Beyond the Static Plan: Why I'm Building N+One
If you’ve followed my recent posts on Vibe Coding and the LLM benchmark wars, you know I’m obsessed with the transition from AI as a "text-box" to AI as an "agent." To truly test the limits of agentic workflows, I decided to apply these patterns to a domain where data is messy, feedback loops are physical, and precision is everything: Endurance Coaching.
For years, the delta between amateur and professional cycling hasn't just been about physiological talent or time - it's been about guidance.
Pros have coaches who analyze their HRV before they even wake up, adjusting intervals based on a poor night's sleep or a stressful travel day. The rest of us? We've been stuck with static PDF plans or "adaptive" algorithms that are really just fancy IF/THEN statements.
In my recent writing, I've discussed the shift from Vibe Coding to Agentic Engineering. This same paradigm shift is happening in sports science. We are moving away from "software as a tool" and toward "software as an agent."
That is why I am building N+One.
The Problem: The "Planning-Reality" Gap
Most training platforms suffer from a rigid architecture. They expect you to be a robot. If you miss a Tuesday VO2 max session because of a late-running deployment or a family commitment, the plan breaks.
You're left with two bad options:
- The Guilt Trip: Try to "make up" the intensity on Wednesday, leading to a fatigue spiral.
- The Guesswork: Manually drag-and-drop sessions, hoping you aren't messing up the periodization.
Human coaches solve this through conversation and context. They don't just look at your TSS; they ask how you feel. But at € 200 per month, that level of nuance is a luxury.
The Agentic Intelligence Layer
N+One is built on what I call the Adaptive Intelligence Layer. We aren't just another dashboard for your Garmin data; we are building a Conversational AI Coach that understands the "Why" behind your training.
By integrating your health stack - WHOOP, Garmin, Strava - the agent gains a holistic view of your readiness. But the real magic happens in the chat interface.
Instead of navigating complex menus to "re-optimize" a week, you simply tell the coach: "I'm feeling a bit blocked after yesterday's intervals and I only have 45 minutes today." The agent doesn't just give you a generic suggestion. It cross-references your current 338W Critical Power, your sleep debt, and your goal event to serve you the exact "Next Best Action."
Why "N+One"?
The name comes from the classic cycling rule for the "correct" number of bikes to own: N + 1, where N is the number you currently have.
In the context of coaching, I see it differently. It's about the Next session + One better decision. It's the marginal gain that comes from never doing a "junk" mile because your plan didn't know you were exhausted, and never missing a breakthrough session because you were afraid to push.
The Vision: Democratizing Elite Performance
We are moving into an era where the barrier between "data" and "action" is dissolving. At N+One, we are applying the same principles of Agentic Engineering - autonomy, reasoning, and real-time adaptation - to the world of endurance sports.
Our goal isn't to replace human coaches, but to make world-class coaching principles accessible to everyone. Whether you're chasing a 4W/kg power-to-weight ratio or just trying to finish your first century ride, you shouldn't have to be a data scientist to train effectively.
The future of training isn't a calendar. It's a conversation.
If you're interested in the intersection of AI and human performance, check out N+One and join our waitlist for early access.
