In the Air Force I had an AAR. Solo, I had nothing.

In the Air Force I had an AAR. Solo, I had nothing.

Issue 1 gave you something to argue with. This is the other half — something that remembers what you said and makes you answer for it.

Paul Cobb
July 27, 2026

Last issue I made the case for an AI sparring partner — something to argue with before you commit real money to a bad read. A few of you replied with the same question, worded different ways: fine, it argues with me in the moment. But who makes me follow through the other six days?

Fair. Arguing is half of it. The other half is memory — something that remembers what you swore you'd do last Monday and sits you down to answer for it. A sparring partner keeps you honest in the moment. A coach keeps you honest over time. As a solo operator you need both, and the trick is the balance.

Here's where this comes from for me. In a former life I was a military officer, and every week ended the same way: an after-action review. What was supposed to happen, what happened, why the gap, what we change next time. It wasn't a mood — it was a structure, run whether the week went well or badly. Then I went out on my own, and that structure just... evaporated. No one runs an AAR on a one-man shop. I missed it more than I expected to. So I rebuilt it, with AI holding the clipboard.

There are three levels to this, and you can start at the first one today.

Level 1 — the weekly prompt.
This is the whole thing in one paste. It's a personal AAR on a weekly clock — review, reckon, re-commit. Steal it:

"You're my weekly accountability coach and sparring partner, and we keep the same thread every week so you remember what I told you.
Start by pulling up the commitments I made last week. Walk them one at a time: done, partial, or slipped? If I slipped, ask me why — one question, enough to surface the real blocker, then move on. Don't let me off the hook, and don't lecture.
Then name my two or three real wins and the one or two blockers that cost me. Keep it short.
Now set this week: push me to three to five commitments that are specific and checkable. If I say something vague like 'work on marketing,' make me turn it into 'publish two posts for the Fairdale rehab by Thursday.' Reject anything you couldn't verify next week.
Close with a quick read: am I energized, steady, or stressed? Is momentum high, medium, or stalled? Then list this week's commitments back to me so you can hold me to them."

Notice what the structure forces. You can't report "worked on marketing" — it makes you commit to something it can check next week. That's the AAR discipline: vague goals can't be reviewed, so they never get done. Specific and checkable, or it doesn't count.

Level 2 — make it a project, keep one thread.
A prompt you run fresh each week is a form. A prompt that runs in one continuous thread is a coach. That's the whole difference, and most people miss it.

Set this up as a saved project or a single ongoing conversation, and never start a new one. When the thread remembers that you've now slipped the same lead-follow-up commitment three weeks running, it stops being polite about it — because the pattern is right there in front of both of you. Continuity is what turns a checklist into something with teeth. My own version runs on a tool in my stack that stores each week's debrief so the thread never resets and the commitments carry forward on their own — but a single long chat gets you most of the way for free.

Level 3 — package it as a skill.
Once the cadence proves itself, stop re-pasting it. Turn it into a skill — a reusable package that carries the whole rubric, your rules, and your voice, so it runs the same way every week without you rebuilding the setup. The major assistants support skills now, and creating one is quick: describe your weekly rubric and ask the AI to build the skill for you, then add it in your settings so it loads every session. Same repeatable process, same artifacts, none of the drift.

I did exactly that with my own weekly review — packaged it into a skill so I'm not rebuilding the prompt every Monday — and I've made a version you can run yourself. It walks the same review-reckon-recommit cadence, holds your commitments across weeks, and pushes when you slip. If you'd rather start with something proven than build from a blank page, take mine for a spin: forvex.app/try.

This is the same move as everything I build: capture a process you're already running by hand, package it once, and let it run like a trained team member instead of a thing you re-explain every time.

Here's the report-back that matters, from my own weeks running it: the value isn't that it hands me a plan. It's that it won't let me slide on the thing I'd quietly dropped. It's not there to be encouraging. It's there to remember.

Where it breaks.
Three ways, all of them mine to prevent. Kill the memory — start a fresh thread each week — and you're back to journaling with extra steps. Let it go soft and agreeable, and it becomes a yes-man; you have to explicitly tell it to push, or the sparring-partner half dies and you're left with a cheerful secretary. And let the commitments stay vague, and the whole review turns into theater — motion without a scoreboard. Structure only works if the structure is real. The coach runs the cadence; you still have to bring the honesty.

The play: this week, write down the three commitments you made to yourself last Monday — even if you made them only in your head. Open one thread, paste the Level 1 prompt, and answer for all three out loud. Then set next week's, specific enough that the thread can check them. Keep that thread. Next Monday, it'll be waiting.

Know an operator who runs solo? Forward it — that's how this list grows. And hit reply: what's the one commitment you keep quietly letting slide? Naming it is half the fix. I read every one.

Next issue: I'm pointing the same approach at the messiest part of my business — rehab. How I use AI to see a finished room, and scope the work to get there, before I move a single sink.

— Written by Paul Cobb — real estate operator (Louisville / Kentuckiana), forVEX builder. I run the plays I write about. More depth: forvex.app/journal