A place for developers to talk about building on Farcaster.

/dev

if you're using an LLM agent to debug a tough issue that ends up going in circles i've found that it helps to tell it to insert console.log statements to test its assumptions

run your failure case and feed the debug logs to the LLM – works pretty well for narrowing down the root cause

One of the most useful use cases of coding tools like Claude Code and Cursor is to insert tons of logging, summarize what it figured out from the logs and then clean it all up.

Print debugging on steroids

I always start with logging and ask to verify with logs before assuming the code is right. Otherwise it tends to assume it fixed the problem and delete the logging afterwards lol.

I'm like hold your horses. I'll tell you whether the bug is fixed.

I always start with logging and ask to verify with logs before assuming the code is right. Otherwise it tends to assume it fixed the problem and delete the logging afterwards lol.

I'm like hold your horses. I'll tell you whether the bug is fixed.

https://farcaster.xyz/gabrielayuso.eth/0x5e52920a

The open web is dead, long live the open web. ⌐◨-◨

Building Farcaster

I don't use this much but I did use it in a one-off personal tool that I had an LLM write, anything that can be run by the LLMs themselves via commands to verify and get data themselves works wonders

Kind of along this vein - another trick I’ve been doing when getting an LLM to implement interfaces that interact with remote APIs is telling it to use curl to get sample data so that it can test as it goes

TDD is unironically pretty great for LLMs lol