Halmos sign in sign up

Queue

Everything the checkers do goes through one queue, and this is it. Nothing here is a simulation — these are the jobs on the box right now.

0
running now
0
waiting
0 ms
to clear the queue
0
parked for compute

Running

Idle. A job submitted now starts immediately.

Waiting — 0 priority, 0 free

Nothing waiting.

The fast lane, and what it is not

One Lean job runs at a time — Mathlib is memory-hungry and this box is shared — so when several people submit at once, there is a line.

Any account holding prepaid credit is served first. That is the whole mechanism; there is no separate product to buy. A free run is never blocked, throttled or degraded by it — only overtaken — and it still runs, in order, in the free lane.

The reason it works this way: the people funding the machine are paying for the checks everyone else gets for nothing. Skipping the line is the only thing that buys, and it buys nothing else.

Add credit to skip the line Your compute

Just finished

ToolResultTookFor
python ran 72 ms a comment
lean refuted 8.4 s A wrong proof, kept up on purpose: \sqrt{4} is irrational
lean proved 8.5 s Every finite division ring is commutative
python ran 39 ms A one-line check that 2^{67}-1 is composite
python ran 64 ms \sqrt{2} is irrational, and the kernel agrees
lean proved 10.1 s \sqrt{2} is irrational, and the kernel agrees

How long each checker usually takes here

Measured from this box's last 20 runs of each tool, not estimated.

ToolTypical
lean9.0 s
sage15.0 s
pari5.0 s
smt22.5 s
python58 ms
oeis7.5 s
plot15.0 s