The Production Doctrine

I've written four pieces about building AI in production, and somewhere in the fourth I noticed they were never four arguments. They were one argument, seen from four sides.

So here it is in one place, as plainly as I can put it: five things I've come to believe about shipping AI that actually works. None of them came from a paper. Each one came the same way: running Easy AI at scale and being wrong first.

Each one is really a decision rule, and each points at the piece that earns it.

The Production Doctrine

1

The model is swappable

It predicts the next token and nothing else. The intelligence is real, but it arrives in a download — and you'll replace it within months. → The Harness

2

The harness is the product

Reliability lives in the loop, the tools, the memory, the gates — the runtime you build, not the model you download. → The Harness

3

Context is the budget

The model re-reads the whole conversation every turn, so you rent attention by the word. Cost and reliability turn out to be one line. → The Economics of a Token

4

The rebuild is the design

A harness is a pile of assumptions about what the model can't do yet. Build it to be replaced, not to last. → Built to Be Replaced

5

The outcome is the only delivery

Shipped measures your end of the wire. Delivered measures the other end — the only end that was ever the job. → Delivered

Reach for this list the next time you're about to spend three weeks chasing a smarter model or a bigger number, and check first whether the thing hurting you is the model at all. Most mornings it isn't. It's the harness, the context, or the metric, and those are yours to fix today. Start from the symptom you actually feel:

Start from the symptom

The agent is flaky in production

Wrong instinctShop for a smarter model

First moveFix the loop, tools, and gates first

The harness is the product

The inference bill climbs as you scale

Wrong instinctNegotiate a lower token price

First moveStop sending words that aren't earning rent; cache the rest

Context is the budget

A better model landed and upgrading hurts

Wrong instinctKeep the scaffolding you're proudest of

First movePull the model-specific cleverness out from behind its seam

The rebuild is the design

The dashboard is green but nothing moves

Wrong instinctCelebrate the output metric

First moveMeasure a number at the customer's end of the wire

The outcome is the only delivery

You're betting the roadmap on the next model

Wrong instinctWait for more parameters

First moveWire today's model in as a part you can swap

The model is swappable

The through-line, read all at once, is almost embarrassingly simple. The model is the part everyone talks about. The harness, the budget, the rebuilds, the outcome (the parts nobody downloads) are the parts that decide whether a person at the far end of the wire has a better morning.

The intelligence arrives finished. The delivery is yours to build. That's the whole doctrine; the four pieces are just the evidence.