Why we’re building Tulina.

Watch a good engineering team work with AI and you notice something: the AI is genuinely useful. It knows the codebase. It knows the conventions, the decisions, and the reasons behind them. It picks up where anyone left off.

They have the same models as everyone else. The difference is what they gave their AI: a shared memory. Plain files that describe how things work. Version history that records every change. One source of truth that everyone reads from and writes to. That setup is a big part of why small engineering teams move so fast.

The rest of the company runs differently. Knowledge sits in static documents that rot, in a dozen tools that never talk to each other, and in people’s heads. The person who understands the pipeline review is the only one who can run it. New joiners work in the dark for months. The sentence we hear most often when we interview operations teams is “I don’t understand how decisions are made.” And every AI chat starts from zero, because the context it needs was never written down anywhere it could reach.

Teams try to close the gap with the tools available to them. Chat assistants with connectors can reach the data, but reach is not judgment: they can’t tell which numbers are canonical, which sources are dead, or why the process works the way it does. Automation platforms ask business people to think in flowcharts, then to hire someone who thinks in flowcharts for them. Knowledge bases hold whatever you type into them, then quietly go stale.

Some teams go further and borrow the engineers’ setup outright: folders of plain files, a repository, scheduled scripts. It works. It also demands branch hygiene, credential management, and a terminal. That infrastructure was never built for the people who run sales, ops, or finance, and asking them to adopt it just to use AI properly is backwards.

So we are building the infrastructure for everyone else.

Tulina connects to the tools your company already uses and writes down how the company works: definitions, processes, decisions, each page linked to its source and dated. The context stays current because Tulina maintains it on its own, folding in every new decision and every lesson as they happen. And the context acts: your team runs daily work from it in plain language. Every run adds to what the company knows. Every decision sharpens it.

We believe a company’s context should belong to the company: owned, current, shared, and useful to every team, whichever AI sits on top of it. Companies that work this way will compound. Companies that keep their context in heads and stale documents will keep starting from zero.

If that sounds like the company you are trying to build, we would like to build yours with you.