Meridian licenses the operational record of real companies — the messages, the tickets, the codebases — to the labs building the next decade of intelligence.
For two decades, we taught machines to read. They learned grammar from encyclopaedias, code from public repositories, taste from forums.
Then we asked them to reason. Public text stopped being enough. Models hit a ceiling against problems no encyclopaedia had solved before.
Now we are asking them to work — to plan, to coordinate, to ship, to follow up six weeks later. And the data that would teach them is locked inside the day-to-day operations of real companies.
That is the data Meridian licenses. Not the internet's reflection of work. The work itself.
Public text is exhausted as a frontier. The next leap doesn't come from another trillion tokens of HTML — it comes from learning how a Series B founder rewrites a brief at 2am, how an engineering lead unblocks a stuck PR, how three teams coordinate a launch across six tools.
Most operating data depreciates the moment it's written. Meridian gives it a second life — as licensed training material, with the company compensated and the privacy of every individual preserved.
Every record is anonymised before it leaves the company. Names, emails, identifiers, secrets — stripped at source by tooling we own. Labs never see who. They only see how.
If your company is three years or older and generates real day-to-day work across messaging, project, and engineering tools — the program may be a fit.
Operational data is the highest-yield, lowest-correlation asset your portfolio companies already produce. Meridian operationalises it without disturbing the cap table.