# Sonoma: driving and car review Public demo · Porsche 991.1 GT3 · 27 September 2026 Published excerpt of prior local analysis. No live AI run or ChatGPT-plan connection is used by this demo. Original recordings remain private. ## Comparison basis Imported from the completed local reports. Driving priorities use 26 pace laps, compared with their own session's quicker quartile, then aggregated equally by session. Mechanical comparisons use matched windows within S2–S4; no pooled solo/passenger degradation claim. ## Data quality S1 GPS/CAN timing mismatch; S4 CAN ends during L27; S5 has no CAN. Within-session comparisons only; observed temperatures are not safety limits. - Passenger load, tire state, fuel, traffic and track conditions remain confounders. Session normalization is not physical load correction. - Brake pressure is quantized to integer bar. No pad temperatures, pedal travel, wheel/ABS or clutch/shaft channels are available. - Firmware Fahrenheit byte packing cannot represent values above 255°F correctly; no observed peak is interpreted as an OEM limit. - Public demo: source IDs are remapped. Original evidence remains private. These are previously generated findings, not a live ChatGPT-plan analysis. - S5L32 pit-entry lap is omitted from the main lap chart. - S1 is excluded from timing-sensitive CAN/GPS mechanical comparisons. - Partial S4L27 only contributes to events completed before its CAN dropout. ## Repeatability priority: Esses T8-9 Esses T8-9 has a 1.16% median within-session gap to the quicker quartile, and ranks among the top three in 4 of five sessions. 26 pace laps, 3–8 per session. The same leading three sections survive 3–5 s selection windows, leave-one-session-out checks and ±20 m section-boundary shifts. These are descriptive execution gaps, not promised time gains. Next step: Compare distance-aligned traces within the same session and test one repeatable entry or linked-corner change. Keep traffic and load context attached. Confidence: medium. ## Repeatability priority: T1-T2 T1-T2 has a 1.04% median within-session gap to the quicker quartile, and ranks among the top three in 4 of five sessions. 26 pace laps, 3–8 per session. The same leading three sections survive 3–5 s selection windows, leave-one-session-out checks and ±20 m section-boundary shifts. These are descriptive execution gaps, not promised time gains. Next step: Compare distance-aligned traces within the same session and test one repeatable entry or linked-corner change. Keep traffic and load context attached. Confidence: medium. ## Repeatability priority: T10 T10 has a 0.93% median within-session gap to the quicker quartile, and ranks among the top three in 4 of five sessions. 26 pace laps, 3–8 per session. The same leading three sections survive 3–5 s selection windows, leave-one-session-out checks and ±20 m section-boundary shifts. These are descriptive execution gaps, not promised time gains. Next step: Compare distance-aligned traces within the same session and test one repeatable entry or linked-corner change. Keep traffic and load context attached. Confidence: medium. ## S2 braking deserves a matched recheck S2 pressure per deceleration rises +1.9% at T7 and +4.6% at T11 over 100→60 mph. Other sessions do not repeat that pattern; brake fade is not established. Six events per S2 zone; first-two versus last-two medians. Changing to 90→60 mph raises T7's estimate to +14.6%, showing speed-band sensitivity. See the source report for all selections and confounders. Next step: Repeat matched-speed stops and record pedal feel. Validate pressure and add identified wheel/ABS channels if available. Confidence: low. ## No progressive early-upshift pattern The loaded upshift RPM envelope stays similar across the available sessions. This does not establish clutch condition or exclude every protection mode. 150 loaded 2→3 / 3→4 upshifts in the prior report. Reported gear state is not verified clutch-engagement timing; S1 is excluded from RPM/GPS ratio conclusions. Next step: Retain Automatic PDK Sport context and prioritize uninterrupted CAN plus verified gear/shaft semantics on the next recording. Confidence: medium. ## Fix coverage before stronger conclusions S1 has an approximate 0.58 s GPS/CAN timing mismatch. S4 drops CAN during L27 and S5 contains GPS without CAN. Existing alignment diagnostics and firmware review; both phones used the same RaceBox Mini and car profile. Phone and video activity changed together, so video causation is not established. Next step: Validate synchronization and continuous recording before the next day; retain recording-device and timing-offset settings. Confidence: high. ## Provenance The companion analysis.json carries source references. sources.json maps demo identifiers to private evidence filenames and original checksums. The public report is an excerpt, not a substitute for that evidence. Session inventories come from the later deterministic preparation run; channel availability does not verify units or clock alignment.