Product tiers — honest

Yield packages

Three paper portfolios with different mechanics: unlevered lending, lending with a Pendle PT fixed rate, and lending with a simulated loop. Each has its own operating status, data, position decision and statistics of the current version. A green status means “the process runs and its data is fresh” — not a risk grade and not approval for real money.

Conservative

Paper · running data complete and fresh

Real capital: not approved paper validation; real capital only after the go-live gates and an owner decision

Research label: PAPER · evidenced track

Mechanic of the current version: unlevered stablecoin lending conservative-lending-v1
Protocol tiers held (registry labels): T1: aave_v3, compound_v3 · T2: fluid_fusdc, maple, morpho_blue_base a tier describes the protocol, the mechanic describes the strategy; one does not replace the other
Position and decision: position open 5 lending positions, cash $5,000 no rebalance: the gain does not beat the cost band
Paper result of the current version: 4.9% annualised, realized on paper 112 days since 2026-06-22 · book value $101,634 on a $100,000 start · measured 2026-10-13 (published weekly) net of modelled gas and slippage of the book’s own moves (charged since 2026-09-10); real execution costs not included
Statistics maturity: reportable — 113 valid days of the current version through 2026-10-14 (report from 30)
Risk profile: Low the status colour is not a risk grade
Observed drawdown: 0.04% over 112 days of the current version — a past observation, not a loss limit
Stop rule: −5% from peak → no new positions (hold and reduce only); −10% → all to cash
Last successful run: 2026-10-14 06:00 UTC daily (cycle at 08:00 local) code: the tree matches origin 186180e2a342 (checked 2026-10-14 06:00 UTC) status from the published snapshot 2026-10-14 06:01 UTC

Research target — not a result, not a forecast: up to 6%

⚠ The real, evidenced number. No leverage, no tail-compensation yield. This is what "safe" honestly means here.

Full risk sheet

Research target band: up to 6% · target band, annualised, before real execution costs

T1/T2 stablecoin lending — Aave, Compound, Morpho, Spark. Deterministic RiskPolicy gate, ≥$5M TVL floor, per-protocol caps. This is the book the desk actually runs on paper today.

The real, evidenced number. No leverage, no tail-compensation yield. This is what "safe" honestly means here.

L3 · paper track (evidenced)

Balanced

Paper · status not confirmed data out of date

no confirmed run for 21.1 h — status not confirmed

the last observation is 21.1 h old

Real capital: refused research paper portfolio outside the go-live track — refused for live capital

Research label: RESEARCH · not live-allocated

Mechanic of the current version: stablecoin lending + Pendle PT fixed rate held to maturity balanced-fixed-carry-v1
Protocol tiers held (registry labels): T2: fluid_fusdc, maple, morpho_steakhouse · tier unresolved: susde a tier describes the protocol, the mechanic describes the strategy; one does not replace the other
Position and decision: position open 1 PT leg(s) held to maturity, $24,874 the PT fixed rate (4.968 %) is below the floor (5.514 %) set by the floating benchmark
Paper result of the current version: Statistics of the current version are accumulating 13 of 30 valid days (through 2026-10-14) — no result is published before 30 net of modelled costs (gas, slippage, Pendle fees on PT trades); real execution costs not included
Statistics maturity: accumulating — 13 valid days of the current version through 2026-10-14 (report from 30)
Risk profile: Medium the status colour is not a risk grade
Observed drawdown: not published while the current version’s statistics accumulate
Stop rule: −8% from the current version’s peak → the book stops trading; positions are kept (a stop trigger, not a maximum-loss guarantee)
Last successful run: 2026-10-14 05:09 UTC hourly; one accounting row a day code: the tree matches origin 186180e2a342 (checked 2026-10-14 06:00 UTC) status from the published snapshot 2026-10-14 06:01 UTC

Research target — not a result, not a forecast: up to 12% Historical tail: ~4.5% drawdown (backtest of an earlier hedged sUSDe book, not the current version)

⚠ Historical tail: ~4.5% drawdown (backtest of an earlier hedged sUSDe book, not the current version)

⚠ earlier period distorted by the identified accounting defect — retained for audit, not shown (39 days)

⚠ Our engine already found this — and its tail. It stays in paper until it clears 30 forward days + a real vol event; that discipline IS the product. Real in the good months, measured in the bad. Join early-access to get in the day it clears.

Full risk sheet

Research target band: up to 12% · target band, annualised, before real execution costs

The Balanced paper book combines floating stablecoin supply with a fixed-rate part: Pendle principal tokens (today PT-sUSDS) bought at a discount and held to maturity, up to 40% of the book and 25% per market. The fixed rate is locked at purchase. No leverage.

Our engine already found this — and its tail. It stays in paper until it clears 30 forward days + a real vol event; that discipline IS the product. Real in the good months, measured in the bad. Join early-access to get in the day it clears.

L2 · paper / backtest

Aggressive

Paper · status not confirmed data out of date

no confirmed run for 21.1 h — status not confirmed

the last observation is 21.1 h old

Real capital: refused simulated leverage, research paper portfolio — refused for live capital

Research label: RESEARCH · simulated loop experiment · refused for live (by design)

Mechanic of the current version: stablecoin lending + a simulated sUSDe/PYUSD loop on Morpho Blue aggressive-susde-loop-v1
Protocol tiers held (registry labels): T2: fluid_fusdc, maple · plus the simulated Morpho Blue loop (sUSDe collateral) a tier describes the protocol, the mechanic describes the strategy; one does not replace the other
Position and decision: position open simulated loop open · debt $116,852 · health factor 1.308 at the last check position open; carry re-checked
Paper result of the current version: Statistics of the current version are accumulating 13 of 30 valid days (through 2026-10-14) — no result is published before 30 net of modelled costs (gas, slippage, swap cost of the loop both ways, borrow interest); real execution costs not included
Statistics maturity: accumulating — 13 valid days of the current version through 2026-10-14 (report from 30)
Risk profile: Higher the status colour is not a risk grade
Observed drawdown: not published while the current version’s statistics accumulate
Stop rule: −25% from the current version’s peak → the loop is unwound and the book stops trading; before that the loop deleverages and unwinds on its own health-factor and USDe-price triggers (a stop trigger, not a maximum-loss guarantee)
Last successful run: 2026-10-14 05:11 UTC hourly; loop supervised every run, one accounting row a day code: the tree matches origin 186180e2a342 (checked 2026-10-14 06:00 UTC) status from the published snapshot 2026-10-14 06:01 UTC

Research target — not a result, not a forecast: up to 20% Historical tail: up to ~50% drawdown (backtest of the earlier Aggressive Lab research book, an unhedged directional book — not the current loop)

⚠ Historical tail: up to ~50% drawdown (backtest of the earlier Aggressive Lab research book, an unhedged directional book — not the current loop)

⚠ earlier period distorted by the identified accounting defect — retained for audit, not shown (39 days)

⚠ This is the 12–20% advertised elsewhere as "yield". Our engine found it AND its ~50% tail — so it REFUSES live capital until it is proven. That refusal is what you are buying: nobody else publishes what they will not touch. Want the validation report first?

Full risk sheet

Research target band: up to 20% · target band, annualised, before real execution costs

The Aggressive paper book holds concentrated stablecoin lending and runs a separate PAPER experiment: a simulated sUSDe-collateral / PYUSD-debt loop on one Morpho Blue market, modelled from that market’s own data (oracle price, debt rate, liquidation threshold, health factor). The loop enters only when its carry beats its costs; until then it holds. No real borrowing, no keys, no transactions.

This is the 12–20% advertised elsewhere as "yield". Our engine found it AND its ~50% tail — so it REFUSES live capital until it is proven. That refusal is what you are buying: nobody else publishes what they will not touch. Want the validation report first?

L2 · paper experiment, statistics accumulating · refused for live

Don't trust us — check us Live track record → Refusal log → Liquidation-NAV by size →

How a tier becomes available

A package is offered ONLY when its strategies pass the DSR gate on real data + the full Yield-Lab lifecycle (yield-source → protocol → liquidity → risk → Red Team → paper test → human approval). Today Conservative is real and evidenced; Balanced and Aggressive are research/paper, shown with the tail, and are NOT live-allocated.

The tail in numbers: 15% vs the realized paper ~4.9% (measured 2026-10-13) → Live risk monitoring → Live track & go-live →

Next — an honest conversation

If the conservative, evidenced track fits how you allocate — this is not an offer, it is a draft for a conversation with a design partner. Non-custodial, human-in-the-loop.

Design-partner pilot →

Non-custodial

We never hold keys, move funds, or sign. Your wallet stays your control.

Honest-first

Every number is evidence-tagged; paper is never shown as live; the tail is always shown.

Public track

A live paper track + go-live, checkable; every hash re-derivable with one zero-dependency verifier.

We show the bad news

A public refusal log, published NO-GOs, and real drawdowns. We publish what we declined.

Paper / advisory. Balanced and Aggressive are outside RiskPolicy, never live-allocated, and never touch the go-live track. Target bands are research targets, not results; a package's result is shown on its card only for the current strategy version and only after 30 valid days. We never present paper/backtest as live.