A small group that treats "it works" as a hypothesis, not a conclusion.
PEGATO Labs runs three systems (Tragent, OCE, and Parallax), each solving a different problem, on a different stack, at a different stage. What holds them together isn't shared code. It's a shared answer to one question: how do you know it actually works?
Built on one non-negotiable premise.
The discipline PEGATO Labs runs on isn't an abstract ideal. It's a hard line: a claim is not evidence, and a system that can't tell the difference will eventually act on something that isn't true. Every rule that follows exists to keep that line impossible to blur, in any system, at any stage, no exceptions.
That's why every system here, regardless of what it does, is built and checked the same way.
Three systems. Zero shared runtime.
Tragent, OCE, and Parallax each keep their own infrastructure, their own port range, and their own release cadence. The only things they share on purpose are a local model runtime and the verification standard every report is held to. If one system needs to change something about how it runs, it can, without a meeting and without risking the other two.
Six rules, applied to every system, every sprint.
A claim needs a number behind it
"Works well," "production-ready," and "fully functional" don't get written down as evidence. They get replaced with the actual measured result, or a worked example someone else can re-check by hand. If something can't be verified yet, the report says so plainly instead of rounding up.
Blocked beats fabricated
If a dependency isn't available to test against, the report says it's blocked, never estimated, never simulated. A number presented as measured has to actually be measured, in every system, without exception.
State lives on disk
The current position of any project is a file, read fresh each session, never reconstructed from memory or re-typed from scratch.
One owner for scope
Architecture and roadmap decisions have a single accountable owner per project. Implementation can flag a concern; it doesn't redefine the plan mid-sprint.
Contradictions get caught before the next step
A report that states a rule and then shows an example that breaks it gets stopped there, every time. Contradictions don't get resolved by picking whichever version sounds more finished.
Independence is a feature
Every system is built so a problem in one can't quietly become a problem in another. That's a design constraint, not an accident of how things ended up.