Build Two

Now make something you would actually use. A reading-level generator, an intake form for a makerspace, a tracker for the thing you currently keep in a spreadsheet.

What changes is not technical.

Your knowledge of the subject no longer starts full. In Build One you were the expert. Here you are solving a problem that touches other people, and what you know about their side of it is thinner than it feels.

The constraints arrive from outside. Somebody else's data, somebody else's rules, a policy you did not write.

And somebody who is not you has to be able to use it.

The gate is a card, not a prompt

Before you open anything, fill in the Anchor card. Seven fields, on paper.

The Anchor card is printable. It asks what professional judgment this sharpens, whose workaround it replaces, whether it makes you more capable or more dependent, and whose evidence you have actually heard.

The field people leave blank is whose evidence. It is also the one that predicts whether anybody uses what you build. A blank there is not a failure to fill in the card. It is the card doing its job, and it is telling you to go talk to somebody before you write a line.

No is a result

A Build Two that ends with a defensible no is a successful Build Two.

Worth saying plainly, because otherwise every card gets filled in a way that justifies building the thing you already wanted to build, and then the gate is decoration.

If the honest answer to "does this make me more capable over time" is that it makes you dependent, you have learned something more useful than a working tool. Write down why and stop.

If you build it

The STEER check is the pre-ship pass, run once the thing works and before anyone else touches it. Five dimensions, one sheet.

The one people skip is Evolving Delegation, because it is the only dimension that asks you to design for taking something away: trust that widens with use and can always be pulled back. A tool nobody can back out of is a tool that quietly makes the decision for them.

What this is really for

Anchor and STEER was written for people designing AI tools that teachers would use. Reading it as a framework is one thing. Being the person it was written about is another, and you cannot get there by reading.

When you run the facilitation tracks later, you will be asking learners to plan before they build and to say who their thing is for. You will be asking it having done it on something that mattered, with somebody else's constraints, and having at least once decided not to build something.

That is the difference the learners will hear.