RAMTT

RAMTT is an endurance platform for athletes and the coaches who train them. It plans training, works out what that training costs in energy, tells the athlete what to eat and drink to pay for it, and reads the file afterwards to see what actually happened. It is made in Denmark by a very small team.

It began as a nutrition plan for one rider. The plan had to answer a plain question: how much carbohydrate does this person need, hour by hour, for the ride in front of them. The tools at hand could log a ride once it was over, but none of them could reason about it beforehand. So the plan became a small program. The program needed to know how hard the ride would be, which meant it needed a model of the rider, and the model needed the rider's history. Each answer asked for the next piece.

Today it is a platform built on more than 300 engines, each one a piece of exercise physiology written as code. They read every second of a session, power, heart rate, speed, position and altitude, and turn it into things a coach can act on: what the session cost, what it built, and what the athlete has left for tomorrow.

The engines contain no interface and no database. They are plain functions that take numbers and return numbers, which is what lets us test them against published data and against our own athletes.

The part we care most about is fueling. Most platforms treat food as a note in the margin of the training plan. We treat it as half of the training. Every planned session in RAMTT has an energy demand in kilojoules and a carbohydrate target in grams per hour, set from the shape of the session and from what the athlete has shown they can absorb.

That last part matters. The gut is trainable, and an athlete who can take in ninety grams an hour did not start there. RAMTT tracks that progression the way other tools track threshold power, session by session, and raises the target only when the evidence is in. Across a week it follows glycogen: how much the long ride took out, how much the days after put back, and whether the athlete arrives at the hard session with enough on board.

Around fueling sits a model of what the athlete can do. Critical power describes the effort a rider can hold for a long time, and a second number describes the limited work available above it. Together they draw the curve of power against duration, and they let us follow that reserve through a race as it is spent and recovered.

We also measure how those numbers drift. A rider four hours into a hard day is not the rider who rolled out in the morning. Durability, the rate at which power fades and heart rate climbs for the same work, is measured in every long session and fed back into the plan.

Training load in RAMTT follows three kinds of strain instead of one: metabolic, neural and peripheral. A long steady ride and a set of sprints can carry the same score in a classic model and leave an athlete tired in completely different ways. Keeping them apart shows a coach which system has had enough and which can take more.

Each morning the athlete answers a short check-in. From it we follow how loaded and how recovered they feel over a rolling week, and we read it next to the objective load, because the two disagree more often than people expect, and the disagreement is usually the interesting part.

Heat, altitude, time away from training and age are handled as corrections. They adjust a target or raise a flag. They never pretend to be a measurement.

Every number in RAMTT carries a grade. Some of our models rest on decades of published work. Others are our own and have not been validated yet, and the product says so, in words, in the place where you read the number. We would rather show a careful estimate with its doubt attached than a confident score nobody can check. The reasoning is written down in Method, the sources are listed in Research, and the terms are defined in Vocabulary.

This is also why there is no single magic score at the centre of the product. There is a summary, because coaches need one, but it opens into the parts it was made from.

We hold the software to the same standard as the science. The app should answer before you have finished asking, and look like someone cared about every screen. A four hour ride with a reading for every second should open as quickly as a note.

RAMTT is built around the weekly loop between a coach and an athlete. Plan the week. Ride it. Look at what happened. Talk about it. Plan the next one. The calendar, the workout builder, the session library and the analysis that follows each ride all exist to make that loop shorter and the conversation inside it better.

RAMTT reads the files athletes already record, from the computer on the bars or the watch on the wrist, and asks nobody to buy new hardware. A coach sees every athlete on one screen: what was planned, what was done, and how far apart the two were.

How the product is made, down to the typeface and the greys, is laid out in the Colophon. Changes are listed in the Changelog as they ship.

We are not open yet. What is finished, what is half built and what we have decided against is listed on Not yet, and it changes often.

RAMTT is made in Denmark.

For the longer version, start with the first letter in Current.

Malte Therkildsen coaches a group of riders and runners, and has done so for long enough to know where a coach's week goes: into planning sessions, reading files, and writing back to the athlete about what the file says.

Most of that week used to happen in tools that were good at storing rides and poor at thinking about them.

RAMTT started on his own desk, as the fueling plan for one of those athletes. When the plan outgrew the spreadsheet it became code, and when the code outgrew the plan it became a product.

He still coaches while he builds it, so every screen is used on real people on a Monday morning before anyone else sees it.

His own library of sessions, several hundred of them, from long low rides to short efforts with spikes in them, is the seed of the workout library in the product.

He writes most of what is published under Current, usually at length, and the shorter notes in Fieldbook.

He cares about two things in roughly equal measure: that the physiology is right, and that the software feels as good as the best tools made for any other profession.

Malte Therkildsen lives and works in Denmark.