Optimisation · DEFLATE
Faster or smaller, proved correct.
Miners replace the LZ77 parser of miniz_oxide's DEFLATE encoder and prove in Lean that its output decodes back to the input. Every accepted parser is timed and measured on a fixed corpus.
- Of validator weight
- 20%
- On the frontier
- 7
- Miners scored
- 3
- Submissions
- 13
- In the queue
- 0
Scored
- Miner
- Reference parser
- Not admitted
- Pareto frontier
- Scoring limit
Every submission
Submissions
Sort by
- 5CAkr5…FDPt3n#12
- Gate
- Accepted
- Admission
- Inconclusive
- Time
- 0.63×
- Size
- 36.06%
28 Sept 2026, 13:29 UTC · Speed gain not established
- 5HjJX7…PsExHb#2
- Gate
- Accepted
- Admission
- Out of bounds
- Time
- 12.1×
- Size
- 34.66%
24 Sept 2026, 14:07 UTC · Slower than the time limit
- 5CvyPx…iKHYRZ#1
- Gate
- Accepted
- Admission
- Inconclusive
- Time
- 7.80×
- Size
- 35.00%
23 Sept 2026, 15:34 UTC · Speed gain not established
How it is scored
The rules
- 01The competition sets 20% of the validator's weights. The rest goes to the treasury.
- 02Only parsers on the Pareto frontier of time and size are paid, and only while they stay on it. How much each one gets depends on how much of the frontier it holds: a point that opens a new trade-off earns more than one squeezed between close neighbours.
- 03If one hotkey has several parsers on the frontier, only the oldest of them is paid.
- 04A parser counts only if it is at most 10× the incumbent's time and compresses to at most 40% on average.
- 05A parser that is faster without compressing better has to prove the gain: the lower end of a 95% bootstrap interval must be above zero.
- 06One submission earns at most 3,600 α over its lifetime. After that its share goes to the treasury.
- 07A miner's source stays private while it is on the frontier. It is published once something beats it.