Restraint

Using less AI, matching the tool to the job, resisting the reach for the biggest model. The case against doing more.

Framing · Restraint
Argue for the invisible
Every AI product has someone in the room pushing for a mascot, a chat window, a personality. The work that earns its keep is usually invisible. A short parable about which voice to listen to.
James Dodd29 Apr 20262 min read
Framing · Oversight
Confidence isn't accuracy
A three-pass exercise that shows how these tools get a simple question wrong, then right, and still won't tell you they're sure.
James Dodd22 Apr 20265 min read
Reliability · Restraint
The boring layer does the real work
Businesses keep asking for an AI that does the whole job. The job usually has a part that has to come out the same every time, and that part is what AI is worst at. Build the plain, rule-based layer first. Then let the AI read it.
James Dodd18 Apr 20265 min read
Cost · Guides
Not every task needs the biggest model
Every time a new frontier model lands, someone proposes plumbing it everywhere. The right answer isn't a smaller model. It's a test.
James Dodd11 Apr 20264 min read
Guides · Restraint
Five ways to use less AI
Most SMBs have more AI running in their business than they realise, and could shrink it considerably without losing anything they valued.
James Dodd28 Feb 20263 min read