Built in the garage

Oct 5, 2026

Zach and I shared his 2022 Mustang Mach 1 at COTA with Chin Track Days. My best beat his by 1.177 seconds, and 0.85 of it came from one corner. What my slow morning in his car actually cost, who wins the theoretical laps, and why the naive way of building them flips the answer.

Oct 1, 2026

Before my first COTA weekend in three years I wanted the real track, seen from above, with a real HPDE lap on it. It didn't exist, so I built it, with a short how-to for every corner.

Sep 13, 2026

At Road Atlanta my Z4 M40i beat Rich's race Mini by 7.318 seconds. At Atlanta Motorsports Park the same two cars finished 0.944 seconds apart, with a wet M440i and a dry E86 along for the ride. What a tight track does to a power car, what an old 1 Hz logger can honestly tell you, where the rain actually took its 14 seconds, and whether my shop is right that the E86 needs a diff.

Aug 24, 2026

The fastest Road Atlanta lap in my dataset came from a driver with ten years of HPDE and a heavily modified M4. Every pro team pairs a driver that fast with a race engineer — someone buried in the data between sessions. HPDE drivers can't rent one. So I measured the road instead: 15 million LiDAR points over the circuit, and the discovery that exactly one corner at Road Atlanta is off camber.

Aug 19, 2026

Two Z4s, one weekend, one racetrack, fourteen years of BMW development between them. The 2022 M40i's 1:40.53 against the 2008 3.0si's 1:48.99, aligned meter by meter — and only a third of the 8.46 seconds turns out to be the engine. The rest is a build plan and a driver.

Aug 15, 2026

A race-prepped E36 M3 and my stock-ish Z4 M40i turned nearly identical Barber laps 31 minutes apart on the same Saturday morning — 1:42.163 to 1:42.70. Aligned meter by meter, the half second turns out to be two completely different cars cancelling each other out.

Jul 27, 2026

I have 40 laps of Circuit of the Americas in a BMW M5 from 2023. Formula 1 publishes its timing data in a coordinate system that doesn't exist on any map. I fit one onto the other, derived all 20 corners from the geometry, and compared them apex by apex — then found the two mistakes that made the first comparison nonsense.

Jul 23, 2026

Same track, same day: Rich's 2,350-lb race Mini against my 382-hp Z4 M40i's personal best. The Mini matched or beat the BMW's cornering speed at six of seven corners — and lost by 7.318 seconds. Where lap time actually comes from, with receipts.

Jul 12, 2026

A 2,350-lb race-prepped Mini against my bone-stock Z4 coupe at Road Atlanta, called corner by corner like an F1 broadcast — built entirely from real GPS data. Six lead changes. And one twist: the race never actually happened.

Jul 12, 2026

A snapped bleeder valve, a seven-day brake thrash, and two track-prepped Z4s — fourteen years apart — in the same paddock. The 2022 Roadster found 2.73 seconds; the 2008 Coupe survived its debut with brakes that were parts in a box five days earlier.

Jun 28, 2026

A heavy, speed-capped M440i beat my lighter Z4's VIR best by 3.47 seconds and came within 0.22s of my all-time M5 lap — with no coach in my car. What the AI half of the loop can do alone, the one fix that beat the math, and the near-incident at 130 mph where the data took the other guy's side.

Jun 5, 2026

A lap identical to my two-year-old personal best — to the millisecond — followed an hour later by the real thing. What a tied PB teaches you that a new one can't, and the true story of the 2022 crash I'd been driving around ever since.

May 31, 2026

An AI read four years of my Barber data and pointed at three corners. A coach in the passenger seat taught me how to actually take them. Neither got me to a personal best alone — and that's the point. First of a series on my AI-coached track weekends.

May 15, 2026

What happens when a data engineer decides the night-before-a-track-day spreadsheet isn't good enough.