Research, experiments,
field notes.
Short essays on the systems we're building, the problems they expose, and the things we'd tell ourselves five years ago.
Why most MVPs fail before code starts
The cause of death is rarely in the codebase. It is in the three weeks before the first commit — where the team agrees to the wrong shape, ships the wrong scope, and pretends the constraints are not real. Notes from twelve years of running that interval well, and badly.
AI agents are workflows before they are intelligence
Every agent that worked in production turned out to be a well-designed workflow with a model wedged in. Every agent that failed was a model wedged in with no workflow at all. Notes from a year of building them — internally for Walter OS, externally for clients.
Designing sovereign medical data systems
Patient data outlives every company that holds it. We are building BioVault on the premise that the patient is the steward, the operator is the custodian, and no single party can take the record hostage. Here is what that means in code.
First note. Hello from the lab.
Two engineers, a network of senior operators, and a short list of problems we think are worth a few years of attention. Here is what Xipher Labs is, what we work on, and what to expect from this page.
A quiet
newsletter.
We don't mail unless we have something worth your time. One short note when a new essay lands. No drip, no funnels.
stored on a sovereign chain · operated by xipher.labs