[ad_1]
Too Long; Didn’t Read
Here are the things I learned in 6 months of working on the most powerful CodeGen dev tool GPT Pilot:
1. The initial app description is much more important than we thought
2. Coding is not a straight line, agents can review themselves
3. LLMs work best when they focus on one problem compared to multiple problems in a single prompt
4. Verbose logs do miracles
5. Splitting the codebase into smaller files helps a lot
6. For a human to be able to fix the code
7. They must be clearly shown what has been written and the idea behind it
8. Humans are lazy
9. It’s hard to get the LLM to think outside the box
[ad_2]
Source link