Who LLMs the LLM?
Level up your code quality with a 3-LLM review process! Learn the 6-step workflow using ChatGPT, Claude Opus (via Replit), and Gemini to tap into the collective intelligence of current state-of-the-art AI. A tiny team's secret weapon for better code, faster.
With a tiny team, vibe coding definitely helps you move very fast, and code quality keeps getting better. But you can really level up by using multiple LLMs to ensure quality.
I’ve already mentioned how I use ChatGPT to provide input on the development of Burrow in Replit (using Claude Opus), which gives me the benefit of two LLMs. The next level is using 3 LLMs for code review.
How this works:
- Refactor using the guidance from the previous blog post and using Opus.
- Check the results into GitHub
- Access GitHub via ChatGPT’s code analysis tool and have it perform a code review
- Take the output from ChatGPT’s feedback and ask Replit to consider the recommendations
- After updating in Replit, check it back into GitHub and repeat until ChatGPT and Replit with Opus are happy with the outcome.
- Repeat Steps 3-5 with Gemini
Following these steps - especially before a major release - gives you the best of all worlds by leveraging the collective intelligence of the current state-of-the-art AI technology. Having a Human-In-The-Loop (HITL) as part of this process is still a massive benefit (if at all possible), but technology keeps improving, as will best practices.
