Two Great Coaches (Rich & Claude) Found Me 2.57 Seconds

Barber Motorsports Park, with the corner numbering Just Track It uses. All corner references below follow this map.

This is the first of many posts about my AI-coached track weekends. The AI — every lap I've ever driven, memorized — is the constant across the series. At Barber I also had a friend coaching from the passenger seat, and this is the weekend I learned why I want both.

The morning number

Sunday morning at Barber, 8 AM, still cool enough that the M440i's tires needed a lap to wake up. Seven laps, best of the bunch a 1:48.56. Fine. Unremarkable. The kind of lap you post and move on from.

Except before Rich ever got in the car, I'd already run the morning data through the AI, and it had flagged three corners where I was leaving time on the table: the T7-T7a right-hander, the Carousel (T13-14), and the Museum hairpin (T5). Not vague "you could be faster here" notes — specific sectors, ranked.

The morning number also came with a piece of trivia I'd completely forgotten. Sitting in a paddock chair, I asked the AI what my best lap at Barber in the M340i had been. That was my 2019–2022 track car — it ended its life at Road Atlanta in a July 2022 crash I get into in the next post. The answer: one session at Barber, ever, January 2022, best lap 1:48.46. My morning best in the M440i was 0.10 seconds off a lap I'd set in a different car, four years earlier, and I found that out between sessions by just asking.

The paddock lookup: one M340i session at Barber, ever — January 2022, 1:48.46.

Then Rich got in the car.

Two ways of finding the same three corners

Here's the twist that makes the coaching matter more, not less: I wasn't at Barber as a student. I was there finishing my MSF-2 instructor certification with JTI, and the weekend's real work was running teaching role-plays with two other instructors, Jeremy and Rich. Our last dry-run happened to be in my car — which is how Rich, a friend as much as a fellow instructor, ended up behind my wheel. He drove it first for the demo, then I drove with him coaching from the passenger seat, then I went solo to see what stuck.

Rich's demo lap wasn't even his fastest of the day (1:49.55) — it wasn't about lap time, it was about showing me lines and brake points I wasn't using.

The demo session on the timer — building from 1:59s down to the 1:49.55.

Then I drove with him coaching live: 1:47.40, a full second under my morning baseline. Afterward, the AI and I looked at where that second had come from, and it was the same three corners it had already flagged before Rich ever sat in the car: T7-T7a, the Carousel, Museum.

Two independent reads on the same laps had landed on the same three problems — one from a guy who's coached hundreds of drivers around this track, one from a g-trace analysis of my own morning session, neither knowing what the other had said. That convergence told me the diagnosis was right. But a diagnosis isn't a lap time. Knowing T7 was costing me didn't make me any faster through it — Rich, sitting next to me and telling me exactly where to brake and where to get back to throttle, did. The AI found where the time was hiding. Rich taught me how to go get it. I needed both, and this is the post where that clicked.

Ten years, one Miata

Since we're on the subject of how I ended up in the instructor seat: my first-ever track day was August 2016 — AMP, with JTI, and an instructor named Nabil who took me out for a few laps in his Miata, the way instructors do. I'd never been in a real track car. It blew me away.

Ten years later I was at Barber as an instructor myself, my 58th JTI event, running the MSF-2 role-plays that were the actual reason for the weekend. And the car Jeremy and I were flying around Barber in for those role-plays? That exact same Miata. Nabil sold it to Jeremy a few years back, and a decade on it still impresses me. That's the JTI family, and getting to help the next batch of drivers start their own obsession is most of why I keep showing up.

Lunch with the other race engineer

Between the coached session and the afternoon, I uploaded the morning data and asked for the full cross-car comparison: this lap against my best in the M340i, the Z4 sDrive30i, and the Z4 M40i, all at this track.

Four cars, one track. And a blunt headline: I was over-slowing the M440i — my minimum corner speed (39.7 mph) was the lowest of any car I'd ever driven at Barber, slower even than the 255-hp Z4 30i.

The analysis ended with a prediction:

"Brake later, lift later, carry the corner… A lap in the 1:45s is in there if you commit."

That was the assignment for the afternoon.

The final solo lap, and the number that didn't make sense at first

After lunch I went back out alone. Five laps. Best one: 1:45.99. Backed up by a 1:47.34 the next lap out, so not a fluke, not a single hero lap with nothing behind it.

The 1:45.99, onboard.

That's 2.57 seconds faster than the morning, in one day, in a car I already knew reasonably well. Here's the breakdown of where it came from:

• T7-T7a right complex: -0.43s
• Carousel (T13-14): -0.38s
• Museum hairpin (T5): -0.33s
• T15 final esses: -0.23s
• T1→T2 transition: -0.22s
• The rest spread thin across the remainder of the lap

That's 1.14 seconds out of the Big 3 corners alone — the same three the AI had called before the day started.

But here's the part that actually confused me when I first looked at the telemetry. My peak braking g went up — from 0.75g in the morning to 1.06g on the PB lap. Rich's demo lap, the one I was supposedly learning from, only showed 0.39g of peak brake. So the lesson I thought I'd absorbed — "brake less, like Rich did" — was backwards. I was braking harder than ever and going faster.

Brake later, not less

What was actually happening, once we lined the data up: I wasn't braking less. I was braking later, harder, and for less time, then getting back to throttle sooner. Rich's demo lap looked gentle because he was carrying speed into the braking zone that I wasn't — by the time he touched the brake at all, he'd already scrubbed less speed than I needed to. My job wasn't to copy his brake pressure. It was to copy his entry speed and brake point, then apply however much force the car and tires could actually take once I got there.

Combined with more cornering commitment — peak lateral grip went from 1.14g in the morning to 1.22g on the PB lap, closing in on the 1.19g I'd recorded in my own (much lighter) Z4 M40i at this same track — the formula for the day was: trust the tires more in the corner, brake later and harder into it, release earlier out of it. More total time on throttle. Less time coasting or trail-braking indecisively.

Top speed ticked up too, 110.9 to 112.5 mph, and minimum speed through the Museum hairpin jumped from 39.7 to 44.3 — nearly 5 mph carried through the slowest corner on track. None of these are dramatic on their own. Stacked together, sector by sector, they're 2.57 seconds.

Where that leaves the M440i

My career-best at Barber, set in the Z4 M40i, is 1:43.91. The M440i's new PB of 1:45.99 leaves a 2.08-second gap. My working estimate — and I said this before the weekend, based on the roughly 600-pound weight difference between the cars — was that the M440i's realistic ceiling at Barber sat somewhere in the 1:45.5 to 1:46.5 range. 1:45.99 lands right in that band. (For the record: my M440i xDrive Gran Coupe runs about 4,167 lb curb weight against the Z4 M40i's roughly 3,575 lb — a ~592-lb gap, close enough to the 600-lb estimate that the ceiling math holds.) Maybe half a second of pure driver improvement left in this car before the rest of the gap is just physics — a heavier car asking more of its tires and brakes to do the same job.

The other database in the paddock

Paddock, Sunday morning. Tire rotation in progress — and yes, that's a SQL plate.

The lap analysis isn't the only place the data does real work. Every tire I own has an identity — a colored cap and a number, matched to a row in the database. That morning at Barber, the rotation plan came straight out of it: Black-2 on the left rear, Red-2 on the right rear, Blue-3 and Green-1 across the front, all from the Pirelli P Zero Corsa set, each with its own running heat-cycle count. When a tire gets swapped to the spare corner or retired, that's a database write, not a mental note.

The database is also honest about what I didn't tell it. The row for this day carries a note: "fronts Blue-3 + Green-1, L/R not recalled." I forgot to record which front went on which side, and rather than guess, the record says so. That's the same discipline that makes the lap analysis trustworthy — the data only ever claims what actually got measured.

What actually got me to 1:45.99

The AI is a side project of mine — it's what became trackdays.ai — and it's genuinely good at exactly one thing: it remembers every lap I've ever turned and tells me where the time is hiding. What it can't do is sit in the passenger seat. This weekend was both halves at once — millions of records pointing at the corners, and a coach in the car teaching me how to actually drive them.

The split is right there in the data. Rich's coaching took me from 1:48.56 to 1:47.40 — 1.16 seconds. Going solo and committing to what both of them had been telling me all day found another 1.41. Neither half happens without the other: without the AI I don't know which corners to spend Rich's time on, and without Rich I'm holding a diagnosis and no treatment. That's the whole argument. If you're chasing lap time, an afternoon with a good coach will do more for you than a fresh set of tires. Rich, I owe you a steak dinner.

I helped Claude write this post

In the interest of honesty: I dashed off a quick Facebook post about this weekend, then handed it to the same AI to help turn it into what you just read. It caught me on the first pass. I'd written that my 1:48.5 came on Saturday — the data disagreed. Saturday was a 1:52; the 1:48.56 was Sunday morning, a few hours before the personal best. My own memory, corrected by my own lap database, in the middle of writing a post about trusting my own lap database.

So yes — I helped Claude write this post. Mostly it helped me, and fact-checked me while it was at it. Which, now that I type it out, is the entire point.

Next up

Five days later I was at Road Atlanta with a different coach, a pre-built attack plan waiting for me before I ever turned a wheel, and a lap time that landed exactly — to the millisecond — on a personal best I'd set two years earlier. Same number. Completely different car underneath it. That one took some untangling.

And my next trip to Barber will be in something a lot more Miata-like: my '08 Z4 (E86), about 1,100 lbs lighter than the Gran Coupe. Not a track car yet. It'll become one.