I'd been trying and failing to lose weight for six months, so I did the only thing I know how to do when a system stops behaving: I sat down and debugged it. The fix turned out to be embarrassingly small. The result was not.
“I decided to use my Computer Science degree to debug my weight loss.”— the most honest sentence I've written
Bear with me, because this sounds like a joke and it mostly isn't. The flaw was real, the math was real, and once I saw it I couldn't unsee it. So let me show you the bug.
The plateau that didn't make sense
The early part of this story is boring in the best way. Basic advice worked great. I started working out about once a week, ate a bit healthier, slept enough — and lost my first twenty pounds with barely any effort. Encouraged, I ramped up: lifting three times a week, tracking calories and macros, sleeping well, playing sports. Over the next year I lost another twenty pounds of fat and put on about ten of muscle. By any honest accounting I had the habits dialed in.
And then, despite keeping all of it up, the scale just stopped. I tried calorie counting, daily weigh-ins, meal planning — nothing moved the needle. The maddening part was that I was nailing every habit except the one that mattered: I could not consistently eat fewer calories than I burned. Every other system in my life had a feedback loop I trusted. This one felt like superstition.
So I broke it down the way I'd break down any other problem. Not "try harder." Debug. And the first thing you do when you debug is stop trusting the thing you've been staring at.
The signal is buried under the noise
Here's the core problem, and it's a measurement problem before it's a willpower problem. What you actually want to lose is fat. But your bodyweight is fat plus water plus whatever's currently in your stomach plus a dozen other things that swing around independently of how disciplined you were yesterday.
The fat you might lose in a single good day is tiny — call it a third of a pound. The day-to-day noise from water and food is several times that. So even if your "true" weight didn't change at all, the scale can read two pounds higher or lower the next morning, and both numbers are lies.
This is a signal-to-noise problem, and the ratio is brutal. You're trying to detect a 0.3 lb signal inside ±2 lb of noise. That's a signal-to-noise ratio of about 0.15 — like trying to hear someone whisper in a thunderstorm. No wonder it felt like superstition. I was trying to read tea leaves.
So I did what you do: I tried every obvious fix first. They all failed, and why they failed is the interesting part.
Every naive fix I tried (and why each one broke)
- Weekly weigh-ins. "I won't see fat loss in a day, but surely a week will show it." Except the noise doesn't vanish — it just moves. Flat for two weeks? You'll read a pound down one Sunday and a pound up the next, and conclude you're winning, then losing, when nothing actually changed.
- Daily weigh-ins. "Maybe more data points cancel out the noise." Kind of — but your real weight can read two pounds either way on any given morning, the numbers jump around, and your brain is a pattern-matching machine that insists on seeing trends that aren't there. The fake progress and fake setbacks wreck your motivation. This is the one that nearly broke me — it pushed me to count every single calorie as a last resort, which lasted about two days.
- Reduce the variation. Same time every morning, same (lack of) clothes, after the bathroom. It helped a little. It did not help enough — the body's own daily swing was still far bigger than any fat I could lose in a day.
- Graph it. Logging into Fitbit and watching a trend line was the closest to working. But the feedback was delayed — a real trend only emerged over many days, so I was always reacting to choices I'd made a week ago. The link between today's dinner and the number on the screen was severed. If you've read Atomic Habits, you know a feedback loop with that much lag barely qualifies as a feedback loop.
At that point I was genuinely exhausted by the whole thing. Every other habit I'd built — lifting, sleep, sport — felt like it was under my control. Weight was the one place I felt like a passenger. The problem, I finally realized, was never the diet. It was the feedback. How could I get an honest, immediate read on whether today's eating actually put me in a deficit?
And then it clicked. It's almost stupidly simple.
The fix: stop tracking your weight, track the delta
Draw a straight line on the graph representing the rate you want to lose weight. Your target on any given day is just your starting weight minus the days elapsed times your daily loss rate. Then — this is the whole trick — stop looking at your weight, and start looking at the delta: target minus actual.
I had ChatGPT spit out a list of daily target weights, one per day, and pasted it into my notes. Each morning I'd compare the scale to that day's target. The magic happened immediately: I stopped caring about the weight number's patterns entirely. My eyes went to the target. Only later did I realize what I'd actually done — I'd quietly switched from tracking a noisy number to tracking the gap from a moving reference line.
“I wasn't tracking my weight — I was tracking my delta.”— from the weight-loss feedback-loop notes
Here's why that one move beats the noise. When you track raw weight, a bad day's mistake just dissolves into the day-to-day churn — invisible, deniable. But the target line keeps dropping 0.3 lb every day whether or not you cooperated, so when you mess up, the deviation doesn't disappear. It accumulates in your delta.
Miss your target today? Your delta reads +0.3. Annoying, ignorable. Keep slipping for a week and your delta is +2.1 — now you're sitting above the daily noise floor, and it's impossible to pretend it's water. The signal-to-noise ratio fixes itself, because you've stopped chasing a tiny daily trend and started keeping one number hovering around zero. Same scale. Same body. Radically clearer information.
And it gives you the thing weekly weigh-ins can never give you: immediate feedback. You find out about yesterday's choices today. Tight loop, clear instruction, trained intuition. If the delta ever drifts past ±2–3 pounds and starts feeling demoralizing, you just redraw the target line back to zero — you've still made real progress, and you've kept the loop psychologically alive, which is the part that actually matters.
The part I didn't expect: it changed my head, not just my data
I went in looking for a measurement fix. What I got was a psychological one, and the shift was instant and a little dramatic.
Instead of feeling helpless in front of patterns that weren't real, I had a target and a verdict every single morning. Behind by half a pound? Tighten the deficit today. Ahead by 0.3? Stay the course, you're fine. The feedback was immediate, the next action was obvious, and progress stopped feeling like a coin flip. After a few weeks I started loosening — adding variety back to my meals, getting sloppier with calorie estimates — and it kept working, because the tight daily loop had quietly trained my intuition about food and portion sizes. I'd internalized the thing I used to have to white-knuckle.
The most important change was the one underneath all of it: I went from feeling like weight loss was random and out of my hands to feeling like I had a reliable system that would deliver if I just did my small part each day. That's the whole game. A daily win, however tiny, that reinforces the habit — until the math quietly guarantees you hit the goal.
I'll be honest about a caveat, because this is exactly the place these systems can turn on you. A loop this tight is incredibly motivating, and motivation pointed the wrong way is how a tool for getting healthier becomes a tool for punishing yourself. The rule I actually live by, the one that outranks every number on the chart:
“abs are a bonus, not a permission slip to exist.”— my own guardrail
The delta is supposed to make you feel in control, not in debt. If it ever flips from "clear instruction" to "moral verdict," that's the bug, and that one's worth debugging too.
The method was never about weight
Here's the thing I find funniest about this whole story: the fix had nothing to do with weight loss. I didn't out-discipline the plateau. I found the real bottleneck — which was signal, not willpower — and built a feedback loop tight enough to act on. That's the same move I make whether I'm learning a language or shipping an app.
“The domains change. The method doesn't.”— from the notes
Decompose the thing. Find the bottleneck that's actually load-bearing, not the one that feels hard. Build a feedback loop. Iterate without ego until it works. I've pointed that engine at a lot of problems, and "my own body" turned out to be one of the most stubborn and most satisfying. (If you want the fuller version of the body-as-a-system story — versioned workout plans, a hypertrophy tier-list, the sports that are really how I make friends — that's over on the Athlete.)
How to actually do it (it's a five-minute setup)
You don't need an app for this. You need two numbers and a column.
- Take your current weight as the starting point.
- Pick a sane daily loss rate — small, sustainable, nothing heroic.
- Generate your target for each day: start − (days × rate). Excel does this in one drag; ChatGPT will hand you the whole list if you ask, which is what I did.
- Each morning, weigh in and compare to that day's target. I'd overwrite the target with my real number in bold, so the row showed the gap at a glance.
- Keep the delta near zero. Behind, push a little; ahead, coast. That's the entire job.
That really is it. The whole thing fits in a notes app.
And then I built it into an app
Of course I did. The notes-app version works but it's tedious — generating the list, overwriting rows, eyeballing the gap — and tedious is exactly the kind of friction that kills a habit. So I built a small app called Weave that does the bookkeeping for you: enter your starting weight, pick a pace, and weigh in once a day. It figures out that day's target and your delta, and just shows you the one number to keep near zero.
I built it carefully, and I'll defend one decision forever: none of the actually-helpful stuff sits behind a paywall, because charging someone to see whether they hit their own target would be evil. I also added a friends layer, so you don't have to do this alone — it keeps your actual weight private and just shares whether you hit your target, so you can keep each other honest without anyone broadcasting a number they're sensitive about.
My mom used it and lost seventeen pounds after years of trying. I lost sixteen. Two people, one straight line on a graph, and a delta we each kept near zero. The bug report that started as "the scale won't move" closed as "shipped." I'll take it.