Explainability
Why your AI flag needs a reason attached
A risk score with no visible basis is a liability. What a defensible explanation actually looks like in a production system, and why post-hoc justification is worse than none.
12 min read
Insights
What we have learned, written down.
Notes from delivery rather than commentary on the news. Mostly about the parts of AI work that are unglamorous and decide whether a project succeeds.
live
Delivery
Build the baseline before you build the model
Most AI projects skip straight to the sophisticated approach. Why we always ship a cheap baseline first, and how often it turns out to be enough.
8 min read
Education technology
What information governance teams actually ask about AI
The questions that stop education AI procurement, in the order they get asked, and how to have the answers ready before the meeting.
10 min read
Engineering
Multi-tenant isolation is a schema decision, not a feature
Retrofitting tenant isolation into a live platform is one of the most expensive mistakes in SaaS. What to get right in the first migration.
9 min read
Search
Technical SEO for products nobody is searching for yet
When your category does not have search volume, keyword-first strategy fails. What to build instead, and how to measure it.
7 min read
Automation
Why we refused to automate LinkedIn with a browser
Browser automation breaches platform terms and risks a permanent restriction. The compliant alternative took longer and is still the right call.
6 min read
Articles are published as they are written rather than to a schedule. If you want a topic covered, say so.
Get these when they are published.
No newsletter cadence and no drip sequence. Email us and we will add you to a list that sends when there is something worth reading.