Engineering notes
Engineering notes for decisionsthat outlast a release.
Practical writing on product systems, data change, integrations, and the trade-offs behind software that must keep working.
Notes ledger
8 issued
What the first week of an engagement actually looks like
After the quote is signed, the useful question is not “when do we see screens.” It is what becomes true in the first five working days so the rest of the work can be reversed if we are wrong.
Read note →AI in a real workflow is mostly the failure path
A demo shows that a model can answer. A production workflow has to decide what happens when the model is wrong, slow, expensive, or unavailable — and who is accountable for that decision.
You do not need a spec to start a useful conversation
Most worthwhile software work begins as a symptom, a deadline, and a system nobody fully trusts. A complete specification is not a prerequisite for a serious first conversation.
A first release is not a product
A launch can prove that screens exist. A product is the workflow, the data, the permissions, the errors, and the operating path that keep working after the announcement.
What a useful handoff actually contains
A delivery that cannot be operated by the people who remain is not finished. A useful handoff is a set of artefacts, access changes, and decisions — not a goodbye email.
Make the next data change reversible
The scary part of a migration is rarely the new schema. It is the hour when old and new must both be true, and the moment you discover they are not.
Integrations fail at the seams
Connecting two systems is easy on a whiteboard. The work is what happens when one side is late, twice, partial, or using a field to mean something else.
Work we decline, and why that helps you
A selective partner is not a luxury. It is how you avoid paying for the wrong shape of help — and how we keep senior attention on the work we actually take.