Authors: @dwddao @sejalrekhan97 @qgolem and Julian Zawistowski
Abstract
Ethereum’s public-goods funding ecosystem has distributed over $150m through mechanisms such as Gitcoin’s quadratic funding, Optimism’s RetroPGF, and Octant’s staking-based allocations. After six years of experimentation, several important problems remain unresolved. This post highlights three of them where we believe a joint research agenda is needed:
-
Deployment Problem: when and how should funds flow?
-
Allocation Problem: whose preferences are reflected, and how are they aggregated?
-
Impact Problem: how are outcomes measured and fed back into future decisions?
Each of these is fundamentally a mechanism design problem: we are choosing rules that map signals and behaviour into allocations under strategic behaviour and noisy information. Ethereum is an unusually good testbed for mechanism design, but actual adoption in live systems is still shallow. The aim here is to outline questions that could guide theory, mechanism design, and empirical work across funding programs, especially as we move toward allocating protocol-level capital at larger scale.
I. The Deployment Problem: when and how should funds flow?
Early Ethereum public goods funding relied on discrete quarterly rounds, Gitcoin’s GR1 (2019) distributed $38,000 to 200 contributors, eventually scaling to $4.4 million in GR15. These mechanisms created predictable problems: donor and project fatigue from “grueling two-week sprints,” capital sitting idle between rounds, builder uncertainty, and concentrated operational overhead.
What We’ve Learned
The ecosystem has evolved from discrete quarterly rounds toward several parallel innovations: streaming mechanisms that eliminate idle capital and provide continuous matching; hybrid epoch models like Protocol Guild’s time-weighted vesting (channeling $30M+ to 190+ contributors) that balance predictability with sustainability; and mechanism plurality acknowledging no single mechanism optimizes across all contexts.
A Selection of Critical Open Problems
Key questions remain unresolved: Can streaming mechanisms achieve robust collusion resistance without prohibitive computational complexity or friction for honest users? What timing models optimize across capital efficiency, donor retention and attack economics? How can we create accountability for funding decisions, through allocation bonds or other mechanisms, without creating plutocratic dynamics? Additionally, as multiple programs (Gitcoin, Optimism, Octant, EF grants) operate in parallel, how should the ecosystem coordinate to build complementary rather than duplicative funding infrastructure?
II. The Allocation Problem: Whose Preferences and How to Aggregate?
Results like Arrow’s impossibility theorem, Gibbard–Satterthwaite, and related social-choice impossibility results prove no voting system can simultaneously satisfy non-dictatorship, Pareto efficiency, independence of irrelevant alternatives, and transitivity. Every allocation mechanism must compromise on legitimacy, efficiency and fairness criteria, the question is which tradeoffs are acceptable for pseudonymous, global communities with misaligned incentives.
What We’ve Learned
Six years of experimentation reveal fundamental tradeoffs with no clear winners: quadratic funding evolved from pure formulas to pairwise-bounded (20-30% dominance reduction) to cluster-matching approaches that embrace rather than prevent coordination; MACI offers theoretical collusion resistance through receipt-free encrypted voting but remains limited to mid-sized settings due to usability challenges; Gitcoin Passport pragmatically weaponizes identity verification costs ; conviction voting enables continuous preference signaling with natural manipulation resistance now deployed by 1Hive and Gitcoin GG24; and Optimism’s two-house system explicitly operationalizes the democracy-technocracy tension through separate Token House (coin-weighted protocol decisions) and Citizens’ House (one-person-one-vote RetroPGF with 20% treasury committed).
A Selection of Critical Open Problems
The deeper challenge is defining and evaluating “good” allocation. Fairness requires distinguishing healthy coordination from collusion, should mechanisms penalize all group behavior or embrace it? Efficiency demands rewarding impact magnitude, yet aggregation rules that resist conflicts of interest (e.g. median-based) compress variance. Rules with higher variance (mean, quadratic, etc.) carry more information but are more fragile. Can we recover some expressiveness without opening obvious attack vectors? Legitimacy creates the most fundamental tension: pure preference aggregation might suffer from rational ignorance and information asymmetry, while expert evaluation concentrates power and lacks democratic mandate. Can we design mechanisms where these tradeoffs become less sharp, or does Arrow’s theorem suggest any allocation mechanism must decide which properties to relax? And how do we compare and evaluate allocation strategies with each other?
III. The Impact Problem: Measuring Outcomes and Creating Feedback Loops
Optimism’s founding insight, “it’s easier to agree on what was useful than what will be useful”, reduces uncertainty but doesn’t eliminate measurement challenges. Four rounds of RetroPGF distributing 71M+ OP reveal that even retrospectively, quantifying magnitude, attributing causality, and aggregating across diverse project types remains fundamentally difficult.
What We’ve Learned
Even retrospective measurement proves difficult: RetroPGF evolved from Round 1’s 24 badgeholders unable to differentiate impact magnitude to Round 3’s cognitive overload to Round 4’s metrics-based pivot that reduced load but introduced selection bias. Analysis confirms no perfect mechanism, median voting resists conflicts but produces low variance (top 1% received only 6x median), while mean/quadratic are more manipulable. Incentive bias persists even retrospectively (tobacco research funded by tobacco companies is 90x more likely to find no harm). Hypercerts solve accounting but not evaluation bottlenecks. The vision of “impact markets” depends on creating credible feedback loops: allocation bonds where early funders/mechanism designers stake capital on their choices and profit if retrospective evaluation validates high impact, or lose stakes if allocations prove ineffective, sending economic signals backward in time to reward good judgment and penalize poor decisions.
A Selection of Critical Open Problems
Fundamental questions remain: How do we attribute causality and quantify magnitude for public goods with long time horizons and compounding network effects, especially when measuring counterfactuals in complex sociotechnical systems? What institutions prevent incentive bias when evaluators have stakes in outcomes, should badgeholder track records be weighted over time? How should mechanisms account for downside risk and negative externalities when retrospective funding cannot impose penalties, only rewards?
IV. A Coordinated Research Agenda
Progress on the three problems require:
-
Continued theoretical innovation bridging mechanism design, cryptography, and social choice theory
-
Rigorous empirical evaluation of live experiments with transparent sharing of failures
-
Coordinated experimentation across funding programs to test complementary approaches
V. How to Get Involved
We propose three concrete next steps for the Ethereum research community:
1. Coordinate empirical research across funding programs
Gitcoin, Optimism, Octant, and others are running parallel experiments. We need shared evaluation frameworks, common metrics, and coordinated timing to enable cross-mechanism comparison. Researchers should engage with program operators to design studies that answer specific theoretical questions.
2. Build open-source infrastructure for mechanism experimentation
Lower barriers to deploying new mechanisms through composable smart contracts, simulation tools, and evaluation dashboards. Modular designs such as Octant v2’s TokenizedAllocationMechanism, exposes a hook-based interface so different allocation rules can be deployed and tested quickly on the same underlying accounting system.
3. Bridge academic research and practical implementation
The game theory, economics, and mechanism design communities contain deep expertise not yet engaged with Ethereum. Events like the Iceland Research Retreat and AI4PG aim to connect researchers with builders. We need more such bridges, workshops at academic conferences, funding for applied research, clearer pathways from theory to deployment.
Call for Feedback and Collaboration
This post represents an attempt to classify three important problems for public goods funding. We need your input:
-
Are we framing the problems correctly?
-
What critical research we’ve missed?
-
What experiments are you running that could inform the broader ecosystem?
-
What theoretical frameworks might unlock progress?
The public goods funding problem is fundamentally one of knowledge aggregation under deep uncertainty with misaligned incentives.
Join the conversation: Comment below with your thoughts, DM researchers working on these problems, or reach out if you’re building tools or running experiments. The ecosystem succeeds when we learn together.
This research synthesis builds on work by teams at Gitcoin, Optimism, Octant, Protocol Labs, and numerous academic researchers. All errors and omissions are mine.