Who We Are
Brood is small.
Our agents are many.
This background was harvested from a failed experiment in extracting correctness signals from model entropy.
We would rather fail at a hard problem than polish an easy one, and we’ll keep trying.
A few of us
Senior people stay on the problem from first conversation through production. The shortest distance between understanding the problem and changing the code is one person.
A lot of them
Agents do the parallelizable work: research, testing, evaluation, documentation, and iteration. More work gets done; responsibility does not get delegated.
How We Work
Most enterprise AI starts backwards.
It starts with a job or a model instead of a specific problem, uses AI where ordinary software would be faster and easier to verify, then mistakes adoption for success.
State the problem. Do not prescribe the method.
Bring us the pain, the constraints, the stakes, and what good looks like, not the tool you think you want.
Automate the task, not the job.
We use AI to supercharge your mission. We don’t automate away jobs, make AI ads, or deliver slop. We break the work apart until we find the places where judgment actually matters, and use AI there.
Software everywhere else.
Most of what people hand to an LLM is deterministic work in a costume. Retrieval, joins, calculations, checks, reconciliations, and routing get ordinary code: auditable, repeatable, fast, and cheap. AI earns its place only where judgment remains.
Production is the test.
A demo proves that something is possible. Production proves that it is useful. We build against a real user, a real workflow, and a definition of done that can be measured.