Staking Systems Research — Aegis Economics

Research Version: 1.0
Author: Bob
Date: 2026-04-17
Status: Research Complete
Drives: Aegis Economics (#12), Validator Security Model

Source Documents

Clark's economics draft: docs/specs/economics.md
Jonto's DeFi in a Box (2024-06-21): https://paragraph.com/@jonto/defi-in-a-box
EigenLayer docs: https://docs.eigenlayer.xyz
UMA Oracle docs: https://docs.uma.xyz
Aegis chain design: aegis-chain-design.md

Staking Systems Evaluated

1. Ethereum Native Staking

Model: POS consensus — validators stake ETH, get rewarded for correct attestation, slashed for double voting / downtime.

What works:

Battle-tested since Merge (Sept 2022), >$50B staked
Slashing mechanics well-understood
Decentralized validator set with no privileged operators
Clear惩罚 (slashing) for malicious behavior

What doesn't work for Aegis:

Designed for consensus, not application-specific screening
No mechanism for task-level rewards (per-tx screening)
Slashing is binary (attested correctly or didn't), not probabilistic
32 ETH minimum = validator centralization risk for smaller chains

Aegis applicability: Base security layer for L2 consensus. Aegis validators must be Ethereum stakers — but native ETH staking alone doesn't solve the screening economics problem.

2. EigenLayer — Restaking

Model: ETH/LST restakers delegate to Operators who secure AVSs (Autonomous Verifiable Services). Restakers earn extra yield; Operators earn fees from AVSs. Slashing flows back to restakers.

Architecture:

Restakers → DelegationManager → Operators → AVSs (Aegis Screening)
                                            ↑
                                    Slash penalties flow back

Contracts (mainnet):

DelegationManager: 0x39053D51B77DC0d36036Fc1fCc8Cb819df8Ef37A
StrategyManager: 0x858646372CC42E1A627fcE94aa7A7033e7CF075A
AVSDirectory: 0x135dda560e946695d6f155dacf6f1f25c1f5af
RewardsCoordinator: 0x7750d328b314EfFa365A0402CcfD489B80B0adda

What works:

AVS ecosystem growing — operators can serve multiple AVSs simultaneously
EigenLayer already deployed on mainnet Ethereum + Base
AgentKit (eigencloud.xyz/agentkit) enables agents to operate as AVS operators — relevant for Aegis agents
Restaked ETH as collateral for Aegis screening is compelling: existing ETH stakers can diversify into screening income without new capital commitment
Slashing penalties flow back to restakers — aligns incentives cleanly

What doesn't work for Aegis:

Cascading slash risk: If an Aegis AVS gets exploited via the validator set, the slash flows back to all restakers who delegated to those operators. This is the main criticism of EigenLayer — correlated slashing.
External dependency: Building on EigenLayer means Aegis's security is coupled to EigenLayer's contract security. If EigenLayer itself has a bug, all AVSs are affected.
AVS registration complexity: AVS must register with EigenLayer's AVSDirectory, define reward/slash parameters. Adds integration overhead.
EigenDA cost: Data availability through EigenDA adds cost; must evaluate vs raw L2 sequencing.

Should Aegis use EigenLayer?

Scenario	Verdict	Reasoning
Aegis validators = ETH restakers via EigenLayer	⚠️ Consider	Clean capital efficiency — ETH stakers earn screening yield
Aegis as AVS on EigenLayer	⚠️ Deferred	External dependency risk for a new chain; revisit post-mainnet
Aegis native staking (no EigenLayer)	✅ Yes for v1	Keep it simple. Add EigenLayer AVS integration in v2.

Recommendation: Build native AEGIS staking for v1. Design the AVS interface so v2 can plug into EigenLayer without changing the core contract logic.

3. UMA — Optimistic Oracle + Bonding

Model: Asserter posts a bond, makes an assertion. Disputer can refute within liveness period. If not disputed → assertion accepted. If disputed → DVM (tokenholder vote) resolves.

Architecture (from UMA docs):

Asserter posts bond → Assertion made
         ↓
    Disputer has window to refute
         ↓
   Undisputed → Accepted
   Disputed → DVM (48-96h vote) → Final resolution

What works for Aegis:

Bond + dispute model maps directly to Aegis escalation: Guardian makes assertion (flag), validator can dispute (escalate), council resolves
Human-in-the-loop for hard cases: DVM (UMA tokenholders) as council — proven mechanism
Economic security = cost of corruption > profit from corruption — the UMA security model
Slow path (DVM) vs fast path (optimistic): Matches Aegis design (fast screening, slow council for disputes)
Bond sizing: Asserter chooses bond size — higher bond = more credible assertion. Aegis could require larger bonds for higher-confidence assertions.

What doesn't work for Aegis:

48-96h DVM resolution is too slow for high-frequency L2 tx — but works for council-level disputes
UMA token dependency: DVM voters must hold UMA — adds another governance token
** UMA token price volatility** affects security budget (cost of corruption measured in UMA)

Aegis applicability:

Council escalation model: Use UMA-like bond + dispute + DVM for council arbitration
Guardian assertions: Bonded assertions at Tier 1/2 level (not needed for every tx — only escalated cases)
Not for core screening: Too slow for 100% tx screening

4. L2 Native Staking (Arbitrum, Optimism, Base)

Model: Sequencer/staker hybrids — validators stake L2 tokens to participate in fraud proof or validity proof systems.

What works:

Base: No native staking token — relies on Ethereum security. Clean for Aegis.
Arbitrum: Uses ETH staking for validators. Validators must stake ETH to run a validator node.
Proven at scale: Arbitrum has $15B+ TVL, Optimism $10B+.

What doesn't work for Aegis:

L2 staking is primarily for sequencing/fraud proof participation, not task-specific screening
No built-in mechanism for per-tx screening rewards

Aegis applicability: Base chain should use Ethereum native staking (no L2 token). Aegis validators stake ETH + AEGIS, not an L2 token.

5. Berachain / Chain-Native DeFi (from DeFi in a Box)

Model: Chain itself deploys core DeFi primitives (AMM, lending) and collects fees. Validator = chain operator + DeFi participant.

What works:

Proven by Berachain: Built-in PMM (Polaris AMM) + lending, chain collects fees
Eliminates protocol TVL migration risk — TVL stays on chain by default
Validator incentives are structural: Validators benefit from chain fee revenue
Agents managing protocols = chain-native: From DeFi in a Box — agents manage protocols, chain provides the infrastructure

From DeFi in a Box (Jonto):

The Core DeFi Stack required for a new chain:

Required: AMM (Uniswap v3/Solidly), Pool Lending (Aave/Compound), Bridging (LayerZero/Across/Native), Governance/Gauges (Curve/Solidly)
Nice to Have: Token/NFT Factory, NFT Marketplace, CDP Stablecoin, Regulated Stablecoin
Hurdles: Oracles (Chainlink), Bridges

For Aegis:

If Aegis runs chain-native DeFi, the validators are also LP providers and borrowers
Screening becomes a fee-generating service ON the chain, not just a security cost
Agents as protocol managers fits the model — agents manage the lending pool, AMM, bridging

Aegis applicability: Strong fit for the DeFi primitives layer. If Aegis embeds core DeFi:

Fees from AMM/lending go to validator pool (security + operations funded)
Agents manage protocols (per DeFi in a Box)
Screening = embedded service, not external cost center

Synthesis: Recommended Architecture

Layer 1: Ethereum Security

Validators stake ETH natively (or via liquid staking tokens)
No new L2 token required for consensus

Layer 2: Aegis Native Staking (v1)

Validator stakes AEGIS + ETH
    │
    ├── Screening rewards (gas surcharge)
    ├── Inflation (bounded bootstrap)
    └── Slashing (negligence)

Layer 3: EigenLayer AVS (v2, deferred)

Restakers → EigenLayer → Aegis as AVS
    │
    └── Extra yield for ETH restakers
        │
        └── Aegis validators as Operators

Layer 4: Council Arbitration (UMA-style)

Guardian assertion (bonded)
    │
    ├── Disputed → Council review
    │       │
    │       └── DVM vote (48-96h)
    │
    └── Undisputed → Accepted

Chain-Native DeFi Primitives for Aegis

Per Jonto's DeFi in a Box framework:

Primitive	Implementation	Purpose
AMM	Solidly-style (chain-native)	Fee generation, LP for validators
Pool Lending	Aave/Compound-style	Validator can borrow against staked position
Bridging	LayerZero or native	Cross-chain asset movement
Governance	Curve/Solidly gauges	Validator voting on protocol params

Agents as Protocol Managers:

Agents don't actively trade — they monitor and manage protocols
Threshold-based intervention: if TVL drops, liquidity incentives auto-adjust
Agents file teacups on protocol anomalies → feeds training pipeline
Largely hands-off, per DeFi in a Box — "agents manage the protocols but they're largely hands off"

Fee Flows:

AMM fees ──────────────→ Validator pool
Lending interest ──────→ Validator pool  
Bridging fees ─────────→ Validator pool
Screening fee (gas) ───→ Validator pool
Inflation (AEGIS) ──────→ Validator pool (bounded)

Decision Matrix

System	Use for Aegis	Rationale
Ethereum native staking	✅ Consensus layer	Required, battle-tested
AEGIS native staking	✅ v1	Core validator bonding + slash
EigenLayer AVS	⚠️ v2 deferred	External dependency risk for v1; compelling for v2
UMA-style council	✅ Escalation/arbitration	Bond + dispute + DVM fits council model
L2 token staking	❌ No	No L2 token for Aegis
Chain-native DeFi	✅ Core primitives	Per DeFi in a Box — AMM + lending + bridging
Berachain model	✅ Reference	Chain collects DeFi fees = sustainable validator income

Open Questions

EigenLayer timeline: When does Aegis v1 ship vs when EigenLayer AVS registration stabilizes?
Chain-native DeFi scope: How deep does Aegis go? Just AMM + lending, or full DeFi stack?
AEGIS token distribution: Is there an inflation schedule? Who gets genesis allocation?
Council composition: UMA-style tokenholders or designated set? Term limits?
Insurance tier: High-TVL protocols pay Aegis for enhanced screening — how is price set?

Recommendations

Skip EigenLayer for v1 — too much external dependency coupling. Design for v2 pluggability.
Use UMA-style bond+dispute for council escalation — proven pattern, matches Aegis Guardian design.
Build chain-native DeFi per DeFi in a Box — AMM + lending as core primitives. Validators = fee recipients.
Agents as hands-off protocol managers — per Jonto's model. Only intervene on threshold breaches.
AEGIS token = staking + governance only — no speculative trading utility in v1.

Design Decision (Jonto — 2026-04-17)

Gas token: ETH or chain-native stable. NOT AEGIS.

Users pay gas in ETH (or stable) — no AEGIS required to transact. AEGIS is staking + governance only.

This avoids:

AEGIS speculative premium baked into every tx cost
Token price volatility affecting user gas costs
New token friction for onboarding users