The lab · internal R&D
Internal projects, published as evidence rather than sold as products. Anyone can call themselves an AI studio; a working system is harder to fake.
A private, single-user AI that lives on hardware we own.
One cockpit for all your AI coding agents. Claude Code, Codex CLI, opencode and Gemini CLI side by side, each on its own branch, with alerts, review and best-of-N merges.
Lab work you can try yourself: an AI Document Reader that shows its evidence for every field, and a Domain & Email Health Check for SPF, DKIM, DMARC, HTTPS and security headers.
Open the tools →Some lab work grows into a product. SyntoraData, our company-data product, is built and run by the same two people: registers, identifiers and a polite crawler, with the source kept on every fact.
About SyntoraData → syntoradata.com →by design, and not for sale
a few retrieved for each turn
composed for each turn
where her identity lives
Why publish this
A lot of what is sold as AI engineering is a hosted API behind a form. That is fine where it fits, but it says little about whether a team can handle the parts that are genuinely hard: training small models of your own, proving a change improved things before it ships, keeping a system honest when a tool did not run, and keeping it working once nobody is watching.
So rather than claim we can do those things, here is the system where we did them, and what it taught us.
The lab work is not for sale, but the capability is. If your data cannot legally or commercially leave your building, we build AI that does not ask it to.