SCALING HIGH QUALITY
HUMAN JUDGEMENT
Driving rewards to open source repos with a market of AIs as the engine and humans as the steering wheel.
THE THREE COMPONENTS OF DEEP FUNDING
The model(s) best aligning with jury spot checks distribute funds
across the dependency graph.
Dependency Graph
A data layer mapping all contributions to an ecosystem
Model Submissions
An open contest to crowdsource the relative value between contributions
Spot Checkers
A jury to express preferences on a randomly selected set of contributions
THE DEEP FUNDING CHALLENGE
Give weights to source repos such that their summation to the target repo = 1
End Goal: Submit 15,000 weights to the Ethereum dependency graph
Level 1: Fund the Farm
Seed Node vs Seed Node
Develop an allocator model assigning weights to 34 source repos
(seed nodes) with the target repo of Ethereum
Level 2: Farm to Forest
Seed Node vs Child Node
Develop an allocator model determining the portion of funds staying
with seed nodes vs passed on to dependencies
Level 3: Fund the Forest
Child Node vs Child Node
Develop an allocator model assigning weights to over 5000
dependencies (child nodes) with the target repo of 34 seed nodes
Submissions: Feb 10 - May 10, 2025
The task of submitters is to give edge weights such that
A + B + C + D = 1 { Level 1: fund the farm }
B vs { B1,B2,B3,B4,B5,B6 } { Level 2: farm to forest }
B1 + B2 + B3 + B4 + B5 + B6 = 1 { Level 3: fund the forest }
Jurors randomly conduct spot checks to answer
"Has A or B been more valuable to Ethereum's success?" Submissions get a higher score if they are more compatible with spot checks
REWARDS WORTH $220k
$170k
to open source repos based on weights given by winning models
$50k
to winning models and knowledge sharing in the contest such as datasets, open source models & other infra
%201.svg)
