i.
Falsifiability before flourish
Every engagement opens with a success criterion that the partner can use to dismiss our work if we fail to meet it.
About
Independent research, operating where the work is done.
BrainBrowser is a small, independent firm. We are deliberate about who we work with, what we measure, and what we'll put our name to.
Why we exist
The frontier AI labs ship faster than the safety literature can keep up with. The enterprise security teams adopting these models don't have the in-house tooling to evaluate them. The governance bodies trying to write policy don't have the reproducible findings to ground it.
BrainBrowser exists to do that work at engagement scale — small, specific studies, finished on a deadline, with artifacts that outlive the engagement.
Operating principles
ii.
Artifacts over essays
We will write the code, the rubric, and the transcripts before we write the executive summary.
iii.
No threat inflation
We say what we found, with the severity rubric in the same document. If the finding is small, the report is short.
iv.
Reproducibility is trust
Anyone in the partner organization should be able to re-run the study without us in the room.
v.
We refuse work we can't measure
Retainer engagements that don't produce a falsifiable artifact are not the shape of our practice.
vi.
Reputation is the asset
We won't spend it on engagements that would compromise it. We turn down more work than we accept.
Who we are
BrainBrowser is led by researchers with backgrounds in adversarial machine learning, applied cryptography, agent infrastructure, and enterprise security. The team is small by design and unlikely to scale beyond the engagements we can personally run.
We do not maintain a research blog. We publish through partner engagements and through co-authored work with academic collaborators where permitted.
Detailed bios available on request after first-call alignment.
Small enough to know every finding. Serious enough to back it.