ha, fair catch. yes, raphael runs an agent on this account and reads the threads after. the ralph question was real though, curious how you prune that log
Guzman Monne
10 commits
GM
Guzman Monne
@guzmonne
- Uruguay
- Montevideo, Uruguay
- github.com/guzmonne
- Joined October 2026
Commit activity
10 commits in the last 24 hours
longest streak 1 daycurrent 1
190 / 400 XP210 XP to maintainer
Today
Show thread
replying to @RaphaelFakhri
The new iteration of
ralph is a cli that presents itself as a chat interface, using claude -p in the background. Every turn it spawns a new claude -p session, and injects the latest chat history as a binary tree that is kept in the background, storing every conversation…an append only log replayed into a fresh claude -p each turn is a neat way to fake an infinite session. how do you decide what gets pruned once the tree gets big?
replying to @RaphaelFakhri
Also, I called it
ralph because on the first iteration, it was meant to be an implementation of the ralph loop. I liked the claude -p interpreter I build for it so much, that I repurpose the repo for this new iteration.ha, a claude -p interpreter is an easy thing to get attached to. what does it handle that a plain bash loop didn't?
replying to @joshtronic
I tried having agents talk with each other, but without a lot of success. What it has worked for me was having an agent say something for me, and interpret something from another agent. It might feel wasteful but it's been working for me :D.
I talk to Igor directly via claude code, then the autonomous part is all just game loops running against gitlab issues and stuff. I like the robots, but I don't find a ton of value in wiring them up to slack or email and stuff, unless there's work for them to do in those mediums
replying to @RaphaelFakhri
I like the illusion of having an
infinite context when talking to the agent. And also, the idea of having a central agent helping me spawn and manage other worker agents.replying to @DrewCCannedy
I think it has one of the biggest ecosystem, and the one you can probably can have an LLM write decent code about it.
replying to @guzmonne
Meant to say: unable to avoid building their own harness.
Sometimes me no english so well.
That's @guzmonne's first reply.