The numbers in NEAR Protocol's HSP-029 proposal reconcile perfectly. That is precisely the problem.
A governance proposal that trims annual issuance from 2.5% to 1.6% over a 24-month linear glide path has been circulating for review. On first pass it looks like disciplined monetary engineering: clean math, a Solana precedent, a supply reduction framed at 4.36%. I ran the cross-checks before reading the narrative. All four arithmetic validations passed. Then I looked at what those validations actually assumed, and the proposal's confidence began to dissolve.
I do not trust the pitch; I audit the structure.
The staking yield equation is internally consistent and externally fragile. Standard L1 yield logic states that APR equals annual inflation multiplied by the share routed to stakers, divided by the staking ratio. At current parameters, 2.5% times 90% divided by 42.71% yields 5.268%. At the target rate, 1.6% times 90% divided by the same 42.71% yields 3.372%. Both match the proposal to the decimal. The deviation is zero in both directions.
Zero deviation is not a virtue here. It is a fingerprint. The calculation holds the staking ratio fixed at 42.71%, which means the proposal's headline yield is a static snapshot, not an equilibrium estimate. It describes a world in which no staker responds to a 36% cut in real yield. That world does not exist. Yield compression drives marginal stakers out, the staking ratio falls, and a falling denominator mechanically pushes per-token yield back up. The realized rate will almost certainly settle above 3.372%. The proposal does not disclose this assumption anywhere in its text. A reader who takes 3.372% as the outcome is reading a number that the protocol's own model contradicts the moment behavior enters the frame.
The supply figures survive reverse engineering, with a caveat that matters. Linear decay from 2.5% to 1.6% produces a two-year average of 2.05%, a 0.45 percentage point annual reduction against the status quo, and roughly 11.8 million fewer tokens issued across the transition. Dividing that figure back through the stated reduction implies a total supply near 1.31 billion NEAR, which sits within the public 1.2 to 1.3 billion range. The six-year cumulative reduction of 4.36% also reconciles, but only against a counterfactual baseline in which issuance never changes. Measured against current supply, the same reduction is closer to 5.0%. Both framings are arithmetically defensible. Only one is flattering. The proposal chose it.
Then the fourth check broke. The per-epoch adjustment of 3.75 basis points does not survive unit reconciliation against the 24-month schedule. Either the epoch count or the basis-point step is misstated, and the proposal does not resolve which. This is the kind of defect I spent six weeks chasing in 2017.
That year I was a security consultant auditing token distribution contracts for three Ethereum ICOs. One project, pitched on a $50 million pre-sale, had a reentrancy vulnerability buried in its distribution logic. The team's answer to every objection was that the numbers added up. They did. The arithmetic was flawless and the contract was still exploitable. I refused to sign off, the launch slipped two months, and the project's momentum died. My clients were furious. The vulnerability was real. Since then I have treated internal consistency as a starting condition, never a conclusion.
The Solana precedent cited in the proposal carries a timestamp of August 2026. I flag this without accusation and without resolution. It may be a forward-dated governance reference, a placeholder, or a transcription error. I cannot verify it, and the proposal offers no corroboration. When a document's supporting evidence cannot be independently checked, the burden shifts to the document's internal logic, and that logic has already failed one unit test.
Here is what the bulls get right, and it is not trivial. A lower issuance rate reduces structural sell pressure from validator rewards. It signals monetary discipline to institutional allocators who model dilution explicitly. And the glide path, rather than a step change, is the correct design choice: gradual decay lets the staking ratio and validator economics adjust without a discontinuous shock. The team that wrote this proposal understands mechanism design. The transition structure is better than most supply-cut proposals I have reviewed.
But mechanism design literacy makes the undisclosed static assumption harder to forgive, not easier. A team that can model a linear glide path can model staker exit behavior. Choosing not to publish that sensitivity analysis is a decision, not an oversight.
Liquidity is a mirage; solvency is the only truth. The same applies to governance math. A yield figure that assumes zero behavioral response is a mirage with a decimal point.
My audit background makes me unfashionably interested in what a proposal omits. The omission here is a single paragraph: an elasticity estimate showing what happens to the staking ratio when yield falls from 5.268% to 3.372%, and what the realized rate becomes under three exit scenarios. Without it, the 3.372% is a marketing number wearing an engineering costume.
The 3.75 basis point discrepancy needs a correction or an explanation in the next revision. The 2026 timestamp needs sourcing. The counterfactual framing of 4.36% needs a footnote. None of these are fatal. All of them are the difference between a proposal that can be audited and one that can only be believed.
Emotion is a variable I exclude from the equation. So is any number that arrives without its assumptions attached. The next revision will tell us whether HSP-029 was written to be verified or merely to be endorsed.