On September 16, a House committee will mark up cryptocurrency tax rules.
That is a fact. It is also nearly empty. No bill text. No committee named โ "House committee" could mean Ways and Means, which holds tax jurisdiction, or Financial Services, which does not. No year attached to the date. No indication whether the draft touches broker reporting, wash sales, staking income, or de minimis thresholds.
Within an hour of the headline, three separate threads priced it as "regulatory clarity." That is a category error, and it is the same category error that gets contracts drained.
The position I intend to defend: the only clause in this bill that matters is the definition of "broker," and it is the clause least likely to appear in the headline. Rates, forms, and deadlines are downstream of that single word.
Context: What a Markup Actually Is
Markup is procedural. A committee takes a draft, amends it line by line, votes, and โ if it clears โ advances it. It is not law. Between markup and statute sit a full chamber vote, a second chamber, reconciliation, and a signature. Twelve to twenty-four months is a normal horizon when nothing goes wrong, and something usually goes wrong.
The United States has been here before. The 2021 Infrastructure Investment and Jobs Act carried a broker definition broad enough to plausibly capture miners, validators, and wallet developers. The industry spent two years litigating that word. Treasury's 2023 proposed digital asset reporting rules reintroduced the same ambiguity rather than resolving it. Form 1099-DA exists because someone eventually had to decide who files it, against what basis, and on what schedule.
What September 16 signals is not that taxes are coming. Taxes were always coming. It signals that the reporting rail is being finalized.
The statutory machinery is old and specific. Section 6045 assigns reporting obligations to custodial intermediaries. Section 6045A covers transfers between brokers. Section 6045B covers organizational actions. Every provision assumes the existence of an account, a custodian, and a payer.
Cryptocurrency does not reliably produce any of those three objects.
Core: The Clause Doing All the Work
A definition is an interface specification. That is the correct mental model, and it is why the broker clause deserves more attention than the rate schedule. A definition states what a system must expose to the outside world. Change the interface and every downstream implementation must be rewritten โ not because the logic changed, but because the contract changed. The art is the hash; the value is the proof. In legislation, the definition is the ABI, and the rest of the bill is a client library that will not compile against an older version.
Consider the two readings.
If "broker" means custodian, only centralized exchanges file. The perimeter is small, the data already exists, and the compliance delta is nearly zero.
If "broker" means "any person who effectuates transfers," then contract accounts enter the perimeter. Routers. Aggregators. Wrapper contracts. Possibly wallet software. Possibly front-ends that never hold a private key and exercise no discretion over execution.
A router has no discretion. It executes a bytecode path chosen by the caller. Calling it a broker is like calling TCP/IP a financial institution โ technically it moves value, legally it has no custodial relationship with anything. But statutes are not written by people who read bytecode, and that gap produced the 2021 fight.
Cost Basis Is a State Reconstruction Problem
Tax law operates on lots. Acquire 1.0 ETH at $2,000 on March 3 and 0.5 ETH at $2,400 on June 11. Dispose of 1.2 ETH at $3,100. Which lots moved? FIFO, LIFO, or specific identification โ the choice changes the gain, and therefore the liability.
Ethereum's account model stores a scalar balance. There is no queue in state. The chain does not know which ETH you sold, because there is no such thing as "which ETH." The sequencing exists only in the custodian's database.
Bitcoin's UTXO model is friendlier โ discrete outputs exist on-chain. But the chain stores no acquisition date and no fiat basis. Those live off-chain, in someone's spreadsheet, accumulated across years of migrations and consolidation transactions.
The ledger does not contain the data the tax code requires. The tax code requires data the ledger was explicitly designed not to hold. Cost basis is not a property of the protocol. It is a property of the intermediary's bookkeeping โ which is precisely why the broker definition carries the entire bill.
This is where the self-custody position is structurally weaker than its advocates admit. Not morally. Structurally. A self-custodial user must now maintain a duplicate, adversarial ledger that the protocol never asked anyone to maintain, and keep it reconcilable against an external authority's schema.
I have run this accounting. In 2020 I reverse-engineered the Uniswap V2 constant product formula and built a Python simulation across more than 500 liquidity pools to test whether published impermanent loss heuristics held for large trades. They did not. Several early lending protocols were running oversimplified models, and the correction forced dashboards to be rebuilt. The lesson: attribution collapses when a position rebalances continuously.
An LP position in ETH/USDC is not a holding. It is a short volatility position denominated in two assets, with a fee stream attached and a continuously shifting ratio between the legs. Tax law asks what you acquired, when, and at what basis. The honest answer is a differential equation. The statute wants a receipt.
The Receipt Problem in DeFi
Constructive receipt is the second fault line. Staking rewards, LP fees, airdrops, MEV rebates โ when is income realized? At accrual, per block, or at claim?
Per-block accrual is technically correct and operationally impossible. Ethereum produces a slot every twelve seconds. A validator accrues roughly 2.6 million reward events per year. Per-event recognition generates a tax record per slot, per validator, priced against a fiat reference that must itself be sourced at that timestamp. No consumer tooling can do this. No institutional tooling wants to.
Claim-time recognition is sane and legally contestable, because constructive receipt does not care about operational convenience.
Then wrapped assets. Convert BTC into a bridged representation and the economic exposure is unchanged. Legally, one asset was exchanged for another and a disposition may have occurred. The chain has no concept of exposure. It records a state transition. Reentrancy doesn't ask permission, and neither does a lot-identification rule โ the difference is that a reentrancy guard is a five-line modifier, while lot identification requires an entire wallet history to be reconstructible on demand, years later, by a third party.
The Reporting Rail Is an Enforcement Substrate
What the markup actually constructs is not a tax schedule.
Mandatory third-party reporting converts private intermediaries into data suppliers. The output is a canonical, per-identity, per-transaction ledger: not held by the state, but submitted to it, in a normalized schema, at scale. Add the counterparty reporting that travel-rule regimes already model, and the graph becomes transitively closed โ your transactions link to your counterparties' transactions, which link to theirs.
Jurisdictions designing retail central bank digital currencies are building the same schema from the opposite direction. One is described as tax administration. The other is described as monetary modernization. They converge on identical fields: identity, amount, timestamp, counterparty, purpose. You do not need to build a surveillance currency when a reporting requirement already produces the ledger.
There is a resilience argument too, and it is the one I trust most. In 2021 I showed that roughly 60% of popular NFT collections failed when gateway providers altered caching policies. The token was on-chain. The metadata was not. Ownership was contingent on an operational decision made by a company nobody had audited. Centralize a dependency and you centralize the failure mode.
Apply that to reporting. If most US-visible volume reports through a handful of entities running a handful of schema implementations, one systematic error โ a misclassified lot-identification default, a timezone bug in timestamps, a mis-specified transfer type โ propagates to every taxpayer downstream. The rail becomes critical infrastructure with no independent verification, tested in production, on users who cannot opt out.
Contrarian: Clarity Is Not a Catalyst
The consensus reading is that regulatory clarity is bullish. I think that reading inverts the mechanism.
Clarity in tax law does not reduce obligation. It converts an undefined obligation into a defined one, and defined obligations are enforceable. What the market prices on a headline like this is the absence of a lawsuit. What it receives is a specification. Those are different assets, and only one of them has a cash flow attached.
The second inversion concerns who pays. Compliance cost falls on the transparent, not on the evasive. The entities that will file under the new regime are the ones already filing. The activity the rule nominally targets sits structurally outside the perimeter: rotated wallets, non-KYC venues, jurisdictional exit. I have argued that most KYC is theater โ the perimeter is bypassed by acquiring a wallet's holdings rather than its identity, while documentation cost is passed to users who never had a choice. Tax reporting inherits that shape exactly.
The third point should bother engineers. Legislation is the only codebase in production that is never audited. We require formal verification before a $50M bridge ships. We do not require an independent technical impact assessment before a statute governing every transaction in a multi-trillion-dollar asset class. In 2018 I refused to sign off on the Parity multi-sig library until the ownership update sequence was patched and proofs were added โ a two-week delay against a Q2 deadline, with management pressure to approve anyway. That standard is unremarkable in code. It is entirely absent in law. Statutes reach production with no peer review and no adversarial testing, and the market absorbs the bugs. Everything here falls under someone's scrutiny except the text itself.
Takeaway
Watch three signals and ignore the rate.
Whether "broker" reaches contract accounts and user interfaces โ this determines whether DeFi remains accessible from US soil at all. Whether wash sale rules extend to digital assets, removing loss harvesting from the last asset class that still permits it. Whether a de minimis threshold appears, which would reveal whether the draft is written for administration or for extraction.
The vulnerability forecast: the first serious enforcement action built on this reporting rail will not target a tax evader. It will target a non-custodial interface that never possessed the data required to comply. We do not build for today.
Watch the ABI, not the rate.
