My alarm won't let me snooze if I'd be late

Description and use

Snooze Oracle lets users safely snooze an alarm by predicting travel time variability, reducing lateness to 5.1 % versus a 10‑minute buffer.

Problem

Users cannot safely snooze an alarm when travel time is variable, risking lateness.

Outcome

The TabPFN‑based alarm was late 5.1% of the time across 12 commute routes, 2.9 minutes earlier on average than a 10‑minute fixed buffer, and outperformed lookup and LightGBM alarms in most scenarios.

Workflow

• 1. Collect past Uber trip data for the desired route and hour.
• 2. Train TabPFN‑3.5 on July 2026 trips (or use pretrained model).
• 3. For each potential wake‑up time, simulate 2,000 mornings (15‑min prep + ride).
• 4. Compute probability of arriving on time; select latest time with ≥90% probability.
• 5. Set alarm to ring 30 minutes before that time.
• 6. Allow user to snooze in 10‑minute increments until the safe threshold is reached.
• 7. When another snooze would drop probability below 90%, disable snooze and display "No safe snooze".

Evidence

Snooze Oracle uses past Uber trips to work out when I need to get up for a meeting.

Source

It finds the latest time I can get up for which the model estimates at least a 90% chance of being on time.

Source

The alarm rings 30 minutes before that, and I can snooze up to 10 minutes at a time until I reach it.

Source

The page is static. The curves were computed with TabPFN on my machine and saved into the page, so nothing on it calls a model.

Source

Across all 12 commute routes, the alarm was late 5.1% of the time and the 10‑minute buffer 8.3%. But you sleep a little less.

Source