We designed it around the principles:
1. Leverage LLMs where they work best: LLMs are faster than humans at putting together prototypes of a few hundred lines of code. By using them early on, and give high level change request commands when the project is small, and then let human devs take over, it’s a productivity boost.
2. non-technical + programmer collaboration: All changes from the app is made as git commits, so that non-technical people (via chat) and programmers are able to contribute to the same code.
LLMs are expensive, but we left it open over the christmas holidays for anyone to try it out: https://run.gptengineer.app/
Would love to hear what you think!