Carlos D. Escobar-Valbuena

@broomva_tech

Building the control layer to make unattended agents safe · Chief AI Officer @ Stimulus

broomva.tech

Joined November 2009

293 Following

86 Followers

Pinned
Carlos D. Escobar-Valbuena
@broomva_tech
Mar 17
Replying to @broomva_tech and @JayScambler
The harness is not incidental. It IS the control substrate. If you want an LLM to actually control systems, you need what classical control theory assumes implicitly but agent frameworks skip entirely: State — not chat history. A typed canonical representation: current
2
Carlos D. Escobar-Valbuena
@broomva_tech
Nov 13
Inspired by @rauchg and @Sci_Phile, tried the same trick on a cloud chamber, after Kyle Hill's video. There's no track-drawing code: ions seed droplets, the droplets grow and fall, and the tracks are what's left. 75 tests check it against NIST and PDG. cloud-chamber-eight.vercel.app
Guillermo Rauch
@rauchg
Nov 8
01:28
Built a physical simulation of a record player with WebGL2 and Web Audio. It's *shockingly* accurate: vinyl-gl.vercel.app. Instead of playing a song, it works by converting it into the physical shape of a groove. Then a simulated stylus rides those walls 384,000 times a
Carlos D. Escobar-Valbuena
@broomva_tech
Nov 10
Everyone leads a fleet now. Five terms decide whether it ships the right thing: Intent: the outcome, and why Bounds: irreversible work waits for a yes Evidence: done means the check passed and you ran it Escalation: asks come with options Review: you don't certify your own work
GIF
1
Carlos D. Escobar-Valbuena
@broomva_tech
Nov 10
How we build with agents, in 2 minutes. Philosophy → design → architecture → structure → standards, then one real change traced end to end: owner ask → gate → isolated agent → cross-model review → policy merge → sensed outcome. Every rule leaves an artifact.
1
Carlos D. Escobar-Valbuena
@broomva_tech
Oct 26
Be bold, explore and let the agents work. I think the progression of idea to materialization is allowing ever growing crazy ideas. The democratization of knowledge and agency it’s going to have one if not the largest impacts on humanity ever. I foresee a (close) future where everyone can learn and build about anything they want and the abundance of wellbeing and wealth would just make the flywheel go even more
Thorsten Ball
@thorstenball
Oct 25
Finally summarized what I've been sending to friends in DMs
Article
How I work with Agents
... as of October 2026. Use the latest and greatest models. Everything else is potentially a waste of time. If a model is wrong 10% of the time, you don't know which 10%, so you end up making multiple

Carlos D. Escobar-Valbuena

Carlos D. Escobar-Valbuena’s posts