Hire a “What’s Missing?” Agent.
You can't see everything, and neither can the agent doing the work. Give one agent the standing job of asking what's missing.
- What I saidHey, what's missing?
- Who asks it nowA reader that never builds
- What it provesEvery gap, with the search
Hey, what's missing?
Right after my advisor caught itself about to build a second copy of my design (Lesson 5), I asked for something new:
Okay, I need something that does this all the time for all my work. […] We need this. We need everything you've mentioned, and we need somebody who says, hey, what's missing? Because I certainly can't see it all. I don't even pretend that I can. […]
Three minutes later it was running, and my advisor described its job:
[…] It can read everything and change nothing. It reads a piece of work with your original ask beside it and looks for what nobody covered: who could cheat it, what happens when it's wrong, what changes underneath it, whether we know it's right and not just consistent, and what bigger thing it could become. […]
Seven minutes after that, its first report showed my advisor had searched too narrowly. That's Lesson 6.
The builder sees what it built.
Every agent that makes something checks it against what it meant to make. That's useful, and it can't find what nobody thought of.
Course 1's Lesson 7, Don't Grade Your Own Homework already sends a fresh agent to check that finished work is true, because “The agent that did the work already believes it.” The gap-finder has a different job. It doesn't check that what was built is right. It looks for what was never built at all.
The questions it asks.
Who could misuse it?
Not who it's for. Who it could hurt, or who could game it.
What happens when it's wrong?
Everything is wrong sometimes. What breaks, and who finds out?
What changes underneath it?
The tools, the data, the rules it depends on.
Is it right, or only consistent?
The same answer twice isn't proof. Lesson 2 has the test.
Who gets left out?
The person it never pictured.
What could it become?
The bigger thing, past the target. Lesson 14.
It proves every gap the way Lesson 6 says: the search, and the same search finding something it knows is there. It reads and never builds, so it has no stake in the answer. And its reports lead with the worst gap (Lesson 8). In my studio it reads the big pieces of work and reports to my advisor, not to me.
Someone has to look for what isn't there.
Give one agent, not the builder, the standing job of asking what's missing. It sets your original ask beside the work, reads and never builds, and proves every gap.
Run it before things ship, not after they break.
Hire one today.
Use a fresh chat, not the one that did the work.
- Hand it the ask and the work.
Paste your original request and the finished thing into a new chat, then this:
Here's what I asked for, word for word, and here's what was made. Don't fix anything. Tell me what's missing, worst first: who could misuse it, what happens when it's wrong, what could change underneath it, whether it's right or only consistent, who it leaves out, and what it could become. For each gap, show me where you looked, and show me the same search finding something you know is there. - Make it standing.
Once it has earned its keep, put it in the routine: anything big gets its read before it goes out.
The builder sees what it built. Someone else has to look for what isn't there.
Where this comes from in course 1. Course 1's Lesson 7 (don't grade your own homework) and Lesson 10 (bring me the choices I don't know about). New here: a standing reader whose only job is what's missing, and its six questions.
Past the Target
Hit the target. Then look past it: what is this really for, who else could it serve, and what will you wish you'd asked for?
← Previous: Lesson 12, Pace the Burn
Every exchange on this page really happened and is quoted exactly, slips included; […] marks where I trimmed. Times are Eastern. Names of clients, people, companies and products are left out. My agents are AI models; the names are left out on purpose.