Every morning at 7am, a machine on Val Town wakes up, trains a logistic regression on my history, and decides how many burpees I should do today. Not a fixed number. A negotiated one — it picks the goal with the highest expected value, and then I can push back.

If you think that's a lot of infrastructure for tracking whether I did some jumping jacks — you're right. But I'd argue it's not enough.

Think about the other side of the ledger. Whole product teams, backed by billions in funding, spend their careers figuring out exactly how to nudge my behavior. The timing of the notification. The color of the button. The streak I'd lose if I stop. They're very good at this.

I'm not going to out-spend them. But I can at least compete — by making the things I actually care about as salient, easy, and hard to ignore as the things they want me to do.

I used to spread this energy across a dozen habits, which meant I was always behind on something and nothing stuck. Now I've narrowed to one keystone metric: VO2 max. It's one of the strongest predictors of healthspan. High-intensity intervals are the highest-leverage way to move it. So: burpees.

One wrinkle. Burpees are a yang exercise — explosive, taxing, hard on the body if you rush in cold. I need yin to balance and to prevent injury. So I do sun salutations first. Five of them. And here's the key: those salutations count toward my burpee goal. The system subsumes its own prerequisite. Yoga is just slower burpees.

Most goal systems silently absorb failure. You miss a day, nothing happens, the number resets. Over time this teaches you that missing is fine.

Beeminder does the opposite. Every goal hit posts a datapoint on a chart that has a required slope — a "yellow brick road." Fall below the road and you lose real money.

The important thing is what happens the next day. The chart doesn't care about your excuse. There's still a road. The question is just: are you on it?

A goal you can negotiate away isn't a goal. An arbitrary number gets ignored. The goal has to feel like it knows something.

Every morning, a logistic regression model sweeps goals 1–50, computes P(success) × goal for each, and picks the maximum expected value. It trains fresh on all my history with 31 features — streak direction, rolling success rates, day-of-week cyclicality, effort ratios, how close I got on days I missed.

The model is sometimes annoyingly right about bad days. When it gives me a lower number than I expected, I've learned to take it seriously. Hit Done or Skip a few times below to see how the goal responds:

https://dcm31--22eabcfe1a4311f1953c42dde27851f2.web.val.run

This is the newest part and I think the most interesting.

When the model proposes a goal, I can submit my own probability estimate — "I think I'm 75% likely to hit this." That gets stored as a feature, the model re-runs with it, and the goal can actually shift. If I'm more confident than the model — slept well, have time today — the goal nudges up. If I'm doubtful, it pulls back.

Drag the slider below to see how your confidence changes the recommendation:

https://dcm31--9cc8f2ac1b2911f18fb042dde27851f2.web.val.run

The goal is no longer just what the model thinks is optimal. It's a negotiation — and my gut read has skin in the game.

Both predictions — the model's and mine — each create a separate question on Fatebook: "If my goal is N, will I complete N burpees today?" The next morning, each question auto-resolves YES or NO from the database.

Over time this builds a calibration record — one for the model, one for me. Those are different questions. The model track measures whether its probabilities are realistic. My track measures whether I know myself. Am I overconfident on Saturdays? Do I underestimate a good week's momentum?

Not enough data yet to draw conclusions — but the infrastructure is there.

This is the underrated property. The goal means nothing if you don't see it constantly, and logging has to be zero friction.

Every morning at 7am a Telegram message tells me the goal — it sits in my messages for the whole day.

The Atom Matrix on my desk shows it in red LEDs — a Cistercian numeral on a 5×5 grid. The bottom row is a 5-dot streak history. When I'm done, it turns green.

The Apple Watch shows it as a complication I see every time I check the time, plus an iOS lock screen widget. Four surfaces total.

The primary logging path: finish an HIIT workout on Apple Watch → workout completion fires an Apple Shortcut → one tap marks it done. Zero extra steps. If I did fewer than the goal, I can log the actual count — the model learns from effort ratios, not just binary hits and misses.

Probably not. But it's the direction to keep building.

The corporations fighting for my attention have compounding advantages: more data, more engineers, more psychological research, more feedback loops. The only way to compete is to be more intentional about my feedback loops — by making the one thing that actually matters for my health just as salient and frictionless as whatever they want me to click next.

One metric. One keystone habit. Every surface it can reach.