Open to work. thinking about agent universe Previous built https://xitter.cf/t.co/dYt3JZ8hgM
Cupertino, CA
Joined November 2020
- Tweets1.1K
- Following231
- Followers188
- Likes1.1K
github.com/pingdotgg/t3code this is the best MIT license repo that teach everyone how to build the best harness product that users love. Just copy and steal it legally.
As a muse $50 user, it cost a 5hour session and just reply hello to me... All other tasks it keeps thinking and does not reply anything.
A smart model is not equal to a useful model
Muse code needs 2mins to reply hello from a brand new chat.
@theo it is not useable… @alexandr_wang can I get my refund?
I find the issue of the agent now. It rarely to no. So human need to design good harness to control it
Agents may should not talk to each other too often. I am trying to use GitHub issue and PR comments as the message board for now.
Running 10-30 threads/agents at the same time.
Building a production-grade agent harness with Codex, extra orchestration, and context taught me a humbling lesson: even making it reasonably usable is hard.
The bottleneck isn't AI. It's how well I understand the problem I'm asking it to solve.
Agent = scrapers
for a long time it felt like the world was fully optimized against you
you go to buy something, you get a bunch of ads and upsells and micro optimizations to get you to spend more
you go to consume some specific content and the algo puts in things that are hard to ignore and designed to get you to spend more time
agents are the first time i've felt like maybe things can go in a different direction. send your robot into that mess and it can ignore it all and do your task
i'm not naive, this is going to be an insane battle with entrenched interests doing their best to fight it
but there's a small chance that in the end we can finally be free of this (until the next new problem)
Current S tier harness:
using T3 code to let claude opus 5.5 be the manager. Then direct a fleet of codex sol 6.1 to do the tasks.
After months of running 3 ChatGPT Pro accounts, 1 Claude Max subscription, OpenCode’s free plan, 1 Mac, and 3 Ubuntu servers, I’ve come to a realization:
Most people aren’t limited by tokens.
They’re limited by ideas and persistence.
AI has made execution cheaper than ever. The real bottleneck is knowing what to build—and refusing to give up.
