AI Did This To Me

Purpose

I was already out of bed. That wasn’t the hard part.

I was sitting in my car, dressed, ready to go train, and I could not make myself leave. I just felt depleted. And the thought that showed up was an old one: I guess I can’t do this.

I know that hole. I’ve fallen into it more times than I can count in my life. It always works the same way. If I can’t go all the way, I don’t want to do anything. All or nothing — and that morning, nothing was winning.

So I did something I’d never done before. I asked a chatbot what I should do.

My ego had a clear answer: go. The chatbot had a different one. It told me to stay home. Rest. Stretch. Take two days.

I listened. And what it gave me that morning wasn’t a workout. It was a little bit of grace.

This wasn’t day one. The program had been running my mornings for a while by then — which is how the machine knew enough to make the call. But I have to back up to explain how it got there.

At 20, I trained to look good naked. I’m 48 now. The math changed. I train so I can keep up with my kids. I train so there’s never a moment where my body is the reason I can’t do something with them, or for them. I’m an airborne paratrooper and I’ve competed in jiu-jitsu at a high level. Being capable is part of who I am. I don’t want limitations that come from neglect.

The why was solid. The how was falling apart.

Work ran 8 to 6. Then dinner, then the kids’ stuff until 8 or 9, then whatever time was left for my wife. Training fit nowhere. My options were a CrossFit class at an hour that made no sense for my family, or nothing. My body had changed the terms, too. At 25 I got out of bed ready to attack the world. At 48 I get out of bed and the first order of business is stretching and moving before anything feels right. I could still perform. It just cost a lot more to recover.

I’d been using AI the way most business people use it — discovery questions, QA test cases, reviewing emails. Somebody told me they’d used it to plan a vacation. I thought: maybe it can plan this.

Here’s what I’ve figured out since. Everybody’s got a Lamborghini now. These AI tools are that powerful, and they’re sitting in everyone’s pocket. But most people are driving theirs 15 miles an hour — a trivia question here, a funny picture there.

I don’t know exactly how fast I’m driving mine. I’m definitely over the speed limit. And the speed was never the point. The speed is just what lets you accomplish the things you actually want in life.

The difference isn’t the tool. It’s what you feed it.

That’s the standard I run now, with all of it. The AI advises. The data verifies. I decide.

Applicability

What you’ll need

• One goal, stated plainly
• Your real constraints — schedule, equipment, food rules (these are inputs, not excuses)
• A data source you trust — a wearable, a scale, a training log
• A doctor who owns your medical decisions — this method organizes programming; it does not prescribe

Short on time? Get the starter file — the setup script from this article, filled in with your answers: Get the starter file

Procedure

Step 1 — Hand the AI your goal and every excuse

I started in January with the Christmas weight still on me and a jiu-jitsu competition in March. I had to get down to my fighting weight, and I had a pattern I was tired of repeating: either I’d show up to a competition underprepared and get my butt beat, or I just wouldn’t sign up at all.

I didn’t start with a program. I started with a problem. I told it I had an energy problem, a time problem, a body that didn’t bounce back, and I asked the most honest question I had: How can you help me?

It walked me through my excuses, one at a time. That’s the part I tell people now: give AI one goal and all your excuses. It will find a path. You just have to stick with it — and keep some feedback metrics along the way so you know you’re gaining traction, and know when to course-correct.

It built me 20-minute workouts, every morning, CrossFit style, using the equipment normal people actually have around the house — dumbbells, a kettlebell, some bands. Not some garage gym fantasy. What I had.

Then it did the thing I didn’t expect. It solved the scheduling war. I’d been asking myself when I could fit CrossFit in around jiu-jitsu. The answer was to stack them in the morning: the 5:30 open mat with the guys who were training for the same competition — tougher rounds than the night classes ever gave me — then home for the 20-minute workout.

One morning it all clicked. Jiu-jitsu from 5:30 to 6:30. Home. Workout knocked out by 7:15. Showered. I woke my wife up at 7:30 and said, “It feels good when you knock out two workouts in a day and your wife’s still sleeping.”

That was the morning I knew. I could do both. Every day.

For the jiu-jitsu, it built me a weekly series of movements for my specific style. I was getting caught in D’Arce chokes, so it taught me the counters — going belly flat, sticking my arm through, getting to a better position, even the submissions that live inside the escape. I can’t tell you the concepts were always right. What I can tell you is it gave me the exact searches to run on YouTube, so I could watch the actual moves from actual grapplers and drill them. The tools worked together. That’s a piece of this people miss.

Step 2 — Feed it your real data, not general questions

I gave it my real preferences — gluten-free, Korean, Mediterranean. Then I started taking pictures of what was in my refrigerator and cabinets, and it built meals out of what I had. Ten minutes at a time. I felt like a world-class cook. I surprised my wife with meals the AI taught me, and they were genuinely good.

Somewhere in there, I had a thought that changed everything: the better the information I gave it, the better the guidance that came back. So I asked what else I had. I had wearable data — a Whoop, a Garmin. I asked what it could do with that.

The Whoop won. And once the data was flowing, the program stopped being just workouts. It built the recovery to match the training — the stretching and active work in the morning, the sauna. When my back acted up, it made me pictures showing exactly where to place the TENS unit and where to foam roll. I’d take still photos of my stretching, and it would analyze my form — I was breaking the curve in my back reaching for my legs, so it cued me on keeping my lumbar curve where it belongs.

Step 3 — Check in daily and let the data set the day’s intensity

It matched the food to the day. Heavy training day, one way. Rest day, another. It sounds simple until you’ve lived your whole life eating the same way on a two-workout day as on a couch day and wondering why you feel empty by Thursday.

A plan that only tells you how to work is half a plan at 48. This one treated recovery with the same seriousness.

Step 4 — Review weekly against the record

So the program had two jobs: the weight, and the jiu-jitsu itself. For the weight, we course-corrected like adults. When the scale stalled, it raised my protein and pulled the carbs down until the weight moved. When I got low, the carbs came back — and my energy on the heavy days came back with them.

Step 5 — Correct the AI when the data disagrees

Early on, AI built me a supplement stack a mile long. Later — same tool — it told me to throw most of it away.

That whiplash was my fault, and it’s the most useful lesson in this whole piece. I was asking general questions, so it answered from general information — which on the internet means marketing. “Men in their 40s need this.” Claims stacked on claims, none of it checked against anything.

Published research is different. It’s been through a process we’ve all agreed gets us as close to true as we know how to get, at that moment. I learned to point the tool at actual studies and academic sources and make it show its work. Pointed at marketing, it’s a salesman. Pointed at science, it’s an analyst. Same Lamborghini. Different fuel.

Verification

1. Your plan names your actual equipment and your actual training days — not a generic split.
2. You feed it specifics — your numbers, your constraints, your actual day — not general questions.
3. Each morning’s intensity call is based on last night’s sleep and recovery data.
4. The weekly review changes the plan based on the record — when the scale stalled, protein went up and carbs came down until the weight moved.
5. Anything medical is decided by your doctor, not the AI — it organizes what’s already there; it doesn’t prescribe.
6. When the AI cites research, you check the citation yourself.

Limitations

I’ll say this part plainly, because I think men should talk about it plainly.

I was on TRT before any of this started — 200 milligrams a week, two doses of 100. I’m on tadalafil, 5 milligrams, for bladder relief; there’s longevity research behind it that I find interesting. Those prescriptions were already in place, worked out with my general practitioner before an AI ever saw my workout schedule. They are the constant in this story, not the variable. Where I was on standard protocols is where I started. Where I got to is where I got to.

The supplements were the part the AI guided — creatine, 10 grams a day; CoQ10; D3 with K2; fish oil — and then I talked through all of it with my doctor, the way you’re supposed to. I’m not on a GLP-1. I’ve been impressed by what I’ve seen from people who are, and I’m not going to pretend that isn’t true either.

I’m not recommending anything here. I’m telling you my story. Talk to your own doctor before you touch any of it.

So why did I believe it that morning in the car, over my own ego?

Because it had the data. And no ego of its own. It knew the week I’d had — the training, the sleep, the load — and it wasn’t trying to prove anything about my toughness. My discipline, the intensity I’ve built my whole life on, dug that hole as often as it climbed out of it. The machine just looked at the numbers and handed me permission to be a human being for two days.

The last competition before this, I got beat — and it hurt, because I knew I could do better. I hadn’t signed up for March yet. “I just won’t sign up” was right there, dressed up like a reasonable decision. That’s a crossroads. It doesn’t announce itself.

I don’t do this to look like anything anymore. I do it so that when my kids want to go, I can go. So there’s no list in my life titled things I can’t do anymore because I let myself go.

Everybody’s driving the same tools. The speed limit is self-imposed. Feed it your real life — your excuses, your equipment, your data. See what it hands back.

Then stick with the program.

Recap

• Give the AI one goal and every excuse — your energy, your time, your schedule, your equipment — and let it find the path.
• Feed it your real data — sleep, recovery, strain, food photos — not general questions.
• Let the morning’s data set the day’s intensity; review the record weekly and course-correct.
• The AI advises, the data verifies, you decide — and your doctor owns anything medical.

the starter setup. A short form — what tools you use, what you’d want help with — and you get the setup file that walks you through building your own version of this, step by step. Mine started with “How can you help me?” Yours can too.

This use case’s starter filehttps://docs.google.com/forms/d/e/1FAIpQLSdZalPMMy25y-6wsH61sBbyMO41qxo9xxOxdIM4hdPNj5mShA/viewform