AI in production fails in three ways: a vendor goes dark, a job fails silently, data leaves your control. This is the record of the rulings I made against each one, on my own stack, with what they cost and what broke. Where something is still untested at scale, I say so.
Build automations with AI
Keep AI output deterministic: let the model decide, let code render
Model output you cannot reproduce is output you cannot sign: the same input comes back different, and nothing in it can be diffed. The rule here takes rendering away from the model, which returns JSON while code renders all a reader sees, and names the model swap that fixed French regressions.
Reliable Agent Systems
The assistant’s self-improvement loop rewrote a rule I had just set
Your last instruction is not the last word on a rule your assistant can edit by itself. This piece shows why a rule stops applying, and what makes one stick: a single writer, an edit proposed before it is written, and a ledger that undoes any of them.
Sovereign & local AI
Cloud by default, local when the data demands it: what runs on my machine and what needs the cloud
Where a task runs decides who can reach its data. This piece lays out why my agent moved from a local default to a cloud one, the data-class rule that keeps short, easy private work on the machine, why the content team never runs locally, and what a 24 GB machine will actually hold.
