Economics & Incentives

Status: Draft (tracks #12) Phase: 0 — design Related: aegis-chain-design.md §Validator Integrity, staking-systems.md Incorporates: Bob's staking-systems.md findings + Jonto's 2026-04-17 gas-token decision.

Purpose

Fund validator inference, reward honest screening, punish negligence — without making the chain economically unattractive or centralizing around the cheapest model.

Goals

Validators profitably run Tier 1+2 on 100% of txs and Tier 3 on escalations.
Users pay a predictable premium over a vanilla L2; target ≤2× base gas at mainnet v1.
Incentives favor model diversity, not a cheapest-model race.
Slashing punishes clear negligence, not judgment calls.
Core DeFi primitives (AMM, lending, bridging) run natively on the chain so fees flow to validators structurally, not just through gas surcharge.

Architecture (after merging `staking-systems.md`)

Four-layer stack:

┌─────────────────────────────────────┐
│ L4: UMA-style council                │  Bonded dispute / DVM vote on escalations.
│      Bond + dispute + vote           │
├─────────────────────────────────────┤
│ L3: EigenLayer AVS (v2, deferred)    │  Restaked-ETH path. Designed-for, not
│      Designed-for pluggability       │  built in v1.
├─────────────────────────────────────┤
│ L2: Aegis native staking (v1)        │  Validator bonds AEGIS + ETH.
│      AEGIS stake, ETH stake          │  Slashable for screening negligence.
├─────────────────────────────────────┤
│ L1: Ethereum consensus               │  Base settlement. ETH is the money.
│      ETH staking, OP Stack L2        │
└─────────────────────────────────────┘

Revenue sources

Source	How	v1 status	Notes
Gas surcharge	`base_gas + aegis_fee`, fee to validator pool	✅ primary	Paid in ETH (or chain-native stable), not AEGIS — Jonto decision 2026-04-17
Chain-native AMM fees	Aegis embeds Solidly-style AMM; swap fees → validator pool	✅	Per DeFi-in-a-Box
Chain-native lending interest	Aave-style pool lending; spread → validator pool	✅	Per DeFi-in-a-Box
Chain-native bridging fees	LayerZero-style native bridge; fees → validator pool	✅	Per DeFi-in-a-Box
Inflation	Native `AEGIS` emission to active validators	✅ (bounded bootstrap)	Capped schedule TBD
Insurance premium	High-TVL protocols pay for enhanced scrutiny	⚠️ optional	Pricing open question
MEV share	Validators get a cut of proposer MEV	❌ v1	Pulls focus from screening; revisit v2
Restaked-ETH yield (EigenLayer AVS)	ETH restakers delegate; extra yield	❌ v1, ✅ v2	External dep risk for v1

v1 mix: gas surcharge + chain-native DeFi fees + bounded inflation + optional insurance tier. Users never pay gas in AEGIS.

Gas token decision (Jonto — 2026-04-17)

Users pay gas in ETH or a chain-native stablecoin. NOT AEGIS.

Rationale:

AEGIS price volatility would propagate directly into user gas cost.
Removes AEGIS speculative premium from every transaction.
Lowers onboarding friction: users need ETH (which they already have) instead of acquiring a new token just to transact.

AEGIS remains required for staking and governance — not transacting.

Cost sanity check

From aegis-training-plan.md: ~$700/mo per validator at ~10M txs ≈ $0.00007/tx. At 50–100 validators the per-tx validator cost pool is $0.004–$0.007 — well within a reasonable L2 surcharge.

Slashing (starting values, to tune)

Condition	Severity	Escalation
Approved a post-facto-confirmed exploit	100%	Immediate after council confirmation
Stale `profile_root` (see soul-hash.md)	1% per offense	Capped at 10% per epoch
Screening timeout >2s	0.1% per offense	Warn at 3, slash at 5 in 24h
Consistently outvoted on escalations	Warning	Pattern review after 100 disagreements

Validator CAPTCHAs

Periodic replay of known exploits injected into the screening queue, marked as drills post-hoc. Failing = warning; repeated failure = slash. Anti-memorization: drills mutate parameters over time.

Council — UMA-style bond + dispute + DVM

For escalation arbitration (not real-time screening), Aegis adapts UMA's optimistic oracle pattern:

Guardian / validator flags suspicious tx
    │
    ▼
Assertion posted with bond (bond scaled to claim severity)
    │
    ├── No disputer within window → accepted automatically
    │
    └── Disputed → council vote (DVM analogue)
                  │
                  ├── Confirmed exploit → 100% slash + bond rewarded to disputer
                  │
                  └── False positive → assertion bond slashed; asserter rewarded nothing

Parameter	v1 value	Notes
Liveness window	15 min	Fast path for L2 cadence; UMA uses hours
Bond size	Scales with claimed loss (0.1% of TVL at risk)	Prevents frivolous claims
Council size	7 (5-of-7 quorum)	Term-limited, staggered rotation
Vote duration	6h (v1) → 24h (later)	UMA DVM is 48–96h; too slow for chain rhythm

Not used for real-time screening (latency too high). Council handles disputes and policy, not the hot path.

Chain-native DeFi — per DeFi-in-a-Box

Core primitives embedded in the chain:

Primitive	Style	Purpose
AMM	Solidly / Uniswap v3	Fee generation; LP for validators
Pool lending	Aave / Compound	Validator can borrow against staked position
Bridging	LayerZero / Across / native	Cross-chain asset movement
Governance / gauges	Curve / Solidly	Validator voting on protocol params

Fee routing:

AMM swap fees ────────────┐
Lending interest spread ──┤→ Validator pool ──→ per-validator share
Bridging fees ────────────┤                      (after infra + T3 cost)
Gas surcharge ────────────┘
Inflation (AEGIS) ────────→ Validator pool (bounded schedule)

Agents as hands-off protocol managers (per DeFi-in-a-Box): threshold-based intervention on TVL / utilization anomalies, not active trading. Anomalies file as teacups → feeds training pipeline (see memory-strategy.md).

Token (AEGIS) utility

Stake — required to validate, slashable
Governance — vote on council seats, slashing params, profile epoch cadence, protocol param changes
Not required for users to transact (gas = ETH or stable; see above)
Not a speculative trading vehicle in v1 — no MEV share, no protocol utility beyond stake + governance

Anti-capture

Model-diversity bonus: small multiplier for validators running uncommon models — discourages everyone running the same cheap inference.
Council term limits + staggered rotation against long-term human-layer capture.
Restaked ETH path designed for (EigenLayer AVS in v2) so the validator set won't be purely AEGIS-token-captured once that ships.
Domain-specific reputation (trust ledger, see memory-strategy.md) so a single captured domain doesn't corrupt the whole validator set.

Numeric model

First-pass per-validator P&L under v1 assumptions lives in economics-model.csv. Summary from the base case (gas surcharge + inflation + insurance only — DeFi primitive fees not yet modeled):

Line	Value
Monthly validator pool (gas + inflation + insurance)	$6.73M
Per-validator share (75 validators)	$89.7k / mo
Per-validator infra + T3 cost	$830 / mo
Per-validator gross profit	~$88.8k / mo
User cost per tx (base L2 + surcharge)	$0.008
Surcharge multiple over vanilla L2	1.67× (target ≤2×)
Break-even surcharge (infra only)	$0.000087 / tx

Not yet in the model — chain-native DeFi fee flows (AMM / lending / bridging). Those are upside over the base case; worth modeling once TVL and fee-tier assumptions firm up.

Every row is tuneable. Open-question rows are explicitly flagged in the CSV.

Acceptance criteria

Open questions (for Bob / Jonto)

Inflation tail — cap total supply, or indefinite low inflation?
Insurance pricing — TVL oracle, or self-declared tier?
Council composition — UMA-style token holders, or designated multisig set with term limits?
EigenLayer timeline — do we publish the v2 AVS interface now so ecosystem can build against it, or wait until v1 ships?
Chain-native DeFi scope — just AMM + lending, or full DeFi-in-a-Box stack (CDP stable, NFT marketplace, etc.)?
AEGIS genesis allocation — who gets what at launch? Validators, early users, treasury, team?