Vic Valenzuela
Everyone's letting AI write their code now. I work on how you know you can trust what it built.
Fifteen years shipping software where mistakes are expensive: auction houses, prosthetic knees, healthcare scheduling. Engineer in Reykjavík.
Writing
Latest posts
I write about making AI-assisted development dependable: trust, memory, process, and the harnesses around coding agents. Start with the externalized trust series.
-
Externalized coordination: the missing primitive for agent swarms
agentic-aiharness-engineeringsoftware-factories -
Externalized proof: confidence is cheap
agentic-aiharness-engineeringdeveloper-tooling -
Externalized process: optional becomes skipped
agentic-aiharness-engineeringbeislid -
Externalized memory: the work has to remember
agentic-aiharness-engineeringmemento-vault -
Externalized trust and the road to software factories
agentic-aiharness-engineeringsoftware-factories
Projects
Things I've actually shipped
From a healthcare workforce platform to AI engineering tools, with prosthetics, auction houses, and radio along the way. The tools are the lab the writing comes from.
dala.care
Iceland's healthcare workforce had no modern coordination layer. We built one.
Fundið
Iceland's lost & found was Facebook groups and police Pinterest boards. I built something better.
Beislið
A human-centric framework for collaborating with coding agents. Hold the reins while agents do the work.
Memento Vault
AI dev sessions produce knowledge that vanishes when the context window closes. Memento captures it.
Rondo
AI agent orchestration is opaque by default. Rondo makes it observable.
About
The agent is the horse, I build the reins.
Mexican-born, Icelandic citizen, based in Reykjavík. For three years I built dala.care's healthcare workforce platform, before that I shipped consumer and healthcare software for over a decade, and lately I've been deep in agentic development, the harnesses and tooling that make coding agents dependable.
All of it circles the same problem, shipping fast with AI and still being able to prove the work is right. That's what I write about, and what teams bring me in to help with.
Newsletter
Get the next post in your inbox
I write about agentic engineering, the harnesses, skills, and trust layer that make coding agents actually dependable. No schedule promises, no spam.
Prefer a reader? Grab the RSS feed.