Run, measure, repeat

The roadmap

• September: Luc Léger beep test to set my maximal aerobic speed, the reference every training pace is derived from.
• Winter: 5 km under 17 minutes, meaning a 3'24" per kilometre pace held without cracking.
• February 2027: the 10 km, the main goal.

What the data changes

Without measurement you run on instinct — usually too fast on easy days and not hard enough on hard days. With heart rate, pace and recovery tracking, every session has a clear role. Easy runs stay easy, which is the hardest part to stick to.

My watch records every run, every night, every heartbeat. That data sits idle inside an app. I wired it into an AI over the MCP protocol, so it can read the data and help me make sense of my training. The gateway runs on my server.

What it enables in practice

• Analysing a session: actual pace against target pace, cardiac drift, consistency.
• Tracking training load and recovery over several weeks, to avoid doing too much — or too little.
• Building workouts and pushing them straight to the watch, ready to run.
• Cross-checking sleep, heart rate variability and performance to spot what actually works on me.

Why the protocol matters

MCP is an open standard describing how an AI can call external tools. Instead of pasting data into a conversation, the AI queries the source itself and gets structured answers. The difference is clear: you move from an assistant that comments on the numbers you hand it, to one that fetches them and cross-checks whatever it needs.