An AI weight loss coach that recalibrates your calories with your real weight

What a week with the adaptive plan sounds like

Kodi

(Monday morning.)

“Morning 🐼 Week reviewed: with your 5 weigh-ins and 13 days of meals, your real expenditure comes out about 120 kcal above what we had. I suggest raising the target from 1,850 to 1,950 kcal while keeping the 130 g of protein: same pace, less hunger. Shall I apply it?”

I only weighed myself once this week, is that a problem?

“Nothing serious: with a single weigh-in I cannot separate your trend from the noise, so I am leaving the plan alone this week. Weigh yourself three or four mornings in a row and I will fine-tune it on Monday.”

What makes an AI weight loss coach different

Losing fat is, at its core, sustaining a moderate deficit for long enough without losing muscle or motivation. The hard part is not the theory, it is the follow-through: knowing your real expenditure, adjusting when it changes and not quitting in the bad week. An AI coach helps with that when it works on your data rather than a template.

• It recalibrates instead of guessing: the starting formula is a starting point; from the second week on, your scale is in charge.
• It caps the changes: no automatic adjustment moves your target more than a bounded step at once, and you always confirm it.
• It keeps protein: when calories come down, protein stays in its grams-per-kilo band instead of shrinking in proportion.
• It talks to you twice a day, in the morning and at close, and answers when you write. No pings at random hours.

How your plan starts

The first plan comes from the standard, well-documented stuff: your basal expenditure by Mifflin-St Jeor, an activity factor and a pace you choose. Gentle, moderate or intensive become an increasing percentage below your estimated expenditure, with a cap on body weight lost per week so that “intensive” never means “reckless”. Protein is set in a grams-per-kilo band taken from ISSN recommendations, and the rest is split between carbs and fat according to your diet type.

That plan is a hypothesis. The app knows it, which is why it starts testing it from the first weigh-in.

How it recalibrates every week

Once a week, the engine looks at your recent weeks. It computes your weight trend (not your last weigh-in, which bounces with water and salt), with its margin of uncertainty, and your average intake. From that it gets an observation of your real expenditure: if you ate 1,900 kcal on average and lost 300 g a week, you burned about 2,230.

That observation does not replace the previous estimate: it is combined with it giving more weight to the more reliable data. Few weigh-ins or a lot of scatter, small step; dense, regular weigh-ins, convergence in a few weeks. The result is an estimated expenditure that homes in on yours without lurching.

• Goal reached: if your trend hits your target weight, the plan moves to maintenance before you keep dropping on momentum.
• Weight drifting: if your real weight moves far enough from the one the plan used, it is recomputed in full with the new weight, anchored to the expenditure already learned.
• Expenditure shifting: if the estimated expenditure drifts from the one behind your target, Kodi proposes the adjustment, capped.
• Missing data: with too few weigh-ins in the window or too many days without food logged, the engine does not adjust blind. It tells you what is missing and tries again next week.

What Kodi does with that

The engine computes; Kodi communicates. In the morning it opens your day with your target, your streak and, when due, the adjustment proposal explained with your data (“with your 5 weigh-ins and 13 days of meals…”). At night it closes the balance: calories, protein, what you achieved and what is up tomorrow. If a week goes badly, it does not punish you: the plan works on averages and absorbs the bad week.

Kodi remembers what you tell it (Thursday dinners out, broccoli never, the knee that cannot take impact) and keeps it in mind when proposing. And if you prefer zero conversation, data-only mode switches it off and leaves the engine with the interface of a metrics dashboard.

Logging without it turning into a job

An adaptive plan needs you to log most days. That is why logging has to take seconds: a photo of the plate, a voice note on the way to the office (“two slices of toast with avocado and a latte”) or the barcode on the yoghurt. A portion estimated from a photo is flagged as such and corrected in one tap if it does not look right. Every entry adds calories, macros and 27 micronutrients against EFSA reference values, so losing weight does not mean losing your iron along the way.

Who it is for, and who it is not for

• It is for healthy adults who want to lose fat with a moderate deficit and a plan that adjusts to their body rather than a table.
• It is for anyone who has stalled on a fixed number and wants to know whether the number is the problem.
• It is for anyone wearing a wearable who wants sleep and training to count in the whole picture.
• It is not for anyone who wants to lose weight fast at any cost: the intensive pace is already the prudent maximum, and the engine does not go below it.
• It is not for minors, pregnancy, breastfeeding or anyone who has or has had an eating disorder: in those cases the reference is a professional, not an app.

Frequently asked questions

Keep going