§ 03 REPORTS / ETHEREUM / 2026-08-04

Ethereum Issuance from First Principles

FILED UNDER ETHEREUM § 2026-08-04

The machine that pays for Ethereum's security, the two-year fight over how much is too much, and the EIP filed this morning that would take net issuance to zero at 50% staked.

tl;dr

Contents

  1. Why issue anything at all
  2. The machine as built
  3. Monetary policy by hard fork
  4. February 2024: the first battle
  5. What the theory says
  6. Where the numbers are now
  7. The proposal on the table
  8. Follow the yield
  9. Five ways out
  10. What is issuance for
  11. Where this goes
  12. Sources and methodology

Why issue anything at all

Start with what issuance is not. It is not yield, not a dividend, not a reward for belief in the roadmap. A proof-of-stake protocol mints new tokens for exactly one reason: to make its history expensive to attack. Everything a validator is paid to do reduces to that one service. Anders Elowsson, the debate's most systematic economist, puts it plainly in "Foundations of minimum viable issuance": "staking is a service, and performing it comes with costs to the home staker, node operator, and delegator." Issuance is the price the protocol pays for that service, and the entire design question is how much attack-cost is enough and what the cheapest way to buy it is.

Proof of stake exists because it buys attack-cost more cheaply than proof of work. Vitalik Buterin's November 2020 "Why Proof of Stake" runs the numbers: per dollar per day of rewards, GPU-mineable PoW buys roughly $0.26 of attack cost (the hardware is rentable), ASIC PoW roughly $486.75, and PoS roughly $2,189 — staked capital neither depreciates nor burns energy, so the honest staker's cost is illiquidity and the capital comes back. He calls it a "5-20x gain in security-per-cost," where in PoW the cost of maintaining consensus is "real electricity being burned in insanely large quantities." PoS also changes what happens after an attack: slashing and minority soft forks destroy the attacker's capital, so each unit of it spends only once. That is the premise. The issuance debate is about whether Ethereum then kept buying more security than anyone ordered.

From the holder's side, issuance is not creation but transfer. Every minted ETH dilutes every existing ETH — what matters is the proportion of supply held, not the token count — and the clean accounting is the proportional yield y_p = (1+y)/(1+s) − 1, nominal yield deflated by supply growth. A non-staker's real return is −s/(1+s): they pay the inflation and receive nothing, a seigniorage tax on holders paid to stakers. At the limit where everyone stakes, as Mike Neuder points out, supply grows at the rate you earn and real yield rounds to zero. Worse, the transfer leaks. "Issuance policy is emphatically not zero sum," in Elowsson's phrase: high yields compel holders to stake defensively, merely to avoid dilution, incurring real costs — hardware, liquid-staking fees and counterparty risk, illiquidity, taxes on income that is substantially inflation. His worked example: a 5% yield with everyone staked and a 20% average tax on staking income dissipates one percent of the entire market cap to tax authorities every year. Excess issuance is negative-sum.

The first-principles frame for everything that follows is a market. The reward curve is the protocol's demand curve for security: at each quantity of stake D, it names the yield it will pay. Holders' reservation yields — the minimum return at which staking beats each holder's costs and risks — aggregate into an upward-sloping supply curve, with an elasticity Elowsson puts around 2 in the mid-range while stressing it "remains empirically uncertain". The stake rate is where they cross, and the curve's shape — not just its level — is the policy instrument, since shape determines how far the equilibrium drifts when staking gets cheaper. Vitalik's Serenity design rationale framed the original choice exactly this way: "If a fixed reward rate is set too low then almost no one will participate, threatening the network, and if a rate is set too high then very many validators will participate, leading to unexpectedly high issuance"; a fixed total reward, meanwhile, invites discouragement attacks, and the inverse-square-root curve "compromises between the two and avoids the worst consequences of each one." A compromise — not an optimum. That distinction carries the next four years of argument.

The machine as built

The machine is small enough to hold in your head. The consensus layer pays a base reward proportional to BASE_REWARD_FACTOR / sqrt(total_active_balance), the factor set at 64 — the constant eth2book calls "the big knob that we could turn if we wished to change the issuance rate," unmoved since Beacon Chain genesis. Per-validator yield falls as 1/√D while validator count grows as D, so total annual issuance grows as √D: at most about 940.87·√N ETH per year for N 32-ETH validators, equivalently ≈166.3·√D for D ETH staked — Elowsson's y = cF/√D with F = 64 and c ≈ 2.6. Half a million validators (16M ETH) earn roughly 665,000 ETH a year at about 4.2%; quadruple the stake and issuance doubles while yield halves. More stake always means more issuance, at a decelerating rate, forever. There is no cap and no peak.

The reward splits by duty, weights summing to 64:

Duty Weight Share of issuance
Timely source vote 14 21.9%
Timely target vote 26 40.6%
Timely head vote 14 21.9%
Sync committee 2 3.1%
Proposal 8 12.5%

Attestations — the steady, every-epoch heartbeat work — carry about 84% of issuance; proposers receive one seventh of what attesters receive for including their attestations, an eighth of the total. Issuance mostly pays for the boring part on purpose.

But issuance is only one of a validator's two incomes, and the other is the architectural fact the whole debate turns on. Priority fees and MEV are paid on the execution layer, out of user transactions and the block-building market — revenue from usage, not minted ETH, and no issuance policy touches them. Cut consensus rewards to zero and a validator still earns fees and MEV: execution-layer income is a floor under staking yield that no reward curve can remove. It also bounds issuance from below — Elowsson's consensus-safety guideline is to keep the protocol-issued share of total validator yield above one half, and never below one quarter, lest honest per-slot behavior become a rounding error next to timing games and MEV-driven reorgs. The machine buys security with one currency while the market pays validators in another, and the two are wired to different controls.

The third component runs in reverse. EIP-1559, live since August 2021, computes a base fee per gas that adjusts up to ±12.5% per block around a gas target and destroys it — "The base fee per gas is burned" — leaving only the priority fee to the block producer, in part because burning "counterbalances Ethereum inflation while still giving the block reward and priority fee" to producers. Net supply change is issuance minus burn, and neither mechanism references the other: issuance is a function of stake, burn of blockspace demand, and "ultrasound money" names the race between them. Justin Drake coined the framing in September 2020 — if capped-supply gold is sound money, decreasing-supply ether is ultra sound money — and for a while the race was won: from the Merge until Dencun, burn exceeded issuance and supply genuinely shrank, Etherscan measuring −0.22% annualized inflation in its 2024 overview. Then Dencun's blob market moved L2 data out of the gas-fee burn, and ETH has not been sustainably deflationary since April 2024, running about 0.74% annualized inflation by that September. The meme outlived the property, and the gap between them is itself a political force in what follows.

Monetary policy by hard fork

Ethereum has never had a fixed monetary policy; it has had a habit. The chain launched in July 2015 with 5 ETH per block, no supply cap, and an understanding that issuance was adjustable by hard fork — and it was adjusted, repeatedly, downward, each time without serious resistance. Byzantium (October 16, 2017) shipped EIP-649, by Afri Schoedon and Vitalik Buterin, cutting the block reward from 5 to 3 ETH; Constantinople (February 28, 2019) shipped Schoedon's EIP-1234, cutting 3 to 2 — the thirdening. Before the cuts, annual inflation had run as high as 7.5%; across the PoW era, mining and uncle rewards issued roughly 50M ETH, about 42% of supply. No central bank did this — a release manager wrote an EIP and the community ratified it by running the code. Both sides of the later war would claim the history: reducers as precedent, opponents as a credibility account that each withdrawal drew down.

The philosophy got a name before it got a formula. During Eth2 design, "minimum necessary issuance" — pay no more for security than security requires — was ethos rather than enforced parameter, documented as a core design principle in Edgington's book. Vitalik tested the harder-money boundary in April 2018 with EIPs issue #960, proposing a hard cap at 120,204,432 ETH — dated April 1, usually read as at least half provocation — which went nowhere. In May 2020 a pseudonymous contributor, lightuponlight, filed consensus-specs issue #1784, arguing a 1% annual issuance ceiling offered "a good Schelling point" for minimum necessary issuance; the curve was never touched, and the issue was eventually swept up in an April 2025 housekeeping close. MVI commanded assent as an ethos and produced no mechanism. Elowsson began turning ethos into economics in October 2021 — "the issuance of new tokens should be high enough to secure the blockchain, but not higher" — now with supply-equilibrium models attached.

Meanwhile the machine kept being rebuilt underneath the debate. London (August 5, 2021) activated EIP-1559 and the burn. The Merge (September 15, 2022) ended PoW issuance entirely — roughly 13,000 ETH per day of mining rewards gone, leaving about 1,600–1,700 ETH per day of staking issuance, a cut of roughly 90% the community called the triple halving. Shapella (April 12, 2023) enabled withdrawals, deleting the last risk discount on staking, and deposits surged. The 2020 reward curve — designed, as EIP-7514's own motivation concedes, before launch, for a world without liquid staking tokens, a matured MEV market, or withdrawals — now faced a question it had no answer to: what happens when staking is easy and the curve never says stop?

The first answer was a stall. EIP-7514, created September 7, 2023 by dapplion and Tim Beiko and shipped in Dencun (March 13, 2024), capped validator activations at 8 per epoch, converting validator-set growth from exponential to linear. The motivation was blunt: "In the event that the deposit queue stays 100% full, the share of ETH supply staked will reach 50% by May 2024, 75% by September 2024, and 100% by December 2024." The stated purpose was to slow the approach to those milestones and allow more time for research into more comprehensive solutions — buying time, explicitly, for an issuance decision nobody had yet proposed. It passed with limited drama because it was small, temporary, and left the reward curve alone; even so, critics in the discussion thread noted that a churn cap is monetary policy, since it alters the growth path of stake and therefore of issuance. By late 2023 Ethereum had admitted, in shipped code, that stake growth was a problem — while deferring every question about the curve that caused it. The deferral lasted five months.

February 2024: the first battle

On February 22, 2024, two Ethereum Foundation researchers, Ansgar Dietrichs and Caspar Schwarz-Schilling, published the answer the churn limit had been buying time for — a case for stake-ratio targeting as the endgame, and, as an interim step aimed at the Electra fork, a concrete change to the reward curve: from y = cF/√D to y = cF/(√D·(1+kD)), k = 2⁻²⁵ — Elowsson's tempered curve, which mimics the current one at low stake and bends yield down hard as deposits grow. The projections: annualized dilution of non-stakers capped near 0.4% against roughly 1.5% under the status quo, with the staking equilibrium pushed toward roughly a quarter of supply. The reasoning was the first-principles case made operational: under the existing curve nearly-everything-staked stays economically viable, the winning liquid staking token then displaces raw ETH as the network's working money, and — in the proposal's words — "For true economic scalability, Ethereum's de facto money should be maximally trustless: ETH."

The backlash arrived within hours, from two directions at once. The staking industry attacked the economics. vshvsh — a handle associated with Lido co-founder Vasiliy Shapovalov of p2p, first reply, same day: lower rewards "will tighten up margins for staking," driving margin-cutting and, he argued, vertical integration — with consequences he predicted would be "quite gruesome for decentralization on staking level." Oisín Kyne of Obol: the change would "harm marginal staking service providers" and "make home stakers non-viable." Valdorff of the Rocket Pool community: it "would essentially end solo staking (with the exception of some very large solo stakers)." Others noted institutional ETH would end up staked regardless while entry barriers rose — almost no net new solo stakers — and that with a ~25% target a custodian like Coinbase could plausibly control 51% of the staked set. The ETH-money camp attacked the legitimacy. Eric Conner, EIP-1559 co-author: "The general disregard for how hard we've worked for a decade establishing ETH being money is concerning." Martin Köppelmann: changing the curve "does not fundamentally improve Ethereum – it just shifts incentives from one group to another." Jon Charbonneau: "these tweaks try to solve an unsolvable problem of fundamental tradeoffs in PoS."

Four arguments did the killing, and they are not the same argument. First, the solo-staker squeeze: yield compression hits the highest-cost, smallest-scale operators first, while professionals with MEV optimization and economies of scale absorb it — a policy sold as decentralization-preserving could centralize the validator set. Second, the social contract: stakers committed capital under known rules, and an ad-hoc curve change read as a retroactive rewrite — a hit to the monetary credibility the PoW cuts, the burn, and the ultrasound identity had spent years accruing. Third, security: either you accept less economic security or you redistribute income among staker classes; the tweaks, critics argued, escape neither horn. Fourth, process: an EF-researcher-authored change to monetary policy, moving on a fork timeline, mid-SEC-pressure, struck even sympathetic observers as the wrong authors moving at the wrong speed. Note that the first and second sit awkwardly together — one defends small stakers' yields, the other defends holders' hard-money expectations — a coalition that would not survive a proposal designed to split it. In 2024 it did not need to: each camp was right about the cost it could see, and the proposal had no answer that satisfied both.

So it died — not by any formal vote, but by the absence of rough consensus. No issuance EIP entered Electra. Pectra shipped on May 7, 2025 with eleven EIPs that restructured validator mechanics — consolidation to 2048-ETH maximum balances, execution-layer exits, in-block deposits — and touched rewards not at all. Fusaka followed on December 3, 2025 with PeerDAS and, again, zero issuance changes. The debate went quiet; the variable it was about did not. Deposits kept arriving through 2025's regulatory opening to institutional staking, and by April 2026 the staked share of ETH supply crossed one third for the first time, with an entry queue that had not cleared in almost a year. The 2024 proposal was defeated on the argument that intervention was too costly. What the next round — taken up in the sections that follow — would have to price was the cost of the alternative.

What the theory says

The issuance debate has an academic spine, and it is thinner than either side likes to admit. The canonical paper is Tarun Chitra's "Competitive Equilibria Between Staking and On-chain Lending" (arXiv:2001.00919, presented at Stanford Blockchain Conference and MIT Cryptoeconomic Systems 2020, written at Gauntlet), and its motivating question is one sentence long: "Suppose that we assume that validators are rational financial agents. Would they not simply move their assets between staking and on-chain lending, depending on which has a higher yield?"

The machinery is deliberately spare. Each of n agents holds a two-asset Markowitz portfolio — staked tokens and lent tokens — and rebalances by mean-variance optimization; lending pays a Compound-style utilization-driven rate, staking pays the protocol's block reward schedule, and risk preferences are tied to time horizons. Chitra then proves the system has a genuine phase transition: his Claim 5 shows the lent supply flipping between supermartingale and submartingale regimes depending on where the lending yield sits — in the paper's own image, the system concentrates into "the house (staked supply) or the gambler (lent supply)." When lending out-yields staking, rational agents drain the validator set together, "coordinated only by rational optimization" — the paper calls it a bank run, and no Byzantine adversary is required. The headline conclusion gets quoted for a reason: "PoS in deflationary systems is unstable and unlikely to work." Deflationary schedules produce rebalancing cascades that scale with the terminal coin supply; polynomial inflation — even simple linear growth — keeps stake sticky. And the churn itself is the wound: observed volatility in the amount staked is "tantamount to dramatically reducing the cost of taking over a staking network." The prescription is calibration, not generosity — "one needs to choose a block reward schedule that increases relative to the yield that these on-chain contracts provide."

Now the footnote that carries more weight than the theorem. Assumption 7 of the paper, titled "No Transaction Fee Revenue," states: "The only revenue that validators receive from staking comes from the block reward" — justified because fee markets "currently have unstable dynamics and are poorly understood." Sit with that. The model most often cited to argue that issuance must be sustained excludes by assumption the exact revenue source — fees and MEV — that a zero-issuance Ethereum would run on. Chitra was not hiding this; he lists fee markets as future work. But it means the paper tells you validator yield must stay competitive at the margin, and is silent on whether that yield may be fees — which is, depending on your priors, either a modeling convenience or the whole argument against a fee-only security budget, smuggled in as a premise.

The sequels complicate the picture in both directions. Chitra and Evans's "Why Stake When You Can Borrow?" (2020) anticipated the stETH economy before it existed: staking derivatives let capital hold the staking and DeFi exposures on the same collateral, which "can potentially mitigate the capital flight issues" of the first paper. The model finds a sharp transition between "safe" and "unsafe" derivative regimes, and — "contrary to previous work" — conditions where derivatives reduce wealth concentration among stakers. But capital flight persists even in the safe region, and consensus now has to care about the derivative's market, because a leveraged claim on stake makes validator default risk consensus-relevant. In 2026 terms: liquid staking tokens blunt the staking-versus-DeFi tradeoff that drives Chitra's bank run — and in doing so they also blunt any policy that relies on low yield pushing stake out the door.

Then Chitra and Kulkarni, "Improving Proof of Stake Economic Security via MEV Redistribution" (ACM CCS DeFi Workshop 2022), the closest the peer-reviewed literature comes to blessing a low-issuance regime. It bolts an MEV revenue flow onto the 2020 dynamical system and shows that validators who capture a redistributed share of MEV "are disincentivized to unstake and lower economic security." Translation: the sustainable flow the Chitra framework demands need not be issuance — fees and MEV can substitute, if they are actually routed to stakers by design. Post-merge Ethereum routes priority fees and MEV-Boost payments to proposers by market convention, not enshrinement; how much weaker that is than the paper's condition is an open question the current proposal does not model.

The broader economics literature loosens the issuance–security link further. Budish's "Trust at Scale" (QJE 140(1), 2025, from the 2018 NBER paper) set the pessimistic baseline: Nakamoto-style security is a flow cost — attack resistance is linear in the recurring payment to consensus participants, which makes trust "expensive to scale." But Budish, Lewis-Pye, and Roughgarden (2024) showed the escape hatch is specific to proof of stake: with slashing, a PoS protocol can be "expensive to attack in the absence of collapse" — attack cost becomes the confiscatable stock of stake rather than a perpetual flow, provided synchrony holds and the adversary is under two-thirds of stake. Security stops being a subscription and becomes a bond. John, Rivera, and Saleh prove staking levels are not even monotone in block rewards — issuance is "an inflationary transfer from short-horizon cryptocurrency investors to long-horizon cryptocurrency investors," and raising it can shrink the dollar value of the staked collateral — while Saleh's RFS 2021 model finds modest rewards actively good for consensus, muting the incentive to perpetuate forks. And the restaking papers (Durvasula–Roughgarden; Chitra–Pai) make the competing-yield problem permanent: even if DeFi lending rates fell to zero, commissions on restaked ETH itself would keep bidding for the same collateral, with formalized cascade risks.

So the honest synthesis runs like this. The literature genuinely supports capping overpayment: security is stock-based under slashing, more issuance does not monotonically buy more of it, and excess yield attracts capital that centralizes the validator set. It is genuinely uneasy about a hard zero: the one formal model of stake-flow dynamics says yield must stay competitive at the margin and that deflationary regimes cascade — and a zero-issuance Ethereum with EIP-1559 still burning edges toward exactly the configuration Chitra flagged, funded by the fee revenue his model refused to touch. Nothing in the formal literature analyzes the conditional case — issuance that resumes below a threshold — which is where the actual proposal lives. One calibration footnote for the record: no Gauntlet eth2 report for the Ethereum Foundation exists, despite the claim's circulation; the real 2020 industry baseline is Hoban and Borgers's "Ethereum 2.0 Economic Review" (ConsenSys personnel, MolochDAO-funded), which estimated a 13.8% stake rate as adequate for security under historical price conditions. Ethereum today has two and a half times that.

Where the numbers are now

Theory says staking yield must stay competitive at the margin. Here is the margin, as of August 4, 2026:

Instrument Yield Denomination As of
US 3-month T-bill 3.78% USD 2026-08-03
Sky Savings Rate (sUSDS) ~3.5% USD 2026-08-04
Aave v3 USDC supply 3.30% USD 2026-08-04
ETH staking, network APR 2.65% ETH 2026-08-04
Lido stETH 2.21% ETH 2026-08-04
Aave v3 WETH supply 1.45% ETH 2026-08-04

Ethereum staking currently loses the yield competition Chitra's model describes — to the T-bill, to the Sky Savings Rate, to lending USDC on Aave. The classic leveraged-staking loop is dead: stETH at 2.21% against Aave's 2.12% WETH borrow rate is a gross spread of about nine basis points, before gas and liquidation risk. Within the ETH-denominated universe plain staking still beats lending ETH; only restaking strategies claim more, at materially different risk, mostly paid in AVS tokens.

The stock: 41.4M ETH staked, 33.98% of supply, across roughly 894,600 validators — an average of ~46 ETH per validator, the signature of post-Pectra consolidation. The entry queue holds 2.5M ETH at a ~43-day wait; the exit queue is effectively zero. The flow: gross issuance is running at ~1.05M ETH per year, about 0.86% of supply, while the EIP-1559 burn has collapsed to roughly 12–20K ETH per year annualized — base gas sat at 0.3 gwei on August 4. Net supply growth is +0.85% per year. "Ultrasound money" is, at current activity levels, a dormant meme: supply is up roughly 1.4M ETH since the Merge, and the deflationary window on the since-1559 chart belongs to the high-burn 2021–2023 era, not to now.

The composition of the validator paycheck is the number that reframes the debate. The displayed 2.65% APR decomposes into roughly 2.54% consensus issuance and about 0.09% execution-layer income — a fresh measurement, not an estimate: total network priority fees over the trailing 30 days summed to ~3,050 ETH on a Dune query run for this report, annualizing to ~37K ETH/yr on 41.4M staked. MEV-Boost proposer payments tell the same story from the other side: ~56 ETH per day across listed builders in the August 4 window — a one-day snapshot, so treat the ~20K ETH/yr annualization as order-of-magnitude — with Titan building 52.6% of MEV-Boost blocks and the top four builders at ~96%. Validator revenue is roughly 96% issuance right now. The "MEV adds 0.5–1%" figures still circulating in secondary sources describe an earlier regime; they would require ten to twenty times the execution-layer flow observably on-chain. Whatever a fee-funded security budget could look like in principle, in August 2026 the fee-funded fraction of the security budget is a rounding error.

Market structure, per on-chain entity attribution refreshed August 4: Lido holds 8.76M ETH, about 21.0% of stake — down from a peak above 30%. Coinbase holds ~1.61M ETH (3.9%), Binance ~3.27M (7.9%), ether.fi ~1.66M (4.0%, down ~30% over six months). Liquid staking plus liquid restaking together account for ~28% of all staked ETH. Solo stakers were last credibly estimated at ~5.4% of stake in June 2024 — a figure now two years stale, with no current measurement in existence, which is itself a data point about who gets measured.

And here is the empirical puzzle the theory section did not predict: staking is losing the yield race and growing anyway. The stake rate bottomed at 27.2% in March 2025 (per Everstake) and has climbed to ~34% — while ETH fell to $1,878, down 62% from its August 2025 all-time high of $4,946, and the APR sagged toward 2.6%. The sequencing inverts the yield-chasing story: the big exit wave hit in September 2025 near the price top; the entry queue swelled through the 2026 drawdown. The parsimonious explanation is that the marginal staker of 2026 is not Chitra's rational rebalancer choosing between staking and Compound — it is an ETF, a treasury company, or a custodian staking because the mandate says to stake. Cumulative Ethereum ETF net inflows reached ~$10.5B by mid-July, and the largest corporate holder, BitMine, holds ~5.6M ETH, about 4.7% of supply (both per Datawallet). Institutional demand for staked ETH looks yield-inelastic within the observed range: it responds to allocation decisions and product structure, not to fifty basis points. Two caveats keep this honest: Sygnum notes that top-ups and post-Pectra consolidations run through the same entry queue as fresh stake, so the 43-day backlog is not a clean demand signal; and no formal study of stake-flow elasticity to yield exists for this period. But the direction is not in doubt — the ratio is rising toward the very threshold the new proposal is built around, at a yield the theory says should be repelling capital.

The proposal on the table

On the morning of August 4, 2026 — the day this report is dated — a draft EIP landed that would end Ethereum's open-ended payment for stake: EIP-8361, "Tapered Issuance Burn", a Core, hard-fork-requiring change whose one-line description is to "burn a fraction of validator rewards that rises with the staking ratio, removing the issuance incentive to stake more than 50% of all ETH." The header says created: 2026-07-14; the PR and its Ethereum Magicians thread went public today. Six authors, in order: pintail (lead — he opened both the PR and the thread), Jérôme de Tychey, dapplion, pa7x1, Ladislaus von Daniels, and Justin Drake, listed last. The press will call it Drake's EIP; that is convenient and wrong. More accurately, it operationalizes the endpoint Drake has campaigned for since his February 2025 "croissant issuance" push, when he told DL News: "Unfortunately, like Bitcoin's issuance, Ethereum's issuance was misdesigned. There's rough consensus [that] the current curve is broken and needs to change," and "To me a 50% staking soft cap feels credibly neutral and pragmatic." The intellectual credits inside the EIP go elsewhere: pa7x1 for the core burn principle, Anders Elowsson for the per-duty structure — credited, but notably not an author of record, despite having spent years building the minimum-viable-issuance case this proposal cashes in.

The mechanism is a burn, not a new curve, and the distinction is load-bearing. Gross issuance is computed exactly as today. Then every validator is charged a deduction for every duty it was assigned — attestation, proposal, sync committee — sized as a fraction of the idealized reward for that duty, and the deducted ETH is destroyed. The burn fraction is

b = (D / SATURATION_BALANCE)^(3/2), clamped at 1

where D is total active stake and SATURATION_BALANCE = 60.25M ETH — the proposal's single permanent new constant, pinned at roughly half of current supply. The 3/2 exponent makes net yield decline linearly to zero at the saturation ratio (the underlying yield curve scales as f^(−1/2), so a linear-in-f deduction picks up an extra half power). The resulting net issuance curve stops being monotone: it peaks at about 19.8% of supply staked and is fully cancelled at 50%. Because the deduction is charged against idealized rewards whether or not the duty was performed, every marginal incentive to actually show up is exactly as strong as today; because it is levied per duty, an honest attester never goes negative in an ordinary epoch. Execution-layer income — priority fees and MEV, which the EIP itself bounds at ≤0.20% annually — is untouched. Fifty percent is a saturation point, not a cap: nothing stops staking above it; issuance just stops paying for it, and the authors expect the market to settle wherever net yield meets the marginal staker's risk premium, strictly below the line.

The EIP's own numbers are candid about the bite. Applied cold at today's ~33% ratio, net yield would drop from ~2.6% to ~1.2% — "enough to prompt a substantial exit of stake on activation," in the proposal's words — hence an 18-month phase-in: BASE_REWARD_FACTOR temporarily doubled from 64 to 128, then walked back down a 65-step staircase (~8.6 days per step) so stakers start near today's yield and glide. The anti-concentration argument gets a number too: an operator holding 50% of all stake stops earning anything from further growth once ~31% of supply is staked, the cutoff arriving soonest for the largest operators — under the current curve no such point exists at any size. And the solo-staker cost gets a number, which is where the trouble starts: the ETH cost of downtime is unchanged, but measured in days-of-net-earnings, recovery from an outage stretches by ~3.8× at today's ratio. The proposal even arrived with client-team code: a self-labeled draft implementation in Prysm, committed July 28 by terence, Prysm's lead developer.

The politics began within hours, because the filing was itself a political act: the draft landed 48 hours before the August 6 "proposed for inclusion" deadline for the Hegotá fork. The first reply on Magicians, from Greg Koumoutsos, made the process objection: "This clearly doesn't leave adequate time for community review of a monetary policy change of this magnitude" — and noted that community fork-planning had placed issuance changes in the fork after Hegotá. Co-author de Tychey's answer was that PFI is the start of the argument, not the end — "Being proposed for inclusion is what opens the floor for feedback, not what closes it" — paired with an urgency case that the staking ratio passes 55% by January 2028 on current trajectory: "Acting now means the market settles into an equilibrium below 50% but acting after the overshoot means correcting a much larger imbalance, with more stake forced to exit and more disruption for every participant." His close: "The gentle path is only available now." The staking industry's first word came from vshvsh (a handle matching Lido co-founder Vasiliy Shapovalov, though the identity is unconfirmed): "Ethereum direct economic security (stake under slashing) is very strong for a very long time" — questioning the premise that stake growth needs arresting at all — and warning that squeezing validator margins "will forcefully transform the security model from decentralization-driven to concentrated professionalism-driven." Solo stakers zeroed in on the 3.8× downtime factor and the composition claim, with goodroot asking the question the EIP does not answer: "What evidence or model supports the claim that lowering staking APR improves the composition of the validator set?" The most constructive early position, from JulianT, called the concerns overblown — the most yield-sensitive cohort is leveraged LST-loopers, whose exit would be hygienic — but proposed a slower taper and a floor of ~1–1.5% "rather than a trend towards actual 0." And Aave founder Stani Kulechov questioned the proposal the same day, per launch-day coverage.

Where it stands as this report goes out: a draft hours old, unmerged, one editor approval short — and, as of this evening, flagged for renumbering: "8361" was self-assigned, that number already belongs to another proposal, and an EIP editor has assigned 8363 in review, so the identifier in the launch-day coverage (and, by necessity, in this report) is unlikely to be the one that ships. No PFI pull request against the Hegotá meta-EIP had been opened as of this evening — the first All Core Devs venue is ACDC #184 on August 6, deadline day itself — and, this deserves emphasis, there is no independent third-party modeling of the post-saturation regime, in which Ethereum's security budget would be funded entirely by the execution-layer flow measured above at 0.09%. Its strongest cards are real: a complete spec, a draft client implementation from Prysm's lead developer, six authors spanning EF research, client development, and community organizations, and a two-year effective runway. Its weakest card is the one this report turns on: the equilibrium it promises — stake settling politely below 50% as yield-sensitive capital exits — is precisely the elasticity the 2026 data does not exhibit, and the fee-only endgame it implies is precisely the regime the theory's founding paper declined to model. The stakeholder fight over all of this comes later in the report. The mechanism, at least, is now on the table with a number attached.

Follow the yield

Every monetary argument is also a cap table. Before adjudicating the steelmen, it is worth writing down who actually earns Ethereum's ~1,054,000 ETH per year of consensus issuance, and what each party has said in public — because the debate's loudest claims come from parties with positions, and its most exposed parties have said almost nothing.

Actor Material interest Stated position
Solo stakers (~5.4% of stake at the last credible measurement, June 2024) Fixed home-operation costs, no fee revenue, taxed on nominal yield, no MEV smoothing. Most exposed to cuts on cohort math; least exposed on the sunk-cost argument. Split. EthStaker's 2024 survey (n=1,024): 50% feel protocol research "either ignores solo stakers or is largely powerless to help solo stakers against monied interests." Both camps claim to be their champion.
Lido Fees on the ~21% of staked ETH it holds today (the 2025 cohort study modeled ~28%) (cohort study); revenue scales with ratio × yield. Studied neutrality plus a warning. reGOOSE (proposal authored by Hasu, Lido strategic advisor; post signed "Lido"): reductions "would likely have severe consequences on the decentralization of the network, in particular affecting solo-stakers and supplier technology such as DVT" — while committing to decentralize stETH regardless of the debate's outcome.
Coinbase ~1.6M ETH attributed on-chain (3.9% of stake; the 2025 cohort study modeled a larger custodial book) at a 25% retail commission; prime execution agent on BlackRock's staked ETF. No public position located. Modeled as the debate's most insulated party: its flows are "largely inelastic," and the cohort study flags it as a relative winner under cuts — and a growing 51%-risk concern.
EigenLayer / ether.fi / restaking Sells supplemental yield stacked on base issuance (EigenLayer ~$5.0B TVL per DefiLlama). Lower base issuance makes restaking yield relatively more important — and makes restaking the live version of the Chitra outside option (below). No formal positions found.
Staked ETFs (Grayscale ETHE, Oct 2025; BlackRock ETHB, Mar 2026) Yield is the product: net distributions of 1.9–2.6% after fees (Everstake). On ETHB, investors receive 82% of staking rewards; the remaining 18% is split between BlackRock and Coinbase, with no public breakdown of the split. No public comments on issuance policy found. Their inflows are, per pintail, the proximate cause of the ratio breaching one-third in April 2026.
L2s Second-order: hold and bridge ETH, benefit from ETH-as-pristine-collateral narratives, pay the blob fees a fee-only L1 would lean on harder. No prominent stated positions in the record.
Ethereum Foundation EF researchers authored every major reduction proposal. Since February 2026 the EF itself stakes ~70,000 ETH of treasury, rewards returning to the treasury to fund ecosystem stewardship. The institution proposing to cut the yield is, modestly, earning it — fair to note; also fair to note ~70k ETH is under 0.2% of stake.
Non-staking ETH holders The named beneficiary of reduction: dilution relief, ETH-not-stETH as money. Diffuse and unorganized — classic concentrated-costs, diffuse-benefits politics. Represented, in practice, by the EF research wing and ultrasound-money X.

One update on the Lido row, because the reflexive version of the critique is out of date. Lido now has a genuinely permissionless entry path: the Community Staking Module has been live since October 2024 and fully permissionless since January 2025 — anyone posting a 1.5–2.4 ETH bond can run validators — and it has grown to ~8.5% of Lido's stake, inside a 9% cap that LDO governance sets. And since June 30, 2025, Dual Governance gives stETH holders a delay-and-rage-quit veto over LDO decisions — shipped, live, never yet triggered. What has not changed: ~90% of Lido's stake still runs through 37 curated operators whose admission — and the permissionless module's ceiling — remain LDO-governed, so the permissionless share is real and growing, and it grows exactly as fast as the governance token the critique targets allows.

Now the steelmen, one per camp, in their strongest verified form.

The pro-reduction case, at full strength, is not about inflation. It is about what issuance buys. Elowsson's framing: every staker has a reservation yield, and issuance above the level that clears the target quantity of stake is not security purchase but pure transfer — an inflation tax paid by holders to intermediaries. And the marginal ETH staked today is almost entirely delegated: SSPs "derive their revenue from SSP fees" and are "shielded from" the dilution downside, so each marginal ETH of issuance buys another validator key run by the same handful of operators — near-zero decentralization per unit spent. Worse, the subsidy compounds into the asset layer. The targeting post's core sentence, quoted exactly because it is the debate's most load-bearing claim: "LSTs compete on money-ness with winner-takes-most if not all dynamics due to network effects." Keep paying everyone to stake and "the de facto money of Ethereum for most use cases besides L1 transaction fees will be some LST(s)" — users "economically quasi-forced" into a token carrying issuer and governance risk. Ethereum's trust model quietly becomes Lido's trust model. And at very high ratios the security story inverts outright: pintail argues excess staking makes Ethereum "more brittle, not more secure," because dominant providers become too big to slash, and Neuder notes that at 80%+ staked a mass-slashing event would destroy so much value that social pressure to intervene becomes overwhelming — accountable safety undermined by its own collateral.

The anti-reduction case, at full strength, is not about protecting yield. It is about who exits first when yield compresses. The unit economics run one direction in every published competitive model: Beccuti, Chantramonklasri, Hafner & Oderbolz find solo stakers the most yield-elastic cohort, crowded out toward exchanges and LSTs that monetize MEV and DeFi yield; the staking-cohort study works through eight cohorts and concludes the losers of a modeled cut are solo stakers (~$1,500 fixed costs, real yield down ~60%, tax due on nominal yield without real gain) and DVT operators, while Coinbase and Lido gain relative share — disclosure: that study is by FranklinDAO and funded by cyber•Fund, a Lido-cofounder-affiliated fund, which does not refute its arithmetic but belongs next to it. Oisín Kyne (Obol) states the asymmetry in one line: a CEX's marginal cost per validator is under $10 a year against $500+ for squad stakers running nodes at home, so capping issuance "creates centralization pressure." Second, the revenue-quality argument: strip out smooth issuance and validators are left holding lottery tickets — lumpy, proposer-concentrated MEV and fees. The canonical instability result here is Carlsten, Kalodner, Narayanan & Weinberg (2016) on fee-only Bitcoin, where it becomes rational to fork wealthy blocks and steal their fees. To be precise about provenance: that paper is strictly Bitcoin PoW, and no anti-reduction principal in the record formally cites it — the transfer to Ethereum PoS is this report's analysis, not the camp's canonical citation. But the mechanism is conceded by the other side's own literature: Drake's MEV-burn design is motivated by "spike smoothing" precisely because reward spikes create reorg and equivocation incentives. Third, Chitra's competitive-equilibria result: stakers rebalance between staking and on-chain lending, and when outside yields beat staking yields the system can cross a phase boundary into a predominantly-lent equilibrium — cut issuance below the market's outside option and stake can flee faster than the curve assumed. Fourth, the social contract. Eric Conner, via Unchained: "The general disregard for how hard we've worked for a decade establishing ETH being money is concerning." Martin Köppelmann, same source: the change "does not fundamentally improve Ethereum — it just shifts incentives from one group to another." Paul Dylan-Ennis read the 2024 blowup as a governance fight, not a curve fight: changing issuance is a red line because it is "reminiscent of central bankers constantly tweaking monetary policy." Capital staked under stated terms is being repriced by the researchers of a foundation that now earns the yield itself. That is the kind of discretion crypto exists to foreclose.

Both steelmen are about intermediaries. One says the subsidy builds an LST monopoly; the other says removing it builds a CEX one. Hold that symmetry — it is the whole debate, and we will come back to it.

Five ways out

Five live paths, each with whatever published modeling actually exists. Where no model exists, that absence is reported as a finding, because unmodeled is not the same as safe.

Path one: do nothing. The sqrt curve keeps a yield floor of roughly 1.5% even at 100% staked — there is no ratio at which the incentive to stake switches off. Elowsson's calibration (hypothetical supply curve, ~300k ETH/yr of realized extractable value) puts the current-curve equilibrium around ~50M ETH staked at ~2.94% yield, rising to ~80–87M under flatter supply curves; Dietrichs and Schwarz-Schilling's supply-curve analysis lands at "100M ETH or ≥80% of all ETH staked" with dilution approaching ~1.5%/yr. pintail's empirical read is blunter: the ratio is "up only," the entry queue has not cleared in almost a year, and above ~1/3 staked a solo staker's dilution-adjusted, tax-adjusted return goes negative while LST holders defer — "once solo stakers are pushed out, there is no mechanism to bring them back." The status quo is not neutral. It is a policy of letting dilution and LST network effects compound, defended most directly by the Maximum Viable Security post (authors affiliated with cyber.fund, Lido DAO, Steakhouse Financial, Progrmd Capital), which wants "corruption costs of hundreds of billions" and opposes predetermined caps.

Path two: temper the curve. The 2024 program — the Electra adjustment (divide the sqrt curve by 1+kD) and Elowsson's Option A. The proposals' own numbers: real staking yield down ~30% at 30M staked, worst-case dilution capped at ~0.4%/yr versus ~1.5%, and under Elowsson's full-reduction calibration the equilibrium moves to roughly ~33–34M ETH at ~2.34% total yield — the yield figure is verbatim, the quantity a graphical reading of his Figure 4. Against it, the two funded-model results: Eloranta & Helminen (funded by cyber•Fund — same disclosure discipline as above) find large pools already earn ~12% higher mean returns than single-validator setups, widening to ~15% under substantial reduction; and Beccuti et al. predict the exit order under compression — solo first, CEX last. Moderate tempering changes where equilibrium lands, not the fact that the curve never switches off.

Path three: EIP-8361. Filed today, August 4, 2026, 48 hours before the Hegotá submission deadline, by pintail, Jérôme de Tychey, dapplion, pa7x1, Ladislaus von Daniels, and Justin Drake. (Anders Elowsson is not an author; the EIP's per-duty burn structure is consistent with the design constraints in his offsets work, but influence is not authorship.) Rather than rewriting the reward curve, it burns a rising fraction of every validator reward — min(1, (stake/60,250,000 ETH)^3/2), applied per duty whether or not the duty was performed — so that net consensus issuance hits exactly zero at a 50% staking ratio. At today's ~33% ratio it would cut net consensus yield from ~2.6% to ~1.2%, hence an 18-month phase-in (base reward factor doubled at activation, decaying back over ~123,300 epochs). It removes the one property no previous plan removed: the yield floor. Above saturation, the marginal staker earns execution-layer fees and MEV only — ~0.2% today — and the market, not the curve, discovers the stopping point. What it does not have is an equilibrium model: there is, as of today, no published analysis of the post-saturation fee-only regime — how stake oscillates around the threshold, what happens to attestation incentives when duty rewards are negligible next to MEV (the failure mode Elowsson's offsets post is built around), or whether LSTs at fee-only yields simply convert staking slots into scarcity rents harvested through the moneyness premium of the wrapper. That last question is argued in the targeting-thread replies and modeled nowhere. And the Chitra phase boundary sits directly across the path: a policy that deliberately pushes marginal staking yield to the fee floor is a standing invitation for outside yield markets to set Ethereum's security level.

Path four: change the architecture, not the price. Rainbow staking (Monnot, Feb 2024) unbundles heavy slashable professional duties from light lottery-based censorship-resistance duties with thresholds as low as 1 ETH; The Scourge floats capping the active set (~2^19 validators) and two-tier designs in which only a bounded risk-bearing slice of stake — Vitalik's example is on the order of one-eighth — carries slashing exposure. This is the only family that addresses solo participation structurally rather than through yield levels, and it has essentially zero published economic modeling: no equilibrium projections, no parameterization, no fork slot. Its known hazards are already named in the literature — caps convert slots into rented scarcity, fair rotation is unsolved, and fixed-quantity targeting maximizes discouragement-attack incentives, per Elowsson's FAQ, Q19: with a fixed target, less stake must leave to drive up the yield, so pushing rivals out pays.

Path five: fix the demand side. MEV burn — or its execution-auction cousins — auctions block-building rights and burns the proceeds. Elowsson's Figure 5 is the quantitative treatment: under full MEV burn his equilibrium shifts down to ~30M ETH staked, with ~300k ETH/yr of burn accruing to all holders, and — the distributional point — it deletes the MEV-variance penalty that makes small validators lottery players. This is the one intervention both camps endorse: Kyne wants it before any issuance cut; Elowsson's solo-staker floor calculations assume it. The fight is sequencing, and the sequencing is infrastructure: enshrined MEV burn needs ePBS, which is Glamsterdam's headliner. The passive variant — do nothing and let restaking absorb the yield — exists as a talking point with a theoretical scaffold (Chitra's phase transition, popularized as "How DeFi cannibalizes PoS security") and no Ethereum-specific model. Its defect is structural: it makes L1 security endogenous to unregulated third-party yield markets, and leverage loops historically push the ratio up, not down.

Path Security Decentralization Monetary properties Feasibility (Aug 2026)
Status quo Maximal stake; brittleness and too-big-to-slash at high ratios Solo real after-tax yield negative past ~1/3; LST network effects compound Supply +~0.85%/yr at today's burn; dilution → ~1.5% at high stake Default; wins every stalemate
Tempered curve ~33–34M ETH equilibrium; attesters keep ≥ half of rewards Large-pool advantage widens 12%→15% (cyber•Fund-funded model) Dilution capped ~0.4%/yr; issuance ≤0.5% supply Rejected once (2024); template for revival
EIP-8361 60M+ ETH collateral, but incentive security unmodeled post-saturation Fee-only margin favors scale; LST rent question open Zero net CL issuance ≥50% staked; strongest ultrasound guarantee Draft filed today; Hegotá slot contested
Structural redesign Bounded consensus weight; selection-rent attack surface Only structural answer for solo participation Unmodeled Research stage; no EIP, no numbers
MEV burn / demand side Removes proposer spike games; equilibrium ~30M Fixes solo variance — both camps want it +~300k ETH/yr burn, deflationary regardless of curve Hostage to ePBS in Glamsterdam

What is issuance for

Underneath the curves is an unresolved question of purpose, and the camps are answering different versions of it. Four teloi are live in the record. Issuance buys security, and only security — Elowsson's minimum viable issuance, on which everything above the market-clearing price of stake is deadweight. Issuance is a decentralization instrument — Vitalik's Scourge framing, on which a cheap but Lido-dominated validator set is a failure even if secure. Issuance policy is a monetary constitution — and here, remarkably, Drake's ultrasound wing and the social-contract conservatives share a premise, that ETH's moneyness is the prize, while disagreeing about which act destroys it: continued dilution, or discretionary repricing. And issuance funds participation as a value in itself — the position in the Practical-endgame replies that wants everyone staking even if that means no real yield. These are not four weights on one objective function. They are four different objective functions, which is why eighteen months of modeling has converted almost nobody.

But the models, read together, do force one honest conclusion, and it is uncomfortable for both camps. Every path that stops stake growth compresses yield, and every published competitive model — Eloranta & Helminen, Beccuti et al., Kyne's OpEx arithmetic — says compression evicts solo stakers first. And the status quo evicts them anyway: pintail's tipping point already puts the diligent home staker at negative real after-tax yield, with LST network effects compounding and no mechanism to bring exited solo stakers back. There is no path on the table under which the home stakers flourish by default. So the debate is not, whatever its participants say, security versus scarcity. It is a choice of intermediary structure: subsidize stake and get an LST-dominated set whose winner competes on money-ness with winner-takes-most dynamics; compress yield and get a CEX-and-ETF-dominated set that survives on scale economics and inelastic customers. Who pays also differs — dilution taxes every holder to fund the first outcome; yield compression taxes stakers to fund the second. The report's read: the sides are not arguing about whether to have a staking oligopoly. They are arguing about which one, and who buys it.

Two facts sharpen the next round. First, the MEV-burn hinge. Both camps' models quietly assume it: the anti camp's variance objection dissolves if MEV is burned, and the pro camp's solo-staker floor is computed assuming it exists. A policy fight this bitter, resting on a shared unbuilt prerequisite, is really a fight about sequencing — and sequencing is decided by fork slots, which is why ePBS in Glamsterdam matters more to issuance politics than any curve parameter. Second, the new veto constituency. In 2024 the opposition was solo stakers and staking businesses. In 2026 it includes regulated products with prospectuses: staked ETFs distributing 1.9–2.6% net, built on a joint 18% gross-reward split between the world's largest asset manager and America's largest crypto exchange. EIP-8361's immediate 2.6%→1.2% consensus-yield cut would push some ETF net distributions toward zero — an arithmetic consequence with no published product-level model, and a constituency with lawyers.

And the Chitra result deserves to be read as its author wrote it, not as either camp deploys it. It does not say "never cut issuance." It says staking yield competes in an open market for the same capital, and the system's equilibrium can move discontinuously when the outside option wins. The proper use is not as a veto but as an instrument check: know what revenue floor you are standing on — fees, MEV, restaking spreads, none of them protocol-controlled — before you remove the one revenue source the protocol does control. The pro camp's plans increasingly acknowledge this (per-duty offsets, phase-ins, floors). The anti camp's invocation of it would be stronger if anyone on that side published the Ethereum-specific version of the model. Nobody has. On the central empirical questions of this debate, the striking fact is how much of the artillery on both sides is still borrowed.

Where this goes

The clock is explicit now. EIPs proposed for inclusion in Hegotá are due August 6 — two days from this writing, 48 hours after EIP-8361 landed, which is precisely Koumoutsos's objection: "This clearly doesn't leave adequate time for community review of a monetary policy change of this magnitude." Whether All Core Devs treats the EIP as a serious Hegotá candidate, defers it to the next fork, or lets it die in review is the first observable — the first scheduled venue is ACDC #184, the consensus-layer call that lands on deadline day. Four more are worth watching. Whether anyone publishes an equilibrium model of the post-saturation fee-only regime — the single largest hole in the literature, sitting under the single most aggressive live proposal. Whether ePBS ships in Glamsterdam on schedule, because MEV burn's feasibility decides whether the two camps' shared assumption becomes real before or after any cut. The staking ratio's climb from one-third toward the taper's bite, with the entry queue a year deep and ETF inflows still arriving. And the first public comment from Coinbase, BlackRock, or an L2 — the actors with the largest positions and, so far, total silence.

What would actually move each camp is also legible. The pro-reduction camp has said its price: evidence that solo stakers are yield-fragile rather than yield-insensitive — a funded, replicated version of the cohort arithmetic without the cyber•Fund provenance question — would break the claim that cuts cost nothing in decentralization. The anti-reduction camp's price is a shipped MEV burn plus a post-saturation model showing duty incentives and solo economics survive the fee-only regime; reGOOSE's own language concedes that its objection is the security budget and the small operator, not the yield itself. If both prices were paid, the remaining disagreement would be the honest one: whether Ethereum's monetary constitution permits repricing staked capital at all.

The 2024 round ended in deferral, and deferral had a beneficiary: every year of status quo enlarges the coalition whose cash flows depend on the curve — more LST TVL, more ETF assets, more DAT treasuries — while the option space shrinks. That is Neuder's "preserving optionality is the main reason to take action" and the anti camp's proof that the social contract has already vested, and both readings are correct, which is the problem. Issuance policy is the one lever Ethereum's social layer must pull against the interests of its own largest constituents or admit it cannot. Either answer settles a bigger question than the curve.

Sources and methodology

This report was compiled on August 4, 2026 — the day EIP-8361 was published — from primary sources, live measurements, and cross-verified reporting.

Primary sources include the EIP-8361 specification text fetched directly from ethereum/EIPs PR #12081 and its Ethereum Magicians thread; the ethresear.ch and Ethereum Magicians threads of the 2024 issuance fight and their replies; the Elowsson corpus (circulating-supply equilibrium, minimum viable issuance, Properties of Issuance Level, tempered-issuance, Practical Endgame, the issuance-reduction FAQ, and the offsets work); Vitalik Buterin's Why Proof of Stake, Serenity design rationale, and The Scourge; the EIP texts of 649, 1234, 1559, 7514; and the academic literature — Chitra (arXiv:2001.00919), Chitra–Evans (arXiv:2006.11156), Chitra–Kulkarni (ACM CCS DeFi '22), Budish (QJE 2025), Budish–Lewis-Pye–Roughgarden (arXiv:2405.09173), John–Rivera–Saleh, Saleh (RFS 2021), Beccuti et al. (arXiv:2503.14385), Eloranta–Helminen, Durvasula–Roughgarden, and Chitra–Pai — read from the papers themselves, with key propositions quoted from the source PDFs.

Quantitative state comes from live aggregators and Dune queries cited inline, including a purpose-run measurement of trailing-30-day priority fees (Dune query 8223420) and on-chain staking-entity attribution (query 2394100), MEV-Boost relay data from relayscan.io, supply and burn figures from ultrasound.money and Etherscan, and market prices from CoinGecko. Where sources disagree — the CL/EL yield split, Coinbase's staked book, total-supply conventions — the disagreement is stated and the on-chain measurement is preferred. Two figures are knowingly stale and labeled as such in the text: the solo-staker share (last credible estimate June 2024) and any pre-drawdown restaking yields.

Every verbatim quote was re-verified against the original thread, post, or paper before inclusion; statements that survive only as paraphrase are presented without quotation marks. Claims that could not be verified were dropped, including several widely circulated ones — there is, for instance, no findable "Gauntlet eth2 report for the Ethereum Foundation." Where a cited empirical study has funding adjacent to a party in this debate, the funding is disclosed in the same paragraph as the finding. Analytical transfers made by this report rather than by a named participant — most notably the application of Carlsten et al.'s fee-only instability result to Ethereum proof of stake — are labeled as this report's analysis at the point of use. All figures and market states are current as of August 4, 2026 unless a different as-of date is given.