FSBK Nogaro recap — what the data taught us

Third round of FSBK 2026, third paddock visited. Nogaro, May 28-31. Four days of riding, track temperatures climbing fast, and most importantly: data flowing back without a hitch.

Here’s what we saw, what worked, and what we’ll fix before Pau Arnos.

What worked — folder cross-session analysis

This was the first real-world outing for folder-level BudAI analysis, shipped just before Lédenon. At Nogaro, we saw it run on actual race weekends.

Typical scenario: a rider creates a “FSBK Nogaro 2026” folder, assigns their eight sessions of the weekend (Day 1 Practice, Day 2 Qualifying 1 and 2, Day 3 Warm-up + Race 1, Day 4 Warm-up + Race 2), and launches the cross-session analysis. One click, one AI credit, and BudAI compares everything.

What the AI surfaced at Nogaro:

• “Your best Race 1 time (1’34.8) beats your qualifying best by 0.3s — you’re stronger in rhythm than in solo effort.”
• “Your braking point at Turn 1 (the long right) moves back 4 meters between Day 1 and Day 3. That’s the main source of your gain.”
• “Your maximum right lean drops from 51° to 47° in Race 2. Either a rear tire issue, or hesitancy after the lap 3 incident.”

Rider feedback: it replaces 20 minutes of spreadsheet work with 30 seconds of waiting. And the AI catches things you don’t see going lap-by-lap.

What worked — coaches actually using rider profiles

Second real-world validation: coaches started writing in their students’ rider profiles.

Quick recap: since April, every rider has a structured profile — strengths, weaknesses, recurring patterns, goals, coach notes. The coach can write in it (with the rider’s consent), and BudAI reads those notes when analyzing a session. The idea: stop the AI from repeating obvious things to the rider, and have it dig into what the coach is actually working on with them.

At Nogaro, we saw coaches writing things like:

• “Tendency to brake straight in fast linked sections — focus on Turn 5-6”
• “Nogaro goal: hold 1’35 in race rhythm over 5 consecutive laps”
• “Pirelli SC1 rear tire, watch degradation after lap 6”

And BudAI actually picked up those elements in its session analysis. The rider no longer has to explain context to the AI every time — their coach already did.

That’s exactly the habit we hoped to see form.

What we’ll adjust — BudAI’s vocabulary

Now the honest critique. Three pieces of feedback came in about how BudAI expresses itself, and we’re taking them all.

1. Too much telemetry jargon, not enough rider language.

BudAI was saying things like: “Your longitudinal deceleration profile peaks at -1.2g, 84m before the apex.”

What the rider wants to read: “You’re braking hard, but 84m before the apex — that’s 15m too early compared to your best lap.”

We’re reworking the prompt to translate numbers into rider intuition. The G-forces stay available for those who want the numbers, but they no longer carry the sentence.

2. Not enough reference to corner names.

At Nogaro, saying “Turn 8” means nothing to anyone. Saying “the Castagner chicane” or “the final hairpin,” everyone sees it. We already have the corner-name database (see our article on named corners) — now BudAI needs to use it when the circuit is identified.

That integration is in flight.

3. Generic advice on fast sections.

On flat-out linked sections (like the long entry sweeper at Nogaro, or the esses at Pau Arnos), BudAI tends to say “keep the throttle on” — which helps no one. We’ll refine so it distinguishes sections where the useful advice is about line (visual reference, pivot point) rather than throttle.

4. Potential gains too optimistic — and that’s on us.

The clearest feedback, and the most important. BudAI was saying things like: “You could gain 4.8s by aligning your best sectors.”

Except that calculation — sum of best sectors, or the “theoretical ideal lap” — is a mathematical ceiling, not a realistic target. No rider, not even a pro, aligns their best inputs at every corner in the same lap. It’s physically and mentally impossible: a perfect braking point in Turn 3 often comes at the cost of a compromise in Turn 4. Promising 4.8s means promising fairy dust.

The result: riders wondering where those mythical 5 seconds went, and eventually losing trust in the analysis.

What we’ll change: BudAI will stop talking about a “theoretical ideal lap.” Instead, we’ll compute a realistic gain — for example, the median of your 3 best passes per sector (which filters out one-off perfect inputs you can’t reproduce), or a target based on what you’ve already held over 2 consecutive laps. The AI will say something like: “Based on what you’ve already done, aiming for 1’34.5 over 3 consecutive laps is reachable — that’s 0.6s of gain.”

Less impressive as a headline, much more useful on track. And more honest.

What’s next — Pau Arnos in 2 weeks

Pau Arnos, June 18-21. Fourth round of FSBK, and probably the trickiest on the calendar — a short, hilly, technical circuit where every tenth is earned through reading the track.

The adjustments above will ship before then. And we keep talking to riders in the paddock — that’s where the best ideas come from.

If you’re riding Pau, come find us. We’ll analyze your weekend for free, and take your feedback live.

And if you weren’t at Nogaro but you have Nogaro files sleeping somewhere — upload them. It’s free up to 5 sessions per month, and it’s exactly the kind of material BudAI improves on.