About
A small lab. We ship what we learn.
Tealbrick is an applied research lab on how people work with AI agents.
The model isn’t the hard part any more.
The hard part is making agents useful, understandable and safe in real work: what they remember, what they may do, who they ask, and how anyone can tell what happened.
Our long-term goal is one steady relationship between a person or a company and their agents, across knowledge, tools, decisions and devices. That’s too big to ship as one product, so we work in small pieces. When a problem keeps coming back in real deployments, usually our own, and is specific enough to solve cleanly, it becomes a mini app.
What we hold ourselves to
- Dated findings
- We write down what we expected, what failed and what we changed, and we date it. If we haven’t measured something, we tell it as a story, not a number.
- Small pieces
- Each mini app solves one recurring problem and is useful on its own. Start with one, and add the next when a client needs it.
- The client owns it
- Mini apps deploy to the client’s own server account, and the code is open source. The data stays theirs.
Two founders, two bricks.
Eyal leads the company. Martin leads engineering. We run agents in our own work, sit in on client deployments and write the lab notes. Message either of us on WhatsApp from the links below; we answer ourselves.
Read the lab notes