Blogs
Notes from the practice.
What we have learned building AI systems and software for organizations that cannot afford for them to be wrong.

Who is responsible when an AI system causes harm
An AI system does not need to become conscious to cause damage. Ethical development means limiting its authority, testing foreseeable failures, and assigning responsibility to the people who can prevent harm.
, 3 min read
Read the article
AI systemsAgentic coding is having its home renovation moment
AI coding tools make software easier to start, much as home improvement stores and YouTube made renovation more accessible. The businesses most exposed sell execution alone. The stronger opportunity is to own the risks that remain after the demo works.
, 4 min read
AI systemsShip the rule, not the model
Roughly half of what arrives described as an AI problem is a rules problem with bad documentation. Telling the difference early is the cheapest thing we do.
, 6 min read
AI systemsWhat an evaluation set costs, and what it saves
Nobody budgets for the evaluation set. It is the single highest leverage artefact in an AI engagement, and the only reason you can change anything later without fear.
, 7 min read
Custom softwareRead the legacy system before you replace it
The undocumented behaviour in an old system is not an accident. It is usually twenty years of edge cases somebody handled correctly and nobody wrote down.
, 5 min read
AI systemsThe approval card
If an automated system can send, publish or merge on your behalf, the interesting design problem is not the action. It is the refusal.
, 4 min read
PlatformMigrations on systems that cannot stop
Expand, migrate, contract. Three deploys instead of one, and every intermediate state is a state you could stay in indefinitely.
, 5 min read
Work with us
Prefer this in practice rather than in prose?
You will speak to an engineer, and you will leave the first conversation with an opinion about your problem whether or not you hire us.