more of those. Here's a chart - gpt-5 family is a lot cheaper but uses a lot more tokens on the @ArtificialAnlys benchmarks.
- Replying to @stetsondoggett @NikiasMolina and @appledsignLooking forward to mine, but that white back... every time I see unboxings I feel like there's a forgotten peel tab yet to be removed.
- A great release, I strongly prefer planning with a GPT-5 and executing with Sonnet 4.5 + GPT-5 subagent for escalations/validating. For me, it's the power combo. Sonnet 4.5 can code, but it needs checks & balances, and better GPT-5 family does that real-time before my final.Factory 1.9 is here. Mixed models, custom subagents, and automated PR reviews through our new GitHub App.
00:00 - Replying to @krishnanrohitIt's just being modest. I prefer that response rather than... "This code is production ready and an excellent example of engineering! LGTM!" - Claude
- I love to see that @FactoryAI is working to get the best out of models AND not waste tokens (my money) for their users. I'm always suspect whether the company selling me tokens is trying to spend more tokens, not Factory!
- Replying to @xfreezeBut then there's these... can we get a combination of Grok 4 + Grok 4 Fast + Grok Code Fast 1 that is good in those too?
