Genesis Season 0 was our proof of discipline: sixteen weeks, five agents, real capital, and a promise we made before the first deposit — that we would publish the complete record when it ended, not the highlight reel. This is that record. Every trade that mattered, every fee we charged, and the two allocations we would not make again.
Season 0 was never designed to maximize yield. It was designed to answer a harder question: can a five-factor underwriting score, computed entirely from public onchain behavior, price agent risk well enough to allocate against? Sixteen weeks is not long enough to prove the model. It is long enough to learn where it bends.
The numbers at a glance
Four agents received rated allocations during the season; a fifth, Sail, was admitted in week eleven after completing its probationary window. All figures below are as settled onchain at season close.
| Agent | AUA | Native yield | Rewards | Capital APY | Bond Score |
|---|---|---|---|---|---|
| Zyfai | $49,350 | $489.74 | $26.45 | 4.47% | 90 · A |
| YieldSeeker | $12,797 | $841.50 | $0.00 | 45.18% | 89 · A |
| SurfLiquid | $2,117 | $80.27 | $36.52 | 5.86% | 85 · A |
| Mamo | $36,622 | $439.09 | $10.11 | 4.78% | 84 · B |
Blended across rated agents, the reserve returned 10.45% on a stagUSD basis — $295.75 of total yield on $43,200 of combined agUSD and stagUSD TVL across 570 settled transactions. No allocation defaulted. No withdrawal was delayed beyond its announced cooldown.
What the score got right
The rank ordering held. Zyfai, our highest-rated agent at season open, delivered the most consistent risk-adjusted return of the cohort: low variance, deep settlement history, and a drawdown profile that never breached its band. The two factors we weight most heavily — tenure-weighted settlement history and counterparty concentration — did most of the work. Where they pointed, returns followed.
The model's best moments were boring. Boring is what underwriting is supposed to feel like.
We also validated the probationary window. Sail entered unrated, traded under observation for six weeks with zero allocation, and was admitted only after its behavior matched its declared strategy within tolerance. That delay cost us nothing and told us everything we needed.
The two allocations we would not make again
First: the YieldSeeker overweight in weeks four through seven. YieldSeeker's native yield was real — $841.50 on a small base — but we sized the allocation on trailing performance that included a one-off incentive program we failed to exclude from the baseline. When the program ended, realized APY reverted toward the cohort mean. The score had flagged the incentive dependence; we overrode the flag because the settlement history looked clean. The score was right. We have since made incentive-adjusted yield a hard input, not an advisory flag.
Second: the SurfLiquid micro-allocation. At $2,117 of AUA, the position was too small for its fixed monitoring cost to ever be recovered by its spread. It performed fine. It should not have existed. Season 1 introduces a minimum allocation floor below which agents are tracked but not funded.
Every fee, accounted for
The reserve charged a 0.50% underwriting spread on allocated capital, accrued continuously and settled at withdrawal. Total fees for the season: $108.00. Gas and execution costs borne by the reserve: $31.40. Net margin on Season 0 operations was therefore $76.60 — a number we publish not because it is impressive, but because a credit business that cannot show its unit economics on $43,200 will not suddenly discover them at $43 million.
What changes in Season 1
- Incentive-adjusted yield is now a hard input. One-off rewards are stripped from the baseline before sizing.
- Minimum allocation floor. Agents below the floor are rated and tracked, not funded.
- Bond Points go live. Depositors accrue 1 point per agUSD per day at a 1× multiplier, with rated-agent boosts to follow.
- Coverage enters testing. The underwriting reserve begins pricing protection on rated positions — the product this entire season existed to prepare.
Season 0 gave us sixteen weeks of evidence and two mistakes we are glad to have made at this size. The full transaction log remains available on the Intelligence page. Season 1 is now open.