The approach
How the lab works.
Stoa Labs runs as a lab: open-ended exploration on one side, disciplined engineering on the other, and a one-way flow of evidence between them. Ideas get followed because they are worth understanding, before anyone asks what they are worth commercially. What survives scrutiny gets engineered into systems that have to hold where the work is real.
Mode one
Exploration
Open-ended investigation across seven research areas of the agentic stack. Success means learning something. An exploration is allowed to end with “this did not pan out”: that is a finding, not a failure, and the room for it is protected on purpose.
The research mapMode two
Development
Findings that survive scrutiny move into development: methods, systems, and eventually products, each claiming only the evidence state it has actually reached. What gets built sends new questions back into the research.
The bet behind the loopClaims carry their evidence
Every material claim we publish names its evidence state, evaluated context, method, and known limits. Our own systems are the first proving ground, and we say so. We will publish what we measure, including what did not work, because the strongest claim we can make is one that can be inspected.
The flow is one-way
What the lab learns flows into the consulting practice, never the other way around:
Research & development
Explorations, methods, tools, core IP
Methods flow →
Nothing returns
Consulting & integration
Independent methods, applied for clients
Stoa Labs develops its research, methods, and products independently from consulting customer information. Client data, workflows, traces, findings, and deliverables remain inside the engagement and are not used to train models, build shared datasets, develop products, or create Stoa’s core IP. Independently developed Stoa methods may be applied for clients, but client work does not become research material.Where the methods get applied