We build production software the way it should be built — with care, with judgment, and with humans owning every decision.
The name is plain: Puyosa, a family name, and a software lab. Nothing more.
Run by Moisés Puyosa, a senior software engineer — remote, GMT-5.
The soul behind it
We're a small lab that believes software is craft, not output. We choose quality over pressure, study what we don't yet know, and use AI as a sharp tool that sharpens our thinking, not a substitute for it. We have judgment, we have standards, and we choose depth over volume — things that work and keep working.
How we work — and why
Sustained quality comes from a deliberate pace, not panic. We design first, decide deliberately, and let the work earn its complexity.
Speed without quality is just debt with a deadline. We move fast where mistakes are cheap and invest carefully where they're costly.
For anything non-trivial, the thinking is written down before the code — an implementable design, not a vibe. Decisions you can audit later.
Clean boundaries, honest abstractions, no clever shortcuts that bite the next maintainer. We abstract on the second instance, not in anticipation.
If tests fail, we say so. If something's a mismatch, we name it. Output is information, not decisions dressed up as conclusions.
AI, used well
AI is part of how we build — structured extraction, retrieval, evaluation, agent orchestration, pair programming. But we don't ship what we can't read, can't test, and can't defend. The tool is fast; the judgment is ours.
AI drafts and accelerates. We review every line, understand it, and own the result. No "the model wrote it" as a defense.
We build the deterministic, no-network paths to test AI systems for real — evals with golden sets, not hope. Quality is measurable here too.
PuyoLabs Learning
So we study — deliberately, continuously, in the open. New stacks, new patterns, the traps that only show up in production. Learning isn't a phase before the work; it's part of the work.
Every gap we find is the next thing we close.
Where we run
Our work runs on the open internet and on our own hardware. Both come with the same duty: keep it safe for the people who depend on it.
We run real software on real infrastructure — self-hosted, with real data and the cost and uptime that implies. Not toy demos: production is the only honest test of whether the work holds up.
A note to the curious and the capable: we're an independent lab — lean, self-directed, accountable to the craft rather than a quarterly roadmap. If you find a hole, disclose it responsibly: raise an issue on GitHub and we'll fix it fast and credit you. We'd rather work with you than get a surprise — security researchers are welcome here.
Work with us
We love results, and we love getting them well. Available for new work — contract or full-time, remote on GMT-5. If you want software built by people who care about the craft and the outcome in equal measure, let's talk.
A free 30-minute call — bring your problem, leave with an honest take on scope, approach, and fit.