01The Lab
Product × Research.
Applied AI is two practices in one room: shipping things that hold up in production, and writing down what we learn while we do it. Both columns below — kept honest, no roadmap theater.
Flagship productBeta
Vertlock
The governed gateway between AI agents and company tools. Scoped permissions per team, one-tap human approvals for dangerous actions, credentials the model never sees, and a tamper-evident receipt for every call.
Blocking AI agents loses the race. Letting them loose loses the audit. Vertlock is the third option.
hash-chained receipts — even admins can't rewrite history
In Production
Client agent suites
LivePrivate AI-agent deployments for operating teams — built, governed, and run through the consultancy practice below.
In Research
- Agent harnessesLive
How the scaffolding around a model — tools, memory, gates — determines whether agents hold up in production.
Why long-running agents degrade, and the feedback structures that keep them on task.
03Consultancy
AI, into the everyday flow.
The other half of the lab. We sit with teams that want to integrate AI into the actual work — not the keynote version, the daily-loop version. Where it earns its keep, where it breaks, what to roll out first, who on the team is going to run it on Monday.
- Operational audit
- Build-vs-buy roadmap
- Tooling + workflow design
- Hands-on team enablement
Working on something adjacent?
We collaborate with teams whose problems push our research forward. Tell us what you're working on.