Proof
Every tool claims accuracy. This one shows its homework.
Including the races it got wrong. Check it against your own results before you pay anything.
Honesty enforced by code, not by good intentions
The app won't start if a recommendation reads as a prescription
Six checks run at boot, and any one of them throws rather than start. The research library is read for causal or prescriptive phrasing. Every sentence the pattern engine can produce is rendered at three sample sizes and checked. A fuel entry without a source URL fails. So does a database table nobody has classified for export or deletion.
A race result can't carry a rank, a score or a personal best
Seventeen field names are rejected outright, because those concepts don't exist in the data model. Competitors ship exactly those as the headline feature.
Inventing a fact is counted, published, and capped
The race-page reader is scored against ten real race pages, hand-labelled by consensus. Declining to answer counts as correct; a confident value where the page said nothing counts as a fabrication. The gate allows at most two across the whole corpus, and the count goes in the run record either way.
The note on the gate reads: fix the change, do not raise the thresholds.
It retracts its own citations in public
Several research figures in the engine have been re-checked and withdrawn in the source after they didn't hold up. That's a position you can defend under scrutiny in a way that "our algorithm" is not.
Built for every runner, not just the ones who got studied
Most fuelling guidance comes from a narrow slice of the running population, then gets handed to everyone as a prescription. Your body isn't that sample, and your answer isn't the average of theirs.
So published guidance is a starting point, your own logged runs are the evidence, and where Absorb applies a research prior it says which one and how far that study actually reaches. Fade over distance is adjusted for sex, and declining to answer is pooled rather than treated as an answer.
The optional cycle space records the phases you name, on the dates you give, in your own words. Absorb never predicts a phase and never adjusts your fuelling or pacing by one, because the conventional four-phase model is wrong for a large share of female endurance athletes.
Prescriptions assume you're the average. Absorb assumes you're not.
Same gel. Different day. Different gut.
Four studies. Four things that were not the carbohydrate.
-
The hour
The same three hour run at 9pm moved through the gut slower than at 9am.
Crossover trial in 16 runners. More gut symptoms, and a much larger cortisol rise.
PubMed 33065594 -
Three days before
The runners worst hit had eaten more polyols in the days beforehand.
Field study at a 56km trail ultramarathon. Total FODMAP intake showed nothing.
PubMed 38637153 -
The night before
One high-FODMAP day raised gut symptoms during the next day's run.
Randomised double-blind crossover, 18 endurance runners, two hours at 35 degrees.
PubMed 31652404 -
Your head
How anxious you were on the start line tracked with how your gut went.
Prospective field study at a 56km ultramarathon, strongest in the final 24 hours.
PubMed 34897557
Running has a prescription problem. Take 60 to 90 grams an hour, as though the gel were the only variable in the room.
Absorb is the other way round. Your runs are the evidence, published guidance is a starting point, and the answer is allowed to be yours rather than the average.
Research figures quoted on this site are population averages, not a recommendation for you. More in the terms.
Absorb never predicts a phase and never adjusts your fuelling by one. It records what you tell it, in your words, and shows you what your own runs say.
Join the list
Be there when it opens
Absorb isn't open yet. Leave your email and we'll tell you on the day it is.
You're on the list
For people aged 16 or over.