ONE INTELLIGENCE · ANALYSIS ARCHITECTURE
Across a three-hour ride that is nearly eleven thousand numbers. ONE Intelligence orders them so they can be compared six months from now and publishes them to your Athlete Portal, where you see your history, your load and ONE Readiness. What it deliberately does not do is decide for anyone: your coach does that.
Published monitoring · coach judgement · experimental engine kept separate
What it is, in one line
Five layers: capture, order, analyse, turn into a tool and, at the end, have someone read it alongside the case. For the athlete, the visible part is the Athlete Portal; the part you do not see is the conversation in which the coach decides what changes. Remove that last layer and what remains is a system producing very precise numbers about questions nobody asked.
What you can see today
The Athlete Portal gathers your power reports in a private access: your best power by season, your load, energy by zone and where you sit against riders of your sex and age. You see the latest published update, with its date, and review it with your coach.
You sign in with your user or with a code sent to your email. If the code does not arrive or you changed email, your coach updates the access.
Open the Athlete PortalPortal status checked on 5 October 2026. The portal and its reports are in Spanish.
Operational
Compares your recent load, over about six days, with your usual load, over about twenty-one, computed from the power of your rides. It also says how stable that state is: how many of 27 reasonable ways of computing it agree. If you log session RPE and the log is sufficient, it comes in as a second opinion.
Limit: It does not measure sleep, heart-rate variability or overall recovery, and it is not a probability of injury or performance.
Operational
The essentials in two minutes: your level, your power profile and how you trained over the last ninety days, with your position against riders of your sex and age.
Limit: It summarises your records; it does not replace the conversation with your coach.
Operational
All the data: best power by duration and season, load, energy by zone, history and the explanation of each method, with data coverage in view.
Limit: Zones and critical power are descriptive estimates, not thresholds measured in a laboratory.
Experimental
Shows what would happen to your profile if you change weight, watts or training hours.
Limit: It is a hypothetical scenario to discuss, not a prediction of how you will perform.
For the coach, the portal also opens the team board: athletes, peaks, state and exports in one view, reviewable from anywhere. Reports and the board require a session and are not indexed by search engines.
How the data arrive
Five steps, and only the last one is taken by a person. That is why the report says when it was updated: the portal is not a live feed.
You record the ride on your head unit or trainer, and it syncs with the platform where your coach sees your training.
ONE software reads the original device file. It separates stops from moving time and discards impossible spikes, duplicates and rides without power.
It computes your best power by duration, load, energy by zone and ONE Readiness, with stated rules and the data coverage of each calculation.
Your report is prepared and published to your private access with the date of that update. You open it from your phone or computer.
Your coach reads it alongside what was planned, your goals, your hours and what you tell them, and decides what stays and what changes.
You see the latest published update, not the ride you finished five minutes ago.
A ride you just finished appears after the next refresh of your data. The report is built from rides recorded with a power meter: if you train with heart rate only, it will have little to show, and monitoring happens with your coach.
Method and evidence
A power you recorded, a calculated zone and a model estimate are not worth the same, even when they show the same number of decimals. The portal keeps them apart, and so does this page.
Your best power by duration, and the time and energy of each ride. It is the firmest part of the report, with two limits: it is only as good as the sensor calibration, and a duration you never pushed fully shows up below your real capacity.
Critical power fitted on your best 3 to 20 minute efforts, the zones derived from it and ONE Readiness. They are useful for organising training and are not laboratory measurements. The first threshold cannot be identified from power alone, and the report says so.
A power–duration model that estimates capacity with probability distributions conditional on its assumptions, and abstains where a duration has no support in your data. It is research, kept separate from the portal readings: calibration of its uncertainty on real riders is still pending, and it is not used to prescribe.
Critical power is not your FTP. They are two different estimates from two different protocols, often close without being interchangeable. The report says which one it uses, and a training zone is not converted from one to the other without saying so.
No test, when the data allow it. Efforts from your normal training are used. If a duration has no recent informative effort, the curve at that duration is not used to prescribe, and an assessment may be needed before setting an intensity.
What is not claimed: validated capacity at eight hours, calibrated uncertainty, an edge over other programmes or a model unique in the world. If any of those claims comes to have evidence, it will be published with its method.See the engine in R&D.
The architecture
Each layer hands something to the next. If one fails, the ones below keep producing numbers and none of them mean anything: unnormalised data cannot be compared, and analysis without a reading changes nothing on Monday.
Layer 1
Where the data come from.
Operational
GURU Ultimate 3D simulator, three-dimensional motion capture, dynamometry, anthropometry under protocol, bike geometry, power files, heart rate, variability, athlete history and training context.
Layer 2
How they are ordered so they can be compared.
Operational
Normalisation, structured storage and a longitudinal profile per athlete, so a measurement taken today can be subtracted from one taken six months ago. Without this layer every session is an island and no comparison means anything.
Layer 3
What is done with them.
Descriptive in operation · Bayesian experimental
Biomechanical rules, descriptive load and power calculations, and statistical analysis, written as in-house software that can be versioned and reviewed. Alongside it runs a Bayesian power–duration engine that is still experimental, and a pattern-classification line under validation. Language models are used as internal support for ordering information: none of them writes a report or decides an adjustment.
Layer 4
What it turns into on the outside.
Operational · with experimental parts
The Athlete Portal, with private power reports, ONE Readiness and express and advanced readings; the team board for the coach; the 11 public calculators on this site, the Alto de Patios Challenge and analytical support for BIKEFIT and BIKEFINDER. The portal simulator is labelled experimental, inside the portal and outside it.
Layer 5
Who decides.
Always human
The last layer is not software. A data point does not become a decision by passing through a model: somebody has to read it alongside the case, the injury history and what the rider says happens at kilometre sixty.
Responsible use
Saying what a tool does is easy. What actually informs whoever is about to use it is where it stops, and that part is rarely published.
Portal reports are prepared by software, with the same format and the same rules for everyone, so a number from today can be compared with one from a year ago. What is not automated is the reading. What that report means for your week and which session changes is decided by your coach, with your goals, your hours and what you tell them.
The portal simulator and the Bayesian engine carry the experimental label wherever they appear, and the lines under construction have a page of their own with the state of each one. It is the difference between learning the limit by reading and learning it halfway through the season.
A measurement is only worth taking if it changes the intervention. If measuring it does not alter what happens next, it is surplus, and an assessment starts from the question in the case rather than the equipment catalogue.
Your report only opens with your session, and only you and your coach see it. The methodology can be published; reports and identities cannot. Published cases are anonymised, and the public calculators compute in the browser without sending anything to a server.
ONE Intelligence does not replace professional judgement. It organises, computes and amplifies the information the specialist interprets.
Where it shows up
Frequently asked questions
No. It orders the information the fit is made with and keeps it comparable over time. What to move, by how much and why is decided by whoever assessed the rider, and the response is checked afterwards on the bike. A system that decided on its own would have to explain why lumbar pain means excess reach in one case and limited hip mobility in another, and that does not fit in a model yet.
Your power reports in a private access: ONE Readiness, the express report and the advanced report, with your best power by season, your load, energy by zone and your position against riders of your sex and age. The simulator is labelled experimental. Coaches also get the team board. You sign in at /portal/ with your user or a code sent to your email; the portal and its reports are in Spanish. Status checked on 5 October 2026.
No. It shows the latest published update, with its date. Each ride goes through import, analysis and report preparation before publication, so the ride you finished an hour ago appears after the next refresh of your data.
Not in an overall sense. It compares your recent load with your usual load, computed from the power of your rides, and says how stable that state is across 27 reasonable ways of computing it. If you log perceived exertion and the log is sufficient, it comes in as a second opinion. It does not measure sleep or heart-rate variability, and it is not a probability of injury or performance: it is a reference to discuss with your coach.
The software prepares the reports with the same rules for everyone, so they can be compared over time. The coach interprets the results alongside your goal, your hours and what you report, and decides the plan adjustments. No report changes a session on its own.
Portal reports use records and descriptive calculations with stated rules. Separately, in the laboratory, there is a Bayesian power–duration engine that is still experimental: it estimates capacity with distributions conditional on its assumptions, its uncertainty is not yet calibrated on real riders, and it is not used to prescribe. Language models are used as internal support for ordering information; none of them writes a report or replaces a reading.
Not always. Efforts from your normal training are used. If a duration has no recent informative effort, the curve at that duration is not used to set intensities and an assessment may be needed. Critical power is not treated as if it were your FTP either: they are different estimates.
Your measurements are used for your process. Published cases are anonymised and carry nothing that identifies you, and the public calculators on this site compute in your own browser: they send nothing. If anything changes on that front, it will be said here beforehand rather than afterwards.
The ONE Alpha Data Stack is the training data layer: load, recovery, monitoring and plan reporting. ONE Intelligence is the whole framework, of which that layer forms one part alongside lab capture, analysis and the public tools.
Tell us what you measure today and which decision is pending. Diagnosis One says what is missing and what is surplus; if you already train with ONE, your report is in the Athlete Portal.
Choose how to start
