Pocket Philosophy

What it is

Pocket Philosophy is a daily game about questions with no right answer. Each day there is one thought experiment, the kind that gets argued over at dinner. Two AI models from different labs take a side each and argue it, then swap sides and argue it again, and a third model referees every turn before anything is published. You read the argument in highlights, about 2 minutes’ worth, with the full transcript a tap away.

Then you answer. Not who won, but where you stand on the 3 to 5 points the argument turns on. Only after that do you see where everyone else landed, how far your gut verdict sat from what your answers implied, and which point pulled you hardest. There is no winner, no leaderboard, no streak and no feed. No question needs an account, there are no ads, and nothing about other players is shown until you’ve answered.

What we built

The game and the machinery behind it. The engine runs the debates, a referee checks every turn, and a separate check makes sure both sides of every point are worded with equal strength. An admin lets a person write or approve every question and schedule it, and the player app has its accounts, email and results.

Behind it

Every guardrail in the design, from “no player data reaches a model” to “no verdict language anywhere in the app”, is an automated test. The debate engine runs on a database account that can’t read the players’ tables, and a test parses every player-facing file and fails on a verdict word. The referee is proven against 5 faults planted in a copy of a real debate, and 1 trap it must leave alone. Every model call is priced from its own token count, and the month’s spend is capped.

Where it stands

In development. The first 10 questions are written. The launch date will appear here when there is one.

More work

IWE, Fetchpawz and the 3D configurator, or all the work.