DELPHI TRAINING FUND · TESTNET REHEARSAL
Every mint helps pay for Delphi's own AI
A share of every mint, royalty and network fee goes into an on-chain fund. Genesis holders vote on what to train, one card one vote. Trainers are paid mostly after evaluators reproduce their results, and every model the network serves is listed here with its weights on Arweave. Nothing leaves the fund any other way.
Pledged: mints
30%
royalties 50% · usage 10%
Deposited
0.004
ETH, all sources
Available
0.003
0 reserved for jobs
Models
1
trained and verified
Is the pledge kept?
Mainnet mints so far: 2 cards, 0.001 ETH to the Safe.
Owed (30% of mints)
0.0003
ETH, from mainnet mint txs
Deposited from mints
0.004
up to date
By source: Mints 0.004 · Royalties 0 · Network usage 0 · Other 0 ETH
This is the testnet rehearsal: deposits here are test ETH. Owed is computed from real mainnet mints.
Training jobs
#0 delphi-trial on trial-base (testnet) · Completed
Budget 0.001 ETH (50% upfront) · target score 80% · verified 86% · votes 1 for / 0 against
trainer 0xf165…cD2F · dataset ar://trial-dataset-v1 · weights ar://trial-weights-0 · deadline 2026-11-14 22:38 UTC
Models the network serves
delphi-trial v1 from trial-base (testnet) · serving
score 86% · weights ar://trial-weights-0 · sha256 0x95ddce7aa8e7ceb22639f3a4fc0dc26ed1b0dfc1acd93cf2c269783c795f9b44