Engagements

Scoped work, not open-ended consulting

Each engagement states what you get and what it produces. We have deliberately not attached fixed durations or prices, because we would be inventing them — scope is agreed per engagement.

Available now

Built on technology you can already verify

Available

AI Opportunity Assessment

For: engineering leaders deciding where AI is worth applying.
You get: a written assessment of candidate problems, ranked by feasibility against effort, with the ones we think are a bad idea named explicitly.
Produces: a decision, not a proposal.

Available

Private AI Evaluation

For: teams blocked by data residency or contractual restriction.
You get: local inference running in your environment on a representative task, with an honest read on whether output quality is sufficient for it.
Produces: a go / no-go backed by your own data.

Available

Agent Architecture Review

For: teams deploying agents with access to systems that matter.
You get: a review of your agent design against the influence-governance questions in our research — who acts, what they may reach, and how much one input can shape a privileged action.
Produces: a risk picture. Vendor-neutral; the fixes are usually in your architecture, not our products.

Available

Custom Development on NC

For: organisations wanting a specific system built on our stack.
You get: a working system, with source, built by the people who built the language.
Produces: software you own and can run without us.

Pilot-first

Where we start you in observe-only mode

Beta

Autonomous Operations Pilot

For: SRE and platform teams with a high volume of repetitive investigation.
How it runs: read-only and in shadow mode first. The agents investigate and propose; nothing touches your infrastructure until you have seen enough to extend that trust.
Produces: evidence of whether agent-led investigation is useful in your environment.

Beta

Agent Governance Pilot

For: teams giving agents write access to systems that matter.
How it runs: AGP sits in front of your existing agents in observe-only mode, recording what it would have refused without blocking anything, so budgets get calibrated against real traffic first.
Produces: a concrete picture of how often an input was steering a privileged action more than you assumed.

How we work

Three commitments

Scope before start

What you get and what success looks like is written down before work begins.

We will talk you out of it

If your existing stack is the better answer, we would rather say so than sell you an engagement that will not help.

You keep the work

Source and outputs are yours, and run without a dependency on us continuing to exist.

Start with the problem

Tell us what it is costing you and we will tell you whether we can help.