CONSULTANCY · ENGINEERING · APPLIED AI

We build, run and reason about complex systems.

For organisations that would prefer their infrastructure stayed boring. Democratising data with AI, since before that meant anything.

See what we do Our products

fig. 1 — a system, simplified beyond usefulness

01 /

What we do

Three practices. One invoice.

Consultancy

We interview your teams, read your architecture diagrams, and tell you which one is fiction. Recommendations arrive in a document short enough to be read.

deliverable: clarity, laminated

Engineering

Software that ships, scales, and survives the person who wrote it. We are comfortable in legacy codebases. Someone has to be.

deliverable: working software, tested

Applied AI

Models where they help, rules where they don't, and a written record of which is which. Democratising data, responsibly vague.

deliverable: an AI strategy you can defend

02 /

How we work

A method. We named the phases.

PHASE 01

Discovery

We ask questions until the real problem admits itself. Usually in week two.

PHASE 02

Cartography

We draw the system as it is, not as the wiki says. The two diagrams differ.

PHASE 03

Intervention

Small changes, placed precisely. Large changes, only when small ones are found to be insufficient in writing.

PHASE 04

Stewardship

Eventually we leave. The system doesn't notice. That is the deliverable.

03 /

Products

Internal tools we decided to name.

Driftwatch

PRIVATE BETA

Infrastructure that tells you when it stopped being itself.

CONFIG DRIFT · LIVE · click a row to acknowledge

Hindsight

WAITLIST

Post-incident reviews, pre-incident.

INC-0042 · SEV2 · RESOLVED, NARRATIVELYpinned · click to unpin

Quorum

COMING 2027

Consensus for teams that don't have any.

ROUND 7 OF ∞consensus: pending

votes: 3 aye · 1 nay · 1 "circling back"

Slowpath

PRIVATE BETA

Observability for the things nobody instruments.

P99 · UNATTRIBUTEDunits: withheld