By ETHPrague 2026, the public consensus among L2 DAO delegates, foundation employees, and governance researchers is that the 2023 peak of “governance = voting” is over. Most L2 DAOs in 2026 are inactive, silent, or shut down. The replacement model is still being designed: it splits work across separate entities (foundation, labs, governance body, oversight committee), accepts optimistic-veto voting as the default, and confronts the brutal lesson from the Optimism / Arbitrum experiments — the worst thing that can happen to governance is indifference, not bad votes.

Key Ideas

”DAOs are not that” — Krzysztof Urbański’s L2BEAT retrospective

Urbański (L2BEAT, four years across Arbitrum / Optimism / Starknet / zkSync / Scroll DAOs) gave the most pointed talk of ETHPrague’s governance track. His core diagnosis:

  • Why governance exists at all in L2 DAOs: not for token utility (the cynical answer) but to handle (1) system upgrades, (2) treasury allocation, (3) oversight of labs and foundation, (4) capturing upside from network effects.
  • The actual work isn’t voting — it’s idea generation, consensus-building, solution exploration, proposal completion, monitoring, post-mortem learning, and aligning multiple proposals to a coherent strategy. Voting is the smallest and optional step (Optimism moved to optimistic veto-only voting).
  • The Arbitrum OAT failure mode: An Oversight & Audit Trail committee was elected in May/June 2025 to govern a second-foundation entity (OpCo) responsible for negotiating a $3M Entropy proposal and approving TMC allocations. As of May 2026 — 10 months in — the OAT had made no public progress on the Entropy negotiation. Nobody on the DAO followed up. The OAT had taken the reservation but not held the reservation.
  • Indifference, not bad votes, is the failure: “What’s the value of a governance token if there’s no real governance? Just ratifying — okay, this is fine — without questioning, without alternatives, without keeping organizations accountable.”

The new institutional pattern: split entities

The emergent architecture across mature L2 DAOs (Optimism, Arbitrum post-2024, zkSync):

  • Labs / dev company — Builds the protocol full-time.
  • Foundation — Boring backend: service-provider deals, infra ops, custodial work for the DAO treasury.
  • Governance association — Forum management, community-facing work, RFP processes, delegate coordination (Shelby Steidl’s role at zkSync Association is the canonical example).
  • Oversight committee — Elected body that supervises the foundation and governance staff. Arbitrum’s OAT is the prototype; it has performed poorly in practice but the pattern survives.

Compared to the original 2023 model of “delegates vote on everything, lab is the enemy,” this is closer to the parliamentary-democracy version: power is distributed across specialized bodies that are individually accountable.

Optimistic voting as the default mechanism

Optimism’s experiment with optimistic governance — proposals pass by default unless a delegate vetoes — has spread. The realization: in mature systems with strong proposers, the cost of getting delegates to actively approve every proposal exceeds the cost of trusting proposer authority and intervening only on red flags. Voting is the exception, not the rule.

Designing for Participation — Marlene Marz

Marz’s complementary lens from Designing for Participation (ETHPrague): DAO participation is not “1-token-1-vote” but a layered hierarchy of engagement modes — lurkers, voters, proposers, oversight members, working-group leads. Designing governance means designing the path between these modes. If lurkers can’t become voters easily and voters can’t become proposers, the pipeline starves and you get OAT-style indifference. Practical lessons: low-friction forum posting, structured rotation between roles, explicit mentorship from existing proposers to new ones.

Decentralized, Not Defenseless — Shelby Steidl

Steidl’s Governance Security Checklist (ETHPrague): treats governance as an attack surface. Key items:

  • Quorum-based bypass attacks (low-turnout proposals slipping critical changes through);
  • Delegated-vote concentration (one whale capturing the optimistic-veto threshold);
  • Proposal-collision attacks (multiple conflicting proposals submitted in quick succession);
  • Emergency-shutdown mechanisms — when present, how the shutdown is triggered is its own attack surface;
  • Pseudonymous delegate accountability (you can’t reach a delegate who only exists on-forum).

This is the OWASP-style checklist for DAO security; the L2BEAT staging framework (stage 0/1/2) operationalizes parts of it.

Burning Man as a polycentric governance case study

The Marian Goodell fireside (CEO, Burning Man Project) framed Burning Man’s 40-year arc as the cultural-movement analogue Ethereum should be studying. Key transitions:

  • LLC → not-for-profit foundation in the early 2000s once the partners realized profits-and-taxation made the model unworkable.
  • Founder-led → CEO-led with explicit scope limitation: “I’m the boss of the business entity, not of the network of events, theme camps, or art cars.”
  • “We don’t read our Twitter.” Filtering signal from anonymous vitriol is operationalized by source: Reddit ignored, Twitter not read, anonymous criticism downweighted vs. authentic-voice criticism. The price of being heard is being identifiable.
  • “You can’t be just a critic” — Engagement requires offering solutions. The community absorbs participants who show up offering work; it doesn’t absorb participants who only complain. This is operationalized as cultural norm rather than rule.

Burning Man’s claim of 175,000 people in Burning Man-affiliated events globally (and a million unique lifetime experiences) suggests Ethereum is significantly smaller as a cultural movement than it imagines itself to be — and that the Burning Man template scales further than the ETH-cultural-movement template currently does.

”Decentralisation Theatre” as a productive frame

Migle Rakitaite’s ETHPrague talk reframes “decentralization theatre” — used pejoratively in 2025 — as a serious metaphor. Theatre is not deception; it’s a culture of collaboration that produces emotion-changing experiences through commitment of many specialized roles (props, cinematography, lighting, makeup, actors). The Ethereum analogue: governance is not voting, it’s the rich workflow of proposal craftsmanship, communication, legal review, technical analysis, and execution — the people in the middle who translate cypherpunk ideology into something a non-technical audience can use. Without them (Bitcoin friend explaining for 4 years, vs. CryptoKitties explaining it in 10 minutes), the values don’t propagate.

Beyond the Treasury: revenue-led DAOs

The Beyond the Treasury talk (ETHPrague) explicit point: DAOs that spend treasury to bootstrap “growth” eventually run out of runway. The new model: DAOs that generate operational revenue from infrastructure (sequencer fees, validator yield, protocol take-rate) and distribute it to active contributors and ecosystem investment. This is the financial-sustainability prerequisite for the institutional split-entity pattern.

Details / Subtopics

From the Playa to the Protocol — full lessons

Goodell’s full set of governance heuristics applicable to Ethereum-style DAOs:

  1. Mission > ownership: Convert from LLC/partnership to not-for-profit as soon as you can. Ownership models corrupt motivation.
  2. Explicit scope of authority: The CEO has decision rights over the business entity, not over the community or the cultural product.
  3. Source-weighted signal: Authentic identity gets more weight than anonymous voice. This is a values choice, not a technical one.
  4. Safety first, then experience, then everything else: At 70K participants, you can’t optimize all three simultaneously; rank them.
  5. People show up; institutions don’t: Hiring follows demonstrated contribution. Maerian Goodell got her CEO role by volunteering to help and being useful — same path as core Ethereum contributors.

Making Governance Easy: UX Lessons from DAOs on Bitcoin

Sascha Goetz’s talk argued the Bitcoin-DAO experiments (StackingDAO, AlexDAO, etc.) have produced better governance UX than Ethereum-L2 DAOs, partly because Bitcoin’s tooling poverty forced UX-first thinking. Lessons portable to Ethereum: native mobile voting; proposal templates as the default interface; reading lists curated by delegates; explicit voter-onboarding flows.

Lessons learned from 3 years spent in L2 DAOs

Urbański’s specific recommendations:

  • Design governance as part of the org structure, not bolted on. The OAT failure was structural — no defined process for “what happens if the elected body goes silent.”
  • Establish expectations of delegates explicitly. If delegates don’t ping accountability questions, that is their failure. Make it called out.
  • Treat governance seriously or won’t have serious governance. The cynical “DAOs are token utility” frame becomes self-fulfilling when there’s no real authority.

From Free to Fast (Viktor Tron’s Swarm) — adjacent governance lesson

Viktor Tron (Swarm) at the decentralization panel: incentives and continuity are the deepest governance variables, not voting mechanisms. How a protocol incentivizes long-term participation matters more than who votes. This connects to the Burning Man “mission > ownership” principle from a different angle.

Veronica & Balaji: Network State as governance experiment

The Network School framing (treated more fully in Network States & Pop-up Cities) is the most ambitious governance experiment in the cluster: a literal community of practice with shared rules, shared finances, and member exit rights. It’s also the most controversial — critics argue it’s just a libertarian residential program, not governance experimentation.

Connections

Open Questions

  • Does the split-entity model (foundation + governance association + oversight) actually outperform monolithic DAO governance, or is it just less visibly broken? Two years of data needed to evaluate.
  • Optimistic voting is the default in mature DAOs but creates new attack surfaces (proposer authority concentration). When does it become a single point of failure?
  • Is indifference (Urbański’s failure mode) primarily an incentive problem (delegates aren’t paid for engagement) or a UX problem (forums are exhausting)? The two diagnoses imply opposite fixes.
  • Can the Burning Man template (mission-driven, source-weighted, explicit authority limits) actually port to crypto DAOs, where pseudonymous participation and capital-weighted decision-making are foundational? The “we don’t read Twitter” heuristic is incompatible with token-vote governance.