We searched ETHGlobal showcases for projects in the AI-agent + governance space. Four direct competitors emerged. This page is honest about what they built well and where AI Treasury Council goes further. Source: internal competitive analysis (Sesja 34) plus public ETHGlobal showcase data.
1. Why we are different (TL;DR)
Three differentiators that no direct competitor matches:
- Source attribution per claim. Every agent statement cites a URL, timestamp, and confidence score. Most competitors output prose with no citations.
- Five trust mechanisms wired in the demo. Source attribution + 48h timelock + 0G audit + ENS reputation + HITL Council Rules. Competitors typically wire one or two.
- Native DAO governance integration. OpenZeppelin Governor + Timelock, 5 contracts deployed and verified on Base Sepolia. Competitors mostly mock this layer.
2. Direct competitors
These are the four projects most adjacent to AI Treasury Council. We evaluated them on what they built, what they did well, and where our approach differs.
Goldman Stacked
| Aspect | Detail |
|---|---|
| What they built | AI-driven trading desk simulation with multiple agent personas executing trades |
| What they did well | Polished UI, clear narrative around AI-as-trader, strong visual debate panels |
| What we do differently | We target DAO governance (not retail trading), publish reasoning per claim with sources, write to on-chain audit trail (0G), enforce 48h timelock before any treasury action |
| Our positioning vs them | They optimize speed for retail; we optimize trust for treasuries. Different markets, no overlap. |
Agentropolis
| Aspect | Detail |
|---|---|
| What they built | City-simulation game where AI agents act as autonomous citizens with economic agency |
| What they did well | Compelling demo of emergent agent behavior, strong storytelling, agent-to-agent interaction patterns |
| What we do differently | Real on-chain treasury actions vs simulation. Audit trail required. DAO governance integration vs game mechanics. |
| Our positioning vs them | Agentropolis explores 'what if agents had economy'; we ship 'what DAOs need today'. Different time horizons. |
Alpha Dawg
| Aspect | Detail |
|---|---|
| What they built | AI alpha-discovery agent for crypto research with sentiment + on-chain signals |
| What they did well | Strong data integration (multiple sources), good signal fusion, real iNFT (ERC-7857) for agent identity |
| What we do differently | Group debate (5+1) vs single agent. Source attribution per claim vs aggregated alpha score. Governance execution layer vs research-only output. |
| Our positioning vs them | They give an alpha researcher a tool; we give a DAO a council. Complementary - their output could be a AI Treasury Council agent's input. |
Ghost in the Machine
| Aspect | Detail |
|---|---|
| What they built | Privacy-focused AI agent execution in TEE (Trusted Execution Environment) with verifiable outputs |
| What they did well | TEE integration is genuinely advanced, hardware-attested execution is rare in hackathons, strong privacy story |
| What we do differently | Public-by-default for governance transparency vs private execution. Multi-agent debate vs single agent. Live deployed on testnet vs concept-heavy. |
| Our positioning vs them | Their TEE story is something we plan for Phase 4 (sensitive personas). For DAO governance, transparency beats privacy at the agent layer - the timelock provides the privacy window. |
3. Where we win
Feature-by-feature comparison. YES means shipped in the demo, partial means working but limited, no means not present.
| Feature | AI Treasury Council | Competitors (typical) |
|---|---|---|
| DAO governance integration (Governor + Timelock) | YES | no (mocked) |
| 0G Storage audit trail | YES | rare |
| ENS deep records (26 text records per subname) | YES | rare / shallow |
| Adversarial agent (forces dissent) | YES (opt-in) | no |
| Source attribution per claim | YES | no |
| Real on-chain demo (5 contracts deployed + verified) | YES | mock-heavy |
| 48h timelock with countdown UI | YES | rare |
| Multi-agent debate (5+1 personas) | YES | 1-3 typical |
| HITL Council Rules JSON Editor | YES | no |
| AgentReputation on-chain (Moat 5) | YES | no |
4. Where they win (honest)
Things competitors do that we explicitly don't - either because we cut scope or because the trade-off didn't serve our use case.
- Real iNFT (ERC-7857): Alpha Dawg ships this today. We have ENS subnames + on-chain reputation but not iNFT yet. Planned for Phase 1 mainnet deploy (Q3 2026).
- TEE (Trusted Execution Environment): Ghost in the Machine has hardware-attested execution. We run in a standard FastAPI process. TEE is on the Phase 4 roadmap for sensitive agent personas where privacy beats transparency.
- Full multi-chain: Some competitors deploy across multiple L2s. We ship on Base Sepolia (governance + treasury) and Sepolia (ENS only) for the hackathon. Phase 1 adds Optimism mainnet.
- Polished animations / 3D UI:Goldman Stacked has more eye-candy. We prioritized correctness, audit verifiability, and judge-readable docs over visual polish. Vela & Aiko have a Phase 1 design pass scheduled.
5. Our positioning
Alpha Dawg targets research, Goldman Stacked targets trading. Long-term, we compete with non-AI governance tools (Snapshot, Tally, Aragon) by giving DAOs a verifiable AI layer that makes their existing governance smarter without replacing the human vote. The 48h timelock and 5-of-7 multisig keep humans in the loop on every action.
For judges: the differentiator that matters most is source attribution per claim. It's the foundation that makes the other four trust mechanisms credible. Without it, you have an opinion engine. With it, you have a council you can audit.