@benjaminbalazs One of the people I met was interested in RL environment we provide for Ai agents. Another couple of them were interested in obsrv. So thatβs a huge win for a start up like us.
Still remember the match against Rafa when he got injured and then came out full of tears with Rafa helping him hop into the court with a badly injured leg. Very happy to see him lift the Rolland Garros finally.
@buildwtim Yes itβs true. And in the meetup I had, few of them were genuinely interested in the product we are building and were more than happy to test it. Thatβs what got me excited.
@AlphaCoreict Hey, I was in tech sales in Dubai for over 10 years, I just leveraged my contacts here to connect with people who were looking for the RL Environment that we are developing at theta lab and I got traction. Nothing to lie about it here. I can only talk from personal experience
Watched a great engineer scroll 4,000 agent traces to find the one that broke. Took an afternoon. Found it at ~3,100.
We have models that can read all 4,000 in seconds.
The scrollbar is not a debugging strategy.
i lose everything cool i find online. tweets, research papers, screenshots of designs straight into the screenshots graveyard.
so i built ctrl+brain.
select anything β ββ§2 β saved to one markdown file on your mac.
no selection? it screenshots, reads the text out of it, and a local ai describes it. all on-device.
want it everywhere? syncs to @supermemoryai so any ai can recall it.
free, open source:
ctrlbrain[.]xyz
Dashboards are the wrong interface for AI systems. They work for metrics you can name in advance: CPU, latency, errors. You can't pre-draw a chart for "misunderstood the user on turn 7." The interesting agent failures are the ones you never thought to instrument. Wrong interface.