My house exit.
I pull onto a 4-lane main road with constant traffic. Two lanes run south-to-north, two north-to-south, separated by a guardrail. For me, that exit is the standard exam for every FSD version.
The car would cross the first two lanes, position itself in the median, and wait there until it found a window.
The car tries to make the full crossing in one move, waiting for a window where both sides are clear at the same time. In real traffic, that moment barely exists.
And when it fails, it doesn't give up gracefully: it sits there blocking the exit. Same exit, same driver, same routine. The regression is brutal.
v···
This table repeats identically every review. Same criteria, version after version. For an interactive view with any pair of versions, open Compare →
Each criterion is scored 1–10 unless otherwise noted. Repeating the same table every release makes comparison direct and verifiable.
Same exit. Same table. Next review.
With every new version I repeat the same format — the house exit, the same criteria, the same table — so progress stays verifiable.