The Liftie SutrasSend It or End It
A true story from the chairlift

The Liftie Sutras

Send It or End It

The rules that decide when a ski lift stops were never written down. We dug them up.

A huge crowd waiting in a long line under a lift at Palisades Tahoe
A lift line at Palisades Tahoe, from r/palisadestahoe.
1

A morning you do not control

It is a powder day. You are in a line that is not moving. The Tram and the Funitel, the two big lifts the rest of the mountain hangs from, have stopped.

No sign says why. No one says for how long. You stand there, cold, while the day drains away.

2

Rules nobody wrote down

Every system has rules like this. In video games, players dig the hidden rules out of the code. That is easy, because the code exists.

The rules that run real life, like when a lift opens, who gets in, when it stops, are not in any code. The only way to learn them is to stand in the line.

3

So we dug

The operators leave traces every morning. They post it: “The Funitel is on wind hold.” We treated those posts like old tablets, and the weather like a carbon date.

7,044posts from the operators across 8 winters
3readers: our parser and two AI models, blind to each other, read the posts
1winter hidden at a time, to see if the rules could predict it
What came out of the ground
I

When the gusts reach sixty, the wind holds begin.

II

When the gateways close, the high lifts follow.

Everything on the upper mountain hangs from the Tram and the Funitel.

III

The low lifts keep spinning while the high ones sleep.

KT-22, Exhibition, Red Dog and Resort Chair usually run.

IV

When snow and avalanche danger build, the dawn comes late.

SEND IT
OR
END IT

Sutra I was tested on winters it never saw; the others are counts over all eight winters. People guess lifts stop around 25 mph. The data says about 60 mph for the day's strongest gust, and about 50 mph if all you know is the wind by 9am.

Eight winters, eight layers

The operators never wrote the rules down, but every morning they left traces. Each layer is a winter. The number on a layer is how well the book called that winter's wind-hold days when it was hidden, from 0 to 1. Click a layer to read what was found in it.

Hide a winter. Predict it.

For each winter, the rules were rebuilt from the other seven and then asked to call the hidden one. The first chart is the main finding. The lift-level numbers below it come from the repository's own results (scripts/baselines.py), counted on expected lift-days: lifts on the morning schedule, plus lifts that normally run on a whole-mountain closure day. Winter 2020-21 ran under COVID operating rules and is left out of those counts.

Nobody graded their own homework

Our parser read every post. Two independent AI readers, Subquadratic (all 7,044 posts) and Meta Muse (7,040; four calls failed), read them blind. Posts where both readers agreed with each other and the parser differed were pulled for review. On 2026-11-07 five models (Claude Opus, Sonnet and Haiku, Meta Muse and Subquadratic) also graded the parser's events by hand on two sets of 120 posts nobody had seen: the five-model majority called 88% and 85% of posts fully right. After more fixes, a frozen sample of 320 posts nobody had seen was graded once by three Claude models: 87.5% were read fully right (95% interval 83 to 91), and 32 of the 35 closure posts. The models agree with each other more than they prove the truth, so a human check of the closure and disputed posts comes next.