Blogs

Notes from inside the transformations.

What worked, what broke, and what we would do differently. Written by the engineers who were in the room.

Let's Speak

Evals

LatestJun 20263 min read

Your eval set is lying to you

Offline scores drift away from production reality quietly. Here's how to notice it before your users do.

Read the article

Deployment
Apr 20263 min read
Week eight is a production date, not a demo
How we scope an engagement so the system is carrying real load before anyone books a steering-committee review.
Read
Governance
Feb 20264 min read
The audit trail is the product
In regulated sectors the hard deliverable is not the model; it is showing exactly why it answered the way it did.
Read

Earlier posts

Forward deployment
Nov 20253 min read
What a forward-deployed engineer actually does all day
Mostly listening, some plumbing, very little modelling: a week-by-week account of an embed that worked.
Read
Ownership
Aug 20253 min read
If you can't export it, you don't own it
A checklist for working out whether your AI stack belongs to you or to whoever invoices you for it.
Read
Agents
May 20253 min read
Most agents are a cron job with better marketing
Before you build an autonomous anything, check whether a queue, a schedule and one good prompt do the job.
Read

Work with us

We want your hardest problems.

We collaborate with ambitious organisations ready to move beyond AI experimentation. If you're looking to transform how your business operates, we'd love to build with you.

Start a conversation

Notes from inside the transformations.

Your eval set is lying to you

Week eight is a production date, not a demo

The audit trail is the product

What a forward-deployed engineer actually does all day

If you can't export it, you don't own it

Most agents are a cron job with better marketing

We want your hardest problems.