Chapter 9: Limitations and Open Problems

Draft v0.1 - 2026-05-25 - awaiting Zaal review

Fractal governance works in production. ZAO Fractal’s 100+ weeks prove it. But the model breaks under specific conditions, fails in visible ways, and faces open problems that no fractal has solved yet. This chapter names those limits openly, not to undermine the system, but to earn its credibility.

Participation Durability and Democracy Fatigue

Weekly synchronous governance is demanding. Research on real-world participation shows a sharp drop-off: after 2-4 months of weekly meetings, attendance stabilizes around a core 40-50% who carry the consensus burden. The rest attend sporadically. The psychological research is clear: webinar fatigue, decision fatigue, and the simple human cost of predictable weekly commitment cause participation collapse.

ZAO Fractal has sustained weekly meetings for 100+ weeks without an announced pause. By the evidence, this should not be possible. The fact that it is possible suggests either that:

1. Music community members are unusually committed to governance, or
2. The ritual is psychologically embedded in the weekly culture in a way that prevents burnout, or
3. We have not yet measured the actual fatigue, and it is accumulating beneath the surface.

Mitigations ZAO uses: We accept that not every member can attend every session. A core group of 4+ unplayed (absent) members triggers an automatic session (no rescheduling required). This lowers the barrier to skip a week. We also rotate facilitators so the burden does not fall on one person. But these are partial fixes. The risk is real: if the weekly ritual becomes a chore instead of a gathering, participation collapses into delegated voting, and the deliberation that makes fractals different evaporates.

Insider Bias and Visibility Bias

Ranking contribution is subjective. Every study on open-source work, organizational bias, and implicit ranking shows the same pattern: visible work (presenting, facilitating, loud talking) is ranked higher than quiet work (infrastructure, documentation, mentorship behind closed doors). Two-thirds of open-source labor is invisible. When invisible labor is ranked, it gets lower compensation, lower status, lower Respect.

ZAO Fractal is vulnerable to this. The Discord bot facilitators, the Discord channel mods, the people who speak in breakout rooms - they accumulate visibility. A developer who fixed a critical security bug in silence may earn less Respect than a community manager who posted memes weekly. Over 100+ weeks, this bias compounds. Early, visible, charismatic members accumulate Respect faster than late-arriving, quiet builders. The governance then drifts toward the preferences of high-Respect members, amplifying the original bias. ZAO Fractal becomes a soft plutocracy, ruled by visibility instead of contribution.

Evidence of this: As of May 2026, only two wallets have ever submitted to the OREC contract (Zaal and civilmonkey.eth). The operating core is concentrated. This is intentional - we designed it this way during growth phase - but it creates a visibility bottleneck. If OREC submission is the mechanism for on-chain governance, only people visible enough to be trusted with OREC signing power can propose on-chain changes. Newer members, quiet contributors, people in other timezones are locked out.

Mitigation: Transparent contribution rubric. Use an explicit checklist for ranking: Did this person ship code? Did they mentor someone? Did they document? Did they curate? Instead of “did they impress me,” use structured judgment. This does not eliminate bias - it just makes it visible. The next mitigation is off-chain ranking on GitHub (Chapter 10 roadmap: Frapp-GH), which makes work verifiable and attributable without visibility bias.

Sybil and Collusion Attacks

Small groups can be gamed. A coordinated ring of 10 people in a 40-person fractal can dominate the Respect flows to the next layer if they rank each other highly in every session. The academic literature on Sybil attacks, collusion in consensus systems, and voting manipulation is clear: a minority that coordinates can capture a system designed for honest majority.

ZAO Fractal has not experienced a documented attack of this kind. But we have not tried hard to prevent it either. The current mitigations are social: we know each other, we would notice if someone was gaming the system, we would call it out. This works at 40 active members. It breaks at 400. And it breaks instantly if someone joins with the intent to subvert.

This is not a theoretical risk. Other DAOs have experienced this. Synthetify DAO on Solana lost $230,000 when an attacker created a governance proposal and voted it through an inattentive circle. The attack succeeded because the governance group was small and the attacker had capital resources.

Mitigation: Periodic account verification (linked to Farcaster FID or Ethereum wallet to prove single-identity-per-account). Transaction cost for circle membership (a small gas fee to participate, preventing low-cost Sybil nodes). Off-chain social proof (references from existing members, reducing entry cost for good-faith newcomers but blocking automated accounts). None of these are perfect. They just make attacks more expensive and visible. ZAO’s roadmap (Chapter 10) includes formal identity-binding for Frapp-GH.

Cold Start: New Members at Zero Respect

A new member joins ZAO Fractal and begins at zero Respect. Existing members have weeks or months of accumulated Respect. The barrier to influence is enormous. This is by design - we do not want to hand out power to newcomers - but the consequence is that newcomers have no voice until they have proven themselves in multiple sessions.

The DAOstar research (2025) on blockchain governance shows that cold-start inequality is a primary barrier to DAO growth. When newcomers feel unheard, they do not return. The insider advantage is permanent unless the system has explicit on-boarding.

ZAO’s mitigation: one-time grants of baseline Respect to new joiners (an amount the council sets). This narrows the gap. But it also dilutes the signal - if everyone starts with 30 Respect, does that Respect still mean contribution? The problem is not solved. It is traded off.

Earned Standing Is Permanent; Active Voice Is Not

Respect accumulates and never leaves a wallet: once minted, a balance changes only when OREC mints again. The structural worry is real: governance power can settle toward whoever arrived first and never left, regardless of recent contribution.

Season 3 answers it on the vote rather than the balance. Voting is gated on a rolling 90-day activation: a member who signed or attended within the last 90 days votes at full weight, a member who did not keeps every point and regains the vote by taking part again (ZIP-2 section 3 in bettercallzaal/zao-papers; outside precedent, from co-operative bylaws to Optimism’s Citizens’ House, in ZAOOS research governance/2558-dao-periodic-reactivation-precedent). The trade-off is turnout: gating the vote on a recurring signal concentrates the electorate in those who show up, and quorum must therefore be defined as a share of the active pool, never as a fixed number of members.

Two rulings that reshape other chapters are recorded here rather than silently applied, because this draft predates them: OG Respect resumes as the one-time achievements ledger and Respect splits by project (ZAO Fractal Respect for governance, ZAO Festivals Respect, WaveWarZ Respect later), both 2026-09-26 (zao-vault decisions/grill-2026-09-26-zao-papers-afternoon.md, rulings 1 and 2); and vote weight becomes OG plus ZOR summed, 2026-10-10 (Chapter 6, Section III). Chapter 10 carries the dates.

Scaling Past Dozens is Unproven

Roy Fractal at 700+ members on EOS proves that the mathematical structure of nested fractals scales. But Roy operates on EOS, a separate ecosystem with different visibility and economics. No Ethereum-side fractal has sustained 400+ members at consistent weekly cadence.

ZAO is at 188 members, 40 active per session. What happens at 200? 400? 1000?

Dunbar’s number (the cognitive limit on human relationships) is ~150. Beyond that, you cannot maintain trust through direct relationships. You need hierarchy, explicit roles, and representatives you have never met. Fractal governance is designed to scale through nested circles, but nesting adds layers of representative risk. At Layer 1 (40-person circle), your voice is 1/40. At Layer 2 (6 reps per circle), you are represented by someone. At Layer 3, you are represented by someone who is represented by someone. The voice attenuates.

ZAO’s roadmap (Chapter 10) includes a decision point: if ZAO grows past 100 active members, do we split into parallel fractals (two breakout-room sets, same Monday time) or do we stay single and accept representative layers? Both have tradeoffs. The problem is unresolved.

Two-Ledger Reconciliation is Incomplete

ZAO maintains two Respect ledgers (OG ERC-20 for Fractals 1-73, ZOR ERC-1155 for Fractals 74+). They are reconciled in Supabase. But on-chain, they are separate contracts. A member who earned 300 Respect in the OG era holds that in one contract. Respect earned in the ZOR era lives in another.

For ORDAO voting, this creates a real gap: does on-chain voting power include the active ZOR ledger? The current answer is no - OREC reads OG balances only, at the moment you vote. New ZOR mints do not change anyone’s voting weight. This means the disadvantage runs toward newer members: a member who joined after the OG freeze (18 December 2025) and holds only ZOR has no on-chain voting weight at all, however much ZOR they have since earned. Early members who hold OG vote at full weight; the active reward ledger confers standing and record, but not yet a vote.

This is not a critical bug - it is the deliberate decoupling described in Chapter 6, seen from its uncomfortable side. But left unreconciled it could poison trust: the members most active today have the least on-chain say. The Chapter 10 roadmap includes giving the ZOR ledger a path to voting weight - publishing an OG-to-ZOR weighting formula and a retroactive claim mechanism - without discarding the OG history the two-ledger model was built to preserve.

Infrastructure Single-Points-of-Failure

ornode (an indexing service for Respect data) is currently DOWN as of May 2026. The on-chain leaderboard at zao-fractal.vercel.app was deleted - there is currently no public leaderboard outside the Discord bot and ZAO OS.

This means that the “immutable on-chain history” of ZAO Fractal is theoretically immutable but practically unreachable for most members. If a member wants to verify their Respect balance without trusting the Discord bot, they would need to call the OREC contract directly on Etherscan or write custom queries. This is not accessible to non-technical members.

Mitigation: Restore the web dashboard at zaoos.com (Chapter 10, June 15 2026 target). This is not complex. It is just infrastructure work. But it is critical - the web dashboard is the proof that governance is real and transparent. Without it, ZAO Fractal feels like a Discord bot that claims to be on-chain, not a blockchain-native system.

Documentation Gap

Tanja’s call on May 18, 2026 identified the #1 onboarding blocker: non-technical members cannot explain ZAO Fractal to peers. There is no single canonical document explaining how it works, why it is different, what the weekly ritual is, how to participate, what Respect means. Members have to ask in Discord. New joiners get lost.

This whitepaper is the start of the answer. But the whitepaper is not an onboarding document. It is a governance paper. ZAO also needs: video tutorials (how to submit a breakout ranking), explainer graphics (Respect earning curves, voting criteria), a FAQ, a constitution (the rules that ZAO operates under, written down).

Mitigation: Publish this whitepaper, then commission the documentation set (Chapter 10, June 30 2026 target). Make ZAO Fractal reproducible. If another music community wants to fork ZAO’s governance and adapt it, they should be able to. Documentation is how that happens.

Operating Core Concentration

As of May 26, 2026, zaal.eth and civilmonkey.eth are the only two wallets that have ever submitted to the OREC contract. This is intentional. During ZAO Fractal’s growth phase (Fractals 1-90+), we centralized submission authority for simplicity and coordination. One or two people could confirm that results were accurate, submit them on-chain, and maintain the contract.

But centralization is intentional only if it is temporary. The longer two wallets remain the sole OREC signers, the more ZAO Fractal looks like a governance theater - voting happens, consensus is reached, but a single gatekeeper executes on-chain. This is not plutocracy (Respect is earned, not capital-weighted). But it is a bottleneck.

Mitigation: Establish a 3+ signer committee for OREC (Chapter 10, June 30 2026 target). Require multi-sig approval for on-chain submissions. This distributes execution authority and makes ZAO Fractal more resilient to single-person failure (what if zaal disappears?).

The Measurement Problem: Can Peers Judge Contribution?

Beneath all of this is a harder question: Can five strangers in a breakout room actually judge whether someone’s work advanced the community’s mission?

The Respect Game assumes yes. We gather in small groups, discuss the week, and reach consensus on a rank order. The consensus is honest. The ranking reflects actual contribution, not politics.

The research on deliberation vs. voting suggests this is true at human scale. Citizens’ assemblies (Ireland, France, British Columbia) use citizen juries of 100-150 people to deliberate on policy. When given time to learn and discuss, ordinary people make judgments that policy experts respect. The Respect Game is a deliberative process. It should work.

But the Respect Game has not been formally studied. No peer-reviewed research exists on the accuracy of peer-consensus ranking as a measure of open-source or creative contribution. This is not because the Respect Game is new - Daniel Larimer invented it in 2020. It is because academic research on governance lags practice. The Respect Game works in Eden and ZAO empirically. But we do not have rigorous evidence of what it actually measures.

This matters because if peers are bad at judging contribution, the entire system is built on a false premise. Mitigation: Formal research. Partner with an academic team (MIT, Stanford, Berkley) to study ZAO Fractal’s rankings for 1-2 seasons. Verify that high-Respect members are actually high-contribution members (by some external measure: code commits, user feedback, outcomes). If the correlation holds, we have empirical proof. If not, we have evidence we need to redesign.

The Franchise Closes Behind You

Measured 2026-09-02; see research/08-zao-fractal-measured-state.md sections 8 and 10.

OREC reads OG for vote weight. OG froze in December 2025. ZOR, the token every weekly award now issues, confers no governance weight at all. The consequence is arithmetic rather than behavioural: no new voter can be created under the current configuration.

Of the 161 members in respect_members holding a wallet, 94 hold OG only, 21 hold both, 18 hold neither, and 28 hold ZOR only and therefore have no vote. Measured across the whole ZOR era, of the 70 members who have ever received an award, 47 - two-thirds - cannot vote. That share has risen monotonically: 25% at period 67, 52% by 74, 58% by 80, 62% by 100, 67% by period 111. It can only continue rising, because every new participant joins the denominator and never the numerator.

The highest-earning member with no vote holds 738 ZOR, which is more than most voting members hold in OG, against a minWeight of 1000 to pass a proposal.

The dashboard does not surface this. computeRespectWeight sums OG + ZOR and labels the result “weight”; for those 28 members that figure is lifetime earnings, not voting power, and the two render identically.

Nothing here is a malfunction. Every component does what it says. It is a design consequence of freezing the governance ledger while continuing to issue a non-governance reward, and it is the single largest gap between what the Respect Game promises participants and what it delivers.

Respect Earned and Never Issued

Measured 2026-09-02; see research/08-zao-fractal-measured-state.md section 11.

The Airtable Respect table carries one column per meeting. “ZAO Fractal 71” has 6 members scored and “ZAO Fractal 72” has 4. Neither period has any mint on chain. Ten member-awards from November 2025 were played for, scored, and never issued, and no system tracks that they are outstanding.

Period 103 has no mints and no Airtable column, but Airtable’s per-meeting columns stop at fractal 98, so its absence there is not evidence either way. Resolving it needs the Discord history.

Related: in the range where both records overlap, periods 67 to 98, each source holds meetings the other lacks - the chain has period 73 and Airtable does not; Airtable has 71 and 72 and the chain does not. There is no single complete record of the ZAO fractal.

Infrastructure Single-Points-of-Failure: Two Measured Instances

The section above names this class. Two concrete instances, both measured 2026-09-01.

The recorder had no credentials for five months. The deployed bot’s .env contains ALCHEMY_API_KEY, DISCORD_TOKEN, FRACTAL_BOT_WEBHOOK_SECRET and WEB_WEBHOOK_URL, and no Supabase credentials at all. Its only write path was a webhook documented in its own source as “fire-and-forget semantics (10s timeout)”. Nothing has been written to fractal_sessions since 2026-03-23 and nobody noticed for five months, because nothing depended on the recording strongly enough to fail visibly.

One key mints everything. Every one of the last twelve ZOR mints was sent by 0x7234…e9afoptimism, which holds 3094 OG against a minWeight of 1000 - clearing the passing threshold three times over, alone, every week. If that key is lost, no Respect can be minted by anyone, since ZOR’s owner is OREC and OREC requires a passed proposal. ch10’s milestone “Establish 3+ Signer Committee for OREC”, dated 2026-06-30, is what this was for. Zaal was shown the measurement on 2026-09-01 and chose to accept the risk for now; recorded here so the acceptance is visible.

Sources

• 05-critiques-failure-modes.md (democracy fatigue research, visibility bias invisibility study, Sybil attack literature, cold-start inequality DAOstar research, scaling limits Dunbar’s number, dormancy risk, subjectivity in ranking)
• 02-live-communities-deep.md (OF pause burnout hypothesis, Aquadac 4-year retention at 20-30 members, Eden Epoch 2 false start Aug 2024, Roy Fractal 700+ scale but documentation sparse)
• 03-optimism-fractal-full-history.md (OF paused Jan 2026 after 15 months, consolidation logic, developer concentration burnout)
• 07-zao-fractal-distinctness.md (OREC 2-wallet concentration zaal.eth + civilmonkey.eth, documentation gap Tanja call May 18 2026, ornode down, zao-fractal.vercel.app deleted, two-ledger reconciliation Doc 115)

Word count: 1,847