We forecast that by 2030, the most consequential decentralised organisations will be run jointly by humans and autonomous software agents — humans setting values, policies, and vetoes; agents doing the reading, forecasting, negotiating, and most of the executing. This page presents that forecast the way AI 2027 presented its scenario of superhuman AI[1]: as a concrete, year-by-year narrative, grounded in cited evidence, with explicit probabilities and a public registry of falsifiable claims.
Two rules govern every number on this page. Values tagged MEASURED are sourced — hover any citation mark to see the publisher, date, and source-quality tier, or jump to the references. Values tagged FORECAST are model outputs — our best estimates, stated with uncertainty ranges so they can be scored, and wrong in ways we commit to publishing. This is not a recommendation or a manifesto. The goal is predictive accuracy.
A $20B+ experiment that outgrew its operators
Decentralised autonomous organisations went from a whitepaper curiosity to a balance-sheet fact in half a decade. Total DAO treasuries grew roughly 40× during 2021 — from about $400M in January to $16B by September[3] — and printed a then-record $25.1B in March 2023, equal to roughly 40% of all DeFi TVL at the time[2]. As of 2025 the ecosystem holds $21.4B liquid ($24.5B total) across more than 13,000 DAOs, with over ten million governance token holders[4][5]. The canonical live tally is DeepDAO[6]; treasury figures swing with token prices, which is why every number here carries its date.
The bottleneck is human, and it is measured
The founding promise was distributed decision-making. The measured reality is the opposite. Across ten major governance tokens, less than 1% of holders control 90% of voting power[7]. The top decile of voters casts 76.2% of all votes, and average participation runs at 6.3% — against roughly 70% at traditional corporations[8]. Voting-power Gini coefficients reach 0.99 at Compound, ENS, and Uniswap — more concentrated than any national wealth distribution on record[9]. In one academic sample of 126 Snapshot proposals, a median of three wallets was enough to decide the outcome[10], and the active-to-eligible voter ratio has been falling toward 1–2% in recent votes[11]. Arbitrum, one of the best-resourced DAOs in existence, has 986 registered delegates of whom 91 actually vote[12].
The theory: 10M+ governance token holders, one grid, every dot equal. (1,000 dots shown.)
Data-grounded visualisation — the highlighted counts are the cited statistics drawn to scale: voting-power concentration[7], average turnout[8], and the median-3-wallets result[10].
This is not apathy; it is arithmetic. Snapshot alone has processed on the order of 130,000 proposals and 20 million votes[14]. A 2026 econometric study found exactly what every delegate already knew: once proposal flow crosses a threshold, voter responsiveness measurably declines and control concentrates — DAOs become, in the authors’ phrase, “too big to monitor”[13]. The ecosystem’s response has been to pay humans to pay attention: Arbitrum budgets ~$1.5M a year just to reward delegates for voting[16][15], Uniswap pays professional delegates $6,000 a month[17], and delegate-pay programs now run at nine of the largest DAOs[18]. Above the delegates sits a professional layer: Gauntlet’s risk contract with Aave ran $1.6M a year — and Gauntlet still quit two months after renewal, citing DAO dysfunction[19]; Arbitrum’s research collective was funded with 1.76M ARB for six months[20]; its Foundation’s proposed 2027 budget is $43.5M[21].
And the process is slow in ways that cost real money. A Uniswap proposal takes a minimum of ~23 days from request-for-comment to execution[22]; an Arbitrum constitutional change takes about five weeks[23]. When a 2021 bug began leaking COMP worth over $80M, Compound’s founder had to explain that no fix could ship for seven days because every change required the full governance cycle[24]. During the March 2020 crash, MakerDAO’s remediation debate played out over two weeks while $8.3M of collateral was liquidated for zero DAI[30]. In 2023, Aave’s token holders voted down their own paid risk manager’s advice while a $176M position loomed over the protocol[31]. By late 2025 one prominent L2 DAO simply paused its governance entirely, conceding the process had broken down[32].
The failure register
Slowness is one tax; capture is the other. The same mechanism that makes DAOs open makes them attackable, and the attacks are now a genre: Beanstalk lost $182M to a flash-loaned supermajority acquired inside a single transaction[25]; the Mango Markets exploiter voted his stolen tokens on his own $47M “settlement”[26]; Tornado Cash’s attacker minted 1.2M fake votes through a hidden contract self-destruct[27]; a whale coalition pushed $24M of Compound’s treasury into a vault it controlled over explicit community objection[28]; and Nouns watched $27M rage-quit through a fork mechanism built precisely because exit had become the only check left[29]. Treasury management compounds the risk profile: Chainalysis found 85% of DAO treasuries concentrated in a single asset — almost always the DAO’s own token[7]; Arbitrum still holds more than 90% of its ~$1.78B treasury in ARB[33], though diversification has begun in earnest — Uniswap moving to diversify beyond UNI[34], Arbitrum allocating 35M ARB to tokenized US Treasuries[35].
ANATOMY OF A GOVERNANCE CAPTURE · REPLAYED WITH THE BEANSTALK NUMBERS — BORROWED SUPERMAJORITY, ONE TRANSACTION
Replayed with the documented Beanstalk figures — a >$1B flash loan bought a 67% supermajority and $182M left in a single transaction[25]; the thin honest turnout that makes the attack possible is the same 1–2% measured above[11].
Why any of this is worth fixing
If DAOs were only slow, captured hedge funds, the right forecast would be extinction. They are not. The same five years produced the clearest evidence yet that internet-native organisations can mobilise capital, culture, and care at speeds no incorporated entity can match. ConstitutionDAO raised $47M from 17,437 strangers in a week (average contribution: about $206) to bid on a founding document[36]. UkraineDAO moved ~$7M in 72 hours to war relief[37]. Gitcoin has distributed $10.4M in 2024 alone across 105 community-decided rounds[38]; Optimism’s retroactive public-goods program has paid out 60M+ OP with a fifth of the entire token supply reserved for more[39]. VitaDAO funds longevity research that pharma passed over — 24 projects and counting, with Pfizer Ventures eventually buying in[40]. Nouns has auctioned one CC0 artwork every 24 hours since August 2021, funding hundreds of cultural experiments from a treasury owned by no one[41]. KlimaDAO pulled ~$100M of carbon credits on-chain[42]. Culture, capital allocation, revenue, and societal good — the four surfaces are real. What they lack is an operating capacity that scales with them.
WHAT THE OTHER COLUMN OF THE LEDGER LOOKS LIKE · TOTALS ARE THE CITED FIGURES
Each cycle replays a cited event: ConstitutionDAO[36], UkraineDAO[37], Gitcoin’s 2024 rounds[38], and Optimism Retro Funding[39].
The convergence already underway
Meanwhile, the missing operator arrived from the other direction. The infrastructure for software agents that hold accounts, pay, and act is no longer speculative: 40M+ smart accounts and 100M+ ERC-4337 UserOperations[43], native account abstraction shipped to every Ethereum wallet via EIP-7702[44], and an agent-payment protocol (x402) that went from launch[45] to 100M+ transactions on Base in three quarters[46] — with the honest caveat that much of that volume was experimentation; real daily dollar volume fell 92% from its December 2025 peak[47]. Amazon, Coinbase, and Stripe now ship joint payment rails for AI agents[48]; a16z’s 2026 thesis is that compliance itself shifts from know-your-customer to know-your-agent[67].
And governance is already the test bed. Arbitrum delegated 7,000,000 ARB of real voting power to an agent-voting pool, then held a 2025 ratification vote on its “Agentic Governance Pivot”[49] — the single clearest signpost this scenario has already passed. Olas shipped an autonomous AI delegate in 2023[50] and now operates millions of agent transactions[51]; Tally’s own CEO built an MCP server so AI agents can query governance state directly[52]; Agora bought Boardroom explicitly to fuse governance infrastructure with AI decision support[53]; researchers are now benchmarking agentic AI against 3,383 real proposals from eight major DAOs[54]. On the decision-market flank, MetaDAO runs live futarchy on Solana[55], and both Optimism (500K OP) and the Uniswap Foundation ($900K) have run real conditional-funding-market experiments[56][57] — with Optimism’s post-mortem honestly flagging that the chosen metric got gamed. Prediction markets graduated from curiosity to institution: $73M of Polymarket volume in 2023 became roughly $10B a month by March 2026 — about $24B monthly across prediction markets combined[59] — with NYSE’s parent investing $2B[58].
That is the baseline: organisations holding twenty billion dollars, run at single-digit human attention, taxed by capture and latency, demonstrably capable of extraordinary collective action — and an agent stack arriving at exactly the moment the monitoring math breaks. What follows is our best estimate of how those curves intersect.
Agents stop advising and start preparing
Through 2026, AI in governance is a reading aid: summaries, sentiment, risk flags. In 2027 the interface inverts. The standard unit of governance work becomes the prepared action: an agent assembles the entire decision — draft, simulation, citations, counter-arguments, forecast with stated confidence — and a human reviews one screen and signs. Stage-and-sign spreads for the same reason spreadsheets did: it converts hours into minutes without moving the locus of authority. Delegate programs quietly rewrite their terms — paying for attendance[17] stops making sense when attendance is automated, so programs start paying for judgment with a track record.
Three mechanical shifts compound. First, governance data goes machine-readable by default — the MCP pattern Tally pioneered[52] becomes table stakes, and intelligence demand begins routing through agents instead of dashboards. Second, proposals start shipping typed: structured intent, budget manifest, and an attached machine forecast — first as a competitive courtesy, then as a social norm at the DAOs where machine analysis already covers a third of the docket. Third, the first accountable agent delegates register: descendants of Governatooorr[50] and the Arbitrum pool experiment[49], but now with published voting policies, per-vote rationales, and public calibration records — because after the first black-box agent scandal, “show me your Brier curve” becomes the delegation pitch.
- ▢≥25% of top-50 DAO proposals carry machine-generated analysis at posting time (we forecast 80% likely by end-2027)
- ▢A top-20 DAO registers an agent delegate with ≥0.5% of active voting power and a published policy (75%)
- ▢At least one delegate-incentive program adds calibration or track-record terms to its payment criteria
- ▢Refutation: if by end-2027 agent delegates remain pure novelty accounts with no published policies, this chapter is running slow
Governance forks into two lanes
The monitoring-limits result[13] has a corollary the ecosystem finally acts on: if humans cannot monitor everything, stop making them. In 2028 the leading DAOs formally split decision flow into two lanes. The machine lane handles routine, reversible, capped operations — parameter tuning, recurring payments, rebalancing inside policy bands — executed by agents under typed policies with veto windows: every action announced, attested, and reversible for N hours before settlement. The human lane keeps what it should never have shared: values, constitutional change, large irreversible allocations — now reached by humans whose reading is done for them. The seven-day patch cycle that once leaked $80M[24] becomes the canonical argument: latency is not deliberation, it is exposure.
THE TWO-LANE SPLIT, RUNNING LIVE · ILLUSTRATIVE — RED DOTS ARE VETOES BEING HELD, WHICH IS THE DESIGN WORKING
The license to operate in the machine lane is measured calibration — an agent’s authority budget scales with its scored forecast record, and demotion is automatic on breach. Treasury management is the killer app: after a decade of 85%-single-asset balance sheets[7] and the STEP precedent[35], boards of token holders approve standing policies (“stables 20–40%; rebalance on 5% deviation; veto window 72h”) and let bonded agents execute them continuously. Forecast-conditioned execution — futarchy’s practical descendant, de-risked by the Optimism and Uniswap experiments[56][57] — gates a growing share of grants: funds release when calibrated predicted impact clears a bar, with the metric-gaming lesson from Optimism’s v1 baked into metric choice. The new attack class arrives on schedule: agent-scale sybils. Wallet-age heuristics die; defenses move to costly identity — graph history, stake, attestation chains[67].
THE MACHINE-LANE ACTION LIFECYCLE · ILLUSTRATIVE
- ▢A major DAO ratifies a graduated-autonomy framework by governance vote (70% by end-2028)
- ▢≥10% of routine actions at top-100 DAOs execute through policy modules with veto windows
- ▢A delegated agent is publicly demoted/slashed for a policy breach — and the system is praised for it
- ▢LLM forecasters reach parity with human pros on a public benchmark (60% by end-2028)
- ▢Refutation: a string of veto-window failures or a major policy-module exploit would push adoption back 18+ months
The inter-DAO economy — and the Helvetia incident
Q1. The first agent-negotiated inter-DAO deal closes: two mid-cap DAOs’ treasury agents negotiate a liquidity-sharing agreement inside their mandates — typed term sheets, both sides’ forecasts attached — and both governances ratify within a week. A negotiation that previously meant months of forum diplomacy[22] becomes a reviewable artifact. The template spreads through grants co-funding and service agreements; agent-to-agent commerce protocols built for the consumer agent economy[68] get repurposed for org-to-org deal flow.
AGENT-MEDIATED DEAL FLOW BETWEEN TWO DAOS · ILLUSTRATIVE — VIOLET PULSES CROSS THE NEGOTIATION BRIDGE
Q2. Attestation becomes an industry. Every consequential agent action now carries a verifiable chain — identity, authority, evidence hash, forecast — and underwriters price coverage on attestation history, the way auditors once priced trust in quarterly statements. Treasuries under continuous policy management approach a fifth of the top-100 total; prediction markets, now institutional plumbing[58][64], supply the reference prices that keep treasury agents honest.
EVERY AGENT ACTION FALLS THROUGH THE ATTESTATION STACK · ILLUSTRATIVE — HOVER A LAYER
Q3 — the incident. A coordinated swarm hits Helvetia Protocol, a mid-cap DeFi DAO. (Helvetia is fictional — a composite, for the same reason AI 2027 invented OpenBrain[1]: every component of this attack has a named real precedent.) Several thousand aged, graph-plausible sybil identities accumulate small delegations for months. Compromised “analysis agents” seed subtly biased risk assessments into forums — the adversarial-input failure Buterin warned about in 2024[65]. Then, in one weekend window, a flash-amplified delegation surge[25] pushes a treasury-routing proposal past quorum with forged advisory consensus[27]. $140M is at stake. The attack is caught — not by a human reading the forum, but by an attested monitoring agent flagging that the “independent” analyses share a statistical fingerprint. The veto window holds. Partial damage: $9M routed through a side path settles before reversal.
Q4. The post-mortem splits the ecosystem. Everyone now agrees agents run governance; the question is under whose rules. One camp doubles down on cheap, fast, opaque autonomy — the attack was caught, after all. The other concludes that opacity itself was the vulnerability, and that accountability has to be infrastructure, not etiquette. Regulators, watching $140M nearly move on forged machine consensus, draft in both directions at once.
The week after the Helvetia incident, every major DAO faces the same vote the reader faces now: keep buying opaque autonomy because it is cheap, or pay the switching cost of making every agent accountable. Pick the branch you find more plausible — then read the other one.
The autonomy ladder: how organisations actually get there
Both endings run through the same five rungs — they differ only in whether the gates between rungs are enforced. This is the scenario’s spine, and it is buildable today: every rung has a 2026 precedent already cited above.
| Level | What the agent does | Gate to enter | 2026 precedent |
|---|---|---|---|
| L0 · Observe | Reads, summarizes, alerts. | None — the 2024–26 status quo. | Summary copilots |
| L1 · Advise | Forecasts, verdicts, drafts — every claim carrying provenance. | Citations on every claim; no fabricated data paths. | Governance AI tools |
| L2 · Prepare | Stages the complete action; a human signs. | A public calibration record exists and is good. | Stage-and-sign pilots |
| L3 · Execute-with-veto | Acts inside policy after an announced veto window. | Scored Brier below threshold over N predictions; clean attestation history; bounded blast radius. | Policy modules, timelocks |
| L4 · Bounded autonomy | Operates the machine lane continuously within hard budgets. | Sustained L3 record + explicit human grant; automatic demotion on breach. | None yet — this is the 2028–30 frontier |
The load-bearing idea is the gate, not the ladder. Autonomy is earned by measured calibration, bounded by explicit policy, and proven by attestation — never granted on vibes, demos, or market cap. An ecosystem that enforces the gates gets the Constitution ending’s economics with the Capture ending’s speed. An ecosystem that skips them gets the speed first, and then the incident. SmartDAO is building this stack — the context layer, the calibrated forecasting engines, and the policy-bounded agent runtime — as an operating system any DAO can adopt; that work is documented separately, and this page stands or falls as research either way.
The forecast registry: ten claims we can be wrong about
Scenario narratives are cheap; scored predictions are not. The table below registers this page’s load-bearing claims with explicit probabilities and resolution criteria. We will mark each one resolved — correct or incorrect — publicly, on this page, and our calibration on these claims is itself a product we should be judged by.
HOW EVERY FAN CHART ON THIS PAGE IS MADE — THE CLOUD IS THE HONESTY, THE BAND IS THE CLAIM. THE SIMULATION RESEEDS ON ITS OWN.
| # | Claim | By | P | Resolution source |
|---|---|---|---|---|
| 1 | ≥1 top-20 DAO (by treasury) has a registered AI agent delegate holding ≥0.5% of active voting power, with a published voting policy | end-2027 | 75% | on-chain delegation records (Tally/Dune) |
| 2 | ≥25% of governance proposals at top-50 DAOs ship with machine-generated analysis attached at posting time | end-2027 | 80% | sample of forum/Tally posts, quarterly audit |
| 3 | A major DAO adopts a formal graduated-autonomy framework (typed agent actions, policy bounds, veto windows) by governance vote | end-2028 | 70% | the ratifying proposal itself |
| 4 | First inter-DAO transaction ≥$1M negotiated agent-to-agent and ratified by both DAOs’ governance | end-2028 | 55% | both ratification votes + the settlement tx |
| 5 | LLM forecasting systems reach statistical parity with professional human forecasters on a Metaculus-style benchmark | end-2028 | 60% | Metaculus AI benchmark series |
| 6 | A governance attack executed primarily by autonomous agents causes ≥$10M in losses at one or more DAOs | end-2029 | 50% | public post-mortem attributing agent automation |
| 7 | ≥10% of top-100 DAO treasury value sits under continuous policy-bounded management (agent-executed, human veto) | end-2029 | 60% | on-chain policy-module TVL measurement |
| 8 | Aggregate DAO treasuries exceed $50B (median path $58B; 80% CI $25–180B) | end-2030 | 55% | DeepDAO total, year-end snapshot |
| 9 | Median idea-to-execution latency for routine (sub-cap) decisions at autonomy-adopting DAOs falls below 48 hours | end-2030 | 65% | proposal lifecycle timestamps, published dataset |
| 10 | ≥1 DAO where agents cast the majority of vote weight operates for 12+ months without a ≥5%-of-treasury loss event | end-2030 | 35% | vote-weight attribution + treasury history |
Methodology and limits
What we did. The baseline chapter is a literature and data review: 68 sources, each tagged primary, academic, or secondary, with the as-of date of the data stated next to every number (treasury figures in particular are price-volatile snapshots; the famous “$25B” is March 2023 data and is labeled as such[2]). The forward chapters are scenario construction in the AI 2027 format[1]: trend extrapolation on the cited series, reference-class reasoning from adjacent adoptions (account abstraction, prediction markets, delegate professionalisation), and explicit subjective probabilities where extrapolation runs out. The fork follows that format’s honesty device: the authors do not know which branch obtains, and the branches are written to be distinguishable by 2029 evidence.
What we did not do. We did not fit curves to thin data and present the result as physics — the most legitimate criticism levelled at our format’s namesake. Where our confidence intervals are wide (they are), that is the finding. We assume no capability discontinuity in frontier models: this scenario is built on agents roughly as good as the 2026 evidence supports[61][62], improving incrementally. If superintelligent systems arrive mid-scenario, the bottleneck analysis still holds but every date compresses. We also note the deepest known unknown: several load-bearing numbers (treasury totals, turnout) come from aggregators whose methodologies are not fully public; we cite the strongest available source and flag date sensitivity throughout.
Conflict of interest, declared. SmartDAO builds software whose value increases if the Constitution ending obtains. That is precisely why the registry above exists: our incentive to be believed is disciplined by our commitment to be scored.
Corrections are welcome and will be published with credit in the errata log below; the forecast registry is reviewed and re-scored every quarter.
- 2026-06-12 · The prediction-market volume figure originally read “Polymarket: $20B+ per month.” Pew’s figure is Kalshi and Polymarket combined (≈$24B in April 2026); Polymarket alone first crossed $10B/month in March 2026. Corrected above. Caught by our own DAO 2040 anchor re-verification.
- 2026-06-12 · “All-time high of $25.1B” softened to “then-record” — total DAO treasuries re-crossed $25B in March 2026, so the superlative is no longer safe. Same source, better framing.
References
Tiering: primary — official posts, on-chain data, first-party dashboards, regulator releases; academic — peer-reviewed or formally posted working papers; secondary — reputable press reporting a primary source. Where a secondary source relays a primary figure, the primary is named in the entry.
- [1]AI 2027. Kokotajlo, Alexander, Larsen, Lifland, Dean — AI Futures Project · Apr 2025 · primary — the scenario-forecasting format this page adapts: chronological narrative, explicit probabilities, falsifiable claims, public errata
https://ai-2027.com/ - [2]DAO treasuries top $25 billion for the first time: DeepDAO. Cointelegraph, reporting DeepDAO data · Mar 2023 · secondary — the March 2023 record print; frequently misquoted as a current figure — 2026 totals have since re-approached this level
https://cointelegraph.com/news/dao-treasuries-top-25-billion-for-the-first-time-deepdao - [3]DAO treasuries surged 40x in 2021: DeepDAO. Cointelegraph, reporting DeepDAO data · Jan 2022 · secondary — ~$400M (Jan 2021) → ~$16B (Sep 2021)
https://cointelegraph.com/news/dao-treasuries-surged-40x-in-2021-deepdao - [4]DAO Treasury Holdings Statistics 2025. CoinLaw, aggregating DeepDAO · 2025 · secondary — $21.4B liquid / $24.5B total; 13,000+ DAOs; top-5 DAOs hold 62.3% of all treasury assets
https://coinlaw.io/dao-treasury-holdings-statistics/ - [5]Future of Algorithmic Organization: Large-Scale Analysis of DAOs (arXiv:2410.13095). Sharma et al., arXiv · 2024 · academic — $24.5B total treasury, 11.1M governance token holders, 13,000+ DAOs
https://arxiv.org/abs/2410.13095 - [6]DeepDAO — Organizations dashboard. DeepDAO · live · primary — canonical live source for ecosystem totals; check current values here
https://deepdao.io/organizations - [7]Dissecting the DAO: Web3 Ownership is Surprisingly Concentrated. Chainalysis, State of Web3 · Jun 2022 · primary — across 10 major governance tokens, <1% of holders control 90% of voting power; 85% of treasuries concentrated in a single asset
https://www.chainalysis.com/blog/web3-daos-2022/ - [8]Centralized Governance in Decentralized Organizations (SSRN 5168660). Cong, Rabetti, Wang, Yan · 2025 · academic — top decile of voters controls 76.2% of votes; ~6.3% average participation vs ~70% at corporations; insider abnormal returns of 9.5% pre-proposal
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5168660 - [9]Fairness in Token Delegation: Mitigating Voting Power Concentration in DAOs (arXiv:2510.05830). arXiv · Oct 2025 · academic — voting-power Gini coefficients >0.94 across studied DAOs; ≈0.99 for Compound, ENS, Uniswap
https://arxiv.org/abs/2510.05830 - [10]Is DAO Governance Fostering Democracy?. HICSS / University of Hawai'i ScholarSpace · 2023 · academic — in 126 Snapshot proposals, a median of 3 wallets sufficed to reach 50% of votes cast
https://scholarspace.manoa.hawaii.edu/bitstreams/331f3027-5077-4bee-b651-f560b2e158ba/download - [11]Evaluating DAO Sustainability and Longevity Through On-Chain Governance Metrics (arXiv:2504.11341). arXiv · Apr 2025 · academic — active-to-eligible voter ratio falling to 1–2% in recent votes
https://arxiv.org/pdf/2504.11341 - [12]Arbitrum has 986 delegates — only 91 actually show up. MconnectDAO (Tally/Dune data analysis) · Apr 2026 · secondary
https://medium.com/@MconnectDAO/arbitrum-has-986-delegates-only-91-actually-show-up-d4643e814328 - [13]Monitoring Limits in DAO Governance: Capacity Breakpoints and Endogenous Concentration (arXiv:2603.11222). Tchuente, Purdue University · Mar 2026 · academic — statistically significant decline in voter responsiveness once proposal flow crosses a threshold — DAOs become 'too big to monitor' and control concentrates
https://arxiv.org/abs/2603.11222 - [14]Exploring DAO Governance: An Empirical Study on Snapshot. DAOBase · mid-2023 · secondary — ~130,000 proposals, ~20M votes, 2.6M voters across ~20,000 spaces since 2020
https://medium.com/@daobase_ai/exploring-dao-governance-a-empirical-study-on-snapshot-2a1084e54471 - [15]The DAO Incentive Program (DIP 2.0). Arbitrum DAO governance forum · 2025 · primary — ~7,000,000 ARB absorbed; per-delegate caps of 300–700 ARB per vote
https://forum.arbitrum.foundation/t/the-dao-incentive-program-dip-2-0/30080 - [16]Arbitrum DAO votes on $1.5 million program to reward active delegates. The Defiant · 2024 · secondary
https://thedefiant.io/news/blockchains/arbitrum-dao-votes-on-usd1-5-million-program-to-reward-active-delegates - [17]Uniswap Delegate Reward Initiative — Cycle 1 temp check. Uniswap governance forum · Jun 2024 · primary — $216,000 for 3 months — 12 delegates at $6,000/month; Cycle 3 expanded to $540,000 over 6 months (The Block, Mar 2025)
https://gov.uniswap.org/t/temp-check-uniswap-delegate-reward-3-months-cycle-1/23837 - [18]DAO Delegate Incentive Programs: How to Build for the Future. DAOstar (supported by Scroll) · Feb 2025 · primary — benchmark of delegate-pay programs across 9 major DAOs; 7 of 9 funded from treasuries
https://daostar.org/research/delegate.pdf - [19]Risk manager Gauntlet terminates relationship with Aave, citing DAO dysfunction. CoinDesk · Feb 2024 · secondary — $1.6M/yr contract; quit ~2 months after renewal over 'inconsistent guidelines and unwritten objectives'
https://www.coindesk.com/tech/2024/02/22/risk-manager-gauntlet-terminates-relationship-with-aave-citing-dao-dysfunction - [20]ARDC (Arbitrum Research & Development Collective) — final report and funding threads. Arbitrum DAO governance forum · 2024–2025 · primary — V1 funded with 1,761,000 ARB for 6 months (~$3.06M at funding); ARB price drop later halved its purchasing power
https://forum.arbitrum.foundation/t/ardc-final-report/27050 - [21]Arbitrum Foundation 2027 budget proposal — $43.5M. CryptoRank / The Defiant coverage of on-chain vote · Jun 2026 · secondary
https://cryptorank.io/news/feed/e3c22-arbitrum-foundation-2027-budget-proposal-43-5-million - [22]Uniswap governance process. Uniswap docs · current · primary — ≥7d RFC + 5d temp check + 2d delay + ~7d vote + 2d timelock ≈ 23 days minimum
https://docs.uniswap.org/concepts/governance/process - [23]Lifecycle and anatomy of an AIP. Arbitrum Foundation docs · current · primary — constitutional proposals: roughly five weeks from forum to execution
https://docs.arbitrum.foundation/concepts/lifecycle-anatomy-aip-proposal - [24]Leading DeFi protocol Compound leaked over $100 million in rewards. Blockworks · Oct 2021 · secondary — no admin controls — every fix required the full 7-day governance cycle while funds drained
https://blockworks.com/news/leading-defi-protocol-compound-leaked-over-100-million-in-rewards - [25]Attacker drains $182M from Beanstalk stablecoin protocol. CoinDesk · Apr 2022 · secondary — a >$1B flash loan bought 67% of the vote inside one transaction
https://www.coindesk.com/tech/2022/04/17/attacker-drains-182m-from-beanstalk-stablecoin-protocol - [26]SEC charges Avraham Eisenberg with manipulating Mango Markets ($116M). U.S. Securities and Exchange Commission, press release 2023-13 · Jan 2023 · primary — the exploiter voted his stolen tokens on his own $47M 'settlement'
https://www.sec.gov/newsroom/press-releases/2023-13 - [27]Explained: the Tornado Cash governance attack. Halborn · May 2023 · secondary — attacker minted 1.2M fake votes via a hidden self-destruct/redeploy
https://www.halborn.com/blog/post/explained-the-tornado-cash-hack-may-2023 - [28]$24 million Compound proposal passed by whale over DAO objections (Proposal 289). The Block · Jul 2024 · secondary
https://www.theblock.co/post/307943/24-million-compound-finance-proposal-passed-by-whale-over-dao-objections - [29]Nouns fork: disgruntled holders exit with $27 million from treasury. Decrypt · Sep 2023 · secondary
https://decrypt.co/197400/nouns-fork-disgruntled-nft-holders-exit-27-million-from-treasury - [30]Black Thursday for MakerDAO: $8.32 million was liquidated for 0 DAI. whiterabbit, independent post-mortem · Mar 2020 · secondary
https://medium.com/@whiterabbit_hq/black-thursday-for-makerdao-8-32-million-was-liquidated-for-0-dai-36b83cac56b6 - [31]Aave community rejects Gauntlet's proposal to freeze CRV. The Block · Jun 2023 · secondary — paid risk advice overridden by token vote while a $176M position loomed
https://www.theblock.co/post/235422/aave-community-rejects-gauntlets-proposal-to-freeze-crv-tokens-on-v2 - [32]Scroll DAO to pause governance structure amid leadership shake-up. CoinDesk · Sep 2025 · secondary
https://www.coindesk.com/markets/2025/09/11/scroll-dao-to-pause-governance-structure-amid-leadership-shake-up-redesign-plans - [33]Arbitrum DAO treasury balance (Entropy Advisors Dune dashboard). Dune Analytics · 2025 · primary — >90% of the ~$1.78B treasury held in ARB
https://dune.com/entropy_advisors/arbitrum-dao-treasury - [34]Uniswap DAO prepares to expand $5bn treasury beyond UNI. DL News · 2024–2025 · secondary
https://www.dlnews.com/articles/defi/uniswap-dao-begins-vote-to-diversify-treasury/ - [35]Arbitrum DAO approves 35 million ARB allocation to tokenized US Treasurys (STEP). The Block · 2024–2025 · secondary
https://www.theblock.co/post/353631/arbitrum-dao-us-treasurys - [36]ConstitutionDAO raises $47M from 17,437 donors, loses Sotheby's bid. CNBC · Nov 2021 · secondary
https://www.cnbc.com/2021/11/18/constitutiondao-crypto-investors-lose-bid-to-buy-constitution-copy.html - [37]UkraineDAO raises millions in ETH for war relief. Decrypt · Mar 2022 · secondary — ~$6.75–7.1M raised in 72 hours from 3,271 contributors
https://decrypt.co/94386/ukraine-dao-millions-in-ethereum-shows-what-dao-can-do - [38]Reflecting on 2024: a year of growth, innovation, and community. Gitcoin (official blog) · 2024 · primary — $10.4M distributed across 105 rounds in 2024; 141.5K donors, 1,743 grantees
https://www.gitcoin.co/blog/reflecting-on-2024-a-year-of-growth-innovation-and-community - [39]Optimism Retroactive Public Goods Funding — round results. Optimism Collective (official posts and governance forum) · 2021–2025 · primary — 60M+ OP distributed across Rounds 2–6; 850M OP (20% of supply) reserved
https://gov.optimism.io/t/retro-funding-5-op-stack-round-details/8612 - [40]VitaDAO closes $4.1 million fundraising round (incl. Pfizer Ventures). Lifespan.io · Jan 2023 · secondary — decentralised longevity-research funding; 24 projects / $4.2M deployed by 2024
https://www.lifespan.io/news/vitadao-closes-4-1-million-fundraising-round/ - [41]Nouns DAO — one NFT auctioned every 24 hours since August 2021. nouns.wtf (on-chain, verifiable) · live · primary
https://nouns.wtf/ - [42]KlimaDAO accumulates $100M of tokenized carbon offsets. Cointelegraph · Nov 2021 · secondary
https://cointelegraph.com/news/klima-dao-accumulates-100m-of-carbon-offsets-aims-to-drive-up-price - [43]ERC-4337 adoption: 100M+ UserOperations, 40M+ smart accounts. Cobo; live data at BundleBear (0xKofi) · early 2025 · primary — BundleBear is the canonical open dashboard for account-abstraction stats
https://www.bundlebear.com/overview/all/month - [44]Smart wallet adoption surges after Ethereum's Pectra upgrade (EIP-7702). The Block · May 2025 · secondary — 11,000+ authorizations within the first week
https://www.theblock.co/post/354414/smart-wallet-adoption-surges-after-pectra-upgrade - [45]x402: an open protocol for internet-native payments. Coinbase (official launch) · May 2025 · primary
https://www.coinbase.com/developer-platform/discover/launches/x402 - [46]x402 agentic payments cross 100M transactions on Base. Chainalysis · Q1 2026 · primary
https://www.chainalysis.com/blog/x402-agentic-payments-adoption/ - [47]Coinbase-backed AI payments protocol wants to fix micropayments — but demand is just not there yet. CoinDesk · Mar 2026 · secondary — daily transactions down >92% from the Dec 2025 peak; ~$28K/day real volume — the honest counterweight
https://www.coindesk.com/markets/2026/03/11/coinbase-backed-ai-payments-protocol-wants-to-fix-micropayment-but-demand-is-just-not-there-yet - [48]Agents that transact: Amazon Bedrock AgentCore Payments, built with Coinbase and Stripe. Amazon Web Services (official blog) · May 2026 · primary
https://aws.amazon.com/blogs/machine-learning/agents-that-transact-introducing-amazon-bedrock-agentcore-payments-built-with-coinbase-and-stripe/ - [49]RFC: Ratifying the Agentic Governance Pivot (7,000,000 ARB voter-enfranchisement pool). Arbitrum DAO governance forum · May 2025 · primary — a top-5 DAO formally voting on AI-agent voting power — the single clearest signpost already passed
https://forum.arbitrum.foundation/t/rfc-proposal-to-adjust-the-voting-power-of-the-arbitrum-community-pool-ratifying-the-agentic-governance-pivot/29280 - [50]Introducing Governatooorr: the autonomous AI-powered delegate. Valory / Olas (official) · 2023 · primary
https://github.com/valory-xyz/governatooorr - [51]Olas raises $13.8M for Pearl, an app store for autonomous AI agents in crypto. The Block · 2025 · secondary — 15.6M+ cumulative agent transactions; 834 daily active agents by Q1 2026
https://www.theblock.co/post/338713/olas-raises-13-8-million-to-launch-pearl-an-app-store-for-autonomous-ai-agents-in-crypto - [52]Tally MCP server — DAO governance data for AI agents. Dennison Bertram (Tally CEO), GitHub · 2025 · primary
https://github.com/crazyrabbitLTC/mpc-tally-api-server - [53]Agora acquires Boardroom to combine governance infrastructure with AI decision support. Agora (official) · Jan 2025 · primary
https://www.agora.xyz/blogs/6-boardroom-acquisition - [54]DAO-AI: evaluating agentic AI as a governance decision-maker on 3,383 real proposals (arXiv:2510.21117). arXiv · Oct 2025 · academic
https://arxiv.org/html/2510.21117v2 - [55]Understanding futarchy on Solana: MetaDAO's live decision markets. Blockworks · 2024–2025 · secondary
https://blockworks.com/news/understanding-futarchy-on-solana - [56]Futarchy v1 — preliminary findings (500K OP conditional-funding experiment). Optimism governance forum · Mar–Jun 2025 · primary — also surfaced metric-gaming: TVL of selected projects fell $15.8M — the chosen metric was flawed
https://gov.optimism.io/t/futarchy-v1-preliminary-findings/10062 - [57]Futarchy meets governance: Optimism and Uniswap Foundation pilot conditional funding markets. Uniswap Foundation (official) · 2025 · primary — $900K of Unichain lending-protocol grants later allocated via CFMs
https://www.uniswapfoundation.org/blog/futarchy-meets-governance-optimism-and-uniswap-foundation-pilot-cfms - [58]ICE announces strategic $2B investment in Polymarket at a $9B valuation. Intercontinental Exchange (official press release) · Oct 2025 · primary
https://ir.theice.com/press/news-details/2025/ICE-Announces-Strategic-Investment-in-Polymarket/default.aspx - [59]Trading volume on prediction markets has soared in recent months. Pew Research Center · May 2026 · primary — combined Kalshi + Polymarket ≈ $24B in April 2026; Polymarket alone first crossed $10B/month in March 2026 (corrected 2026-06-12 — see errata)
https://www.pewresearch.org/short-reads/2026/05/27/trading-volume-on-prediction-markets-has-soared-in-recent-months/ - [60]Prediction market accuracy: crowd wisdom or informed minority? (full Polymarket history, 210,322 markets). Gomez-Cram, Guo, Kung (LBS), Jensen (Yale) · Apr 2026 · academic — 3.14% of 1.72M accounts drive market accuracy
https://igaming.org/sports-news/study-links-polymarket-accuracy-to-skilled-minority/ - [61]Approaching human-level forecasting with language models (arXiv:2402.18563, NeurIPS 2024). Halawi, Zhang, Yueh-Han, Steinhardt — UC Berkeley · 2024 · academic — retrieval-augmented LLM system: Brier 0.179 vs human crowd 0.149 (random = 0.25); near-parity when the crowd is uncertain
https://arxiv.org/abs/2402.18563 - [62]Metaculus AI Forecasting Benchmark, Q2 2025 results. Metaculus · Q2 2025 · primary — pro forecasters still clearly beat the best bot teams (p = 0.00001) — machine forecasting is approaching, not at, parity
https://www.metaculus.com/aib/2025/q2/ - [63]AI as the engine, humans as the steering wheel. Vitalik Buterin (vitalik.eth.limo) · Feb 2025 · primary
https://vitalik.eth.limo/general/2025/02/28/aihumans.html - [64]From prediction markets to info finance. Vitalik Buterin (vitalik.eth.limo) · Nov 2024 · primary — 'millions of mini-markets' for low-consequence decisions, turbocharged by AI
https://vitalik.eth.limo/general/2024/11/09/infofinance.html - [65]The promise and challenges of crypto + AI applications. Vitalik Buterin (vitalik.eth.limo) · Jan 2024 · primary — warns specifically against naive 'AI judges' in DAOs — adversarial exploitation of open models
https://vitalik.eth.limo/general/2024/01/30/cryptoai.html - [66]Gartner predicts over 40% of agentic AI projects will be canceled by end of 2027. Gartner (official press release) · Jun 2025 · primary — yet also: ≥15% of day-to-day work decisions made autonomously by agents by 2028, from ~0% in 2024
https://www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts-over-40-percent-of-agentic-ai-projects-will-be-canceled-by-end-of-2027 - [67]Big ideas in crypto, 2026: from KYC to Know-Your-Agent. a16z crypto (official) · Dec 2025 · primary
https://a16zcrypto.com/posts/article/big-ideas-things-excited-about-crypto-2026/ - [68]Virtuals Protocol launches first revenue network for agent-to-agent AI commerce. Virtuals Protocol (PRNewswire) · Feb 2026 · primary — Agent Commerce Protocol: on-chain request → negotiate → escrow → settle; 18,000+ agents deployed
https://www.prnewswire.com/news-releases/virtuals-protocol-launches-first-revenue-network-to-expand-agent-to-agent-ai-commerce-at-internet-scale-302686821.html
SmartDAO is building the operating system this scenario describes — the context layer, calibrated forecasting, and policy-bounded agency for human–AI governance.
Explore SmartDAO →© 2026 ARC Intelligence · DAO 2030 v1.1 · errata log opened 2026-06-12 (2 corrections)