As in F104-05, being straight with you: this is a category, not a named incident with a figure. These failures are private and the people they happen to do not publish them.
What is documented is the mechanism, in two halves.
The burden of substantiating basis rests with the taxpayer. F108-03 is the demonstration: the 1099-K recipients who resolved their notices quickly were the ones who could produce a reconciliation, and the ones who could not spent months and money establishing something that had been true the whole time.
And the record source most people rely on is a counterparty. Mt. Gox took a decade to distribute. FTX and Celsius entered bankruptcy. Cryptopia was liquidated. In every case the customer's own transaction history sat inside an entity that had stopped operating normally.
No obligation runs to any of them keeping your history reachable for the years you might need it. F107-04's counterparty exposures apply to your data exactly as they apply to your balance, and nobody thinks to size that one.
There is no bad decision here either. The decision is the one deferred. Records are cheap to capture at the moment of a transaction and expensive to reconstruct afterwards, and the gap between those two costs grows every year while the reason to close it stays invisible until a letter arrives.
United States federal framing, education rather than advice. The habit generalises; the tax consequences around it do not.
Six fields
At the moment a transaction happens, capture these. It takes fifteen seconds and it is the entire lesson.
1. Date and time. With the timezone. Prices move within a day and a reconstruction that is out by a day is out by a number.
2. What left. Asset and amount.
3. What arrived. Asset and amount. For a purchase, this is the currency. For a swap, F108-01 reminds you that both sides matter.
4. The fiat value of both, at that moment. This is the field people skip and the one that cannot be recovered casually, because "the price that day" is a range and you need a defensible point in it, from a source you can name.
5. The fee. In whatever asset it was paid. Fees usually adjust basis or proceeds, and paying a fee in crypto is itself a disposal.
6. Where it happened. Which wallet, which account, which venue. This is the field almost nobody records and it is now compulsory in effect, because F108-02's per-wallet rule means basis without a location is basis you cannot place.
Why yours and not theirs
Two independent reasons, and either one alone would be sufficient.
The burden of proof is yours. Not the broker's, not the software's. When a number is questioned, you produce the support or the number does not stand, and F108-02's arithmetic showed that an unsupported basis behaves as zero at six times the liability.
Every external source is a counterparty. The autopsy lists four venues that stopped operating normally. Add to that: exchanges migrate systems and purge old data, tax software companies are acquired and shut down, block explorers change their APIs, and a wallet application can stop supporting a chain.
The one source that does not disappear is the chain itself, which is the useful half of F101-05: your on-chain history is permanent and public and reconstructable in principle. What the chain does not record is what anything was worth in dollars at the moment, or which wallets were yours, or what you paid on a venue that never touched a chain. Those are the fields that only exist if you wrote them down.
The annual reconciliation
Once a year, while the year is still fresh in your memory. Two or three hours.
1. Export everything. Every venue, every wallet. Full history, not the current year, because the tooling will need earlier lots.
2. Import into a tax tool. They are imperfect and they are far better than doing it by hand.
3. Reconcile against the venue's own records. Where the tool and the venue disagree, find out why. This is where you catch the third failure mode in F108-03.
4. Find at least one thing the tooling got wrong. F108-L requires this because it always happens. The usual candidates: transfers between your own wallets classified as disposals, which inflates your gains; a swap where one side's price was pulled from a thin market; a fee treated as a disposal or not treated at all; an airdrop valued at a moment you did not control it; a bridge classified as a sale.
5. Produce the reconciliation document. Not the raw exports. A statement saying: these disposals, with these proceeds, correspond to these acquisitions at these prices from these sources. This is the artefact that resolves a notice, and F108-03's autopsy is what it is worth.
If you do one thing from this course, do this one, and do it before the reconciliation rather than as part of it.
Export your full history from every venue and every wallet you currently use, now, while they all still exist. Save the raw files somewhere durable.
The reconciliation can wait for a quiet weekend. The export cannot, because it depends on parties continuing to operate, and the autopsy is four of them that stopped. An exchange that closes takes your history with it and gives you no notice, and the moment you discover this is the moment you needed it.
Then repeat the export every year, so that a venue's disappearance costs you at most one year of convenience rather than everything you did there.
Storage and retention
Retention. Records supporting a disposal need to survive well past the year of the disposal, and periods can be extended in cases of substantial omission. The conservative rule that avoids the whole question: keep basis records for as long as you hold the asset, plus a generous margin after you dispose of it. For an asset held ten years, that is a fifteen-year document, which is a different design problem from a shoebox.
Storage. Plain formats that will still open. CSV rather than a proprietary export. A copy that does not depend on one device, and one that does not depend on one company, which is F104-04's redundancy argument applied to a different asset.
And a note that this material is sensitive. A complete transaction history is exactly the identity-joined-to-balance pairing F106-04 spends a lesson telling you to avoid creating. Do not put it in a public place, do not hand it to a tool you have not evaluated, and keep it out of the AI tools in F111-06's Tier 2 unless the terms support it.
The habit, minimised
Realistically, most people will not log six fields per transaction forever. So here is the version that survives contact with real life.
At the time: if a transaction is unusual, large, or involves something you received rather than bought, write one line somewhere durable. Everything else can be reconstructed from exports.
Monthly: ten minutes. Export anything new. Note anything that would be hard to explain in two years.
Annually: the full reconciliation above.
Whenever you change the architecture: a new wallet, a new venue, a tier restructure. F108-02's per-wallet rule means a new location is a new set of books, and the moment to record what moved into it is when it moves.
I will just reconstruct it from the block explorer if I ever need to.
The chain is permanent, public and complete, and it is genuinely a better base layer than most asset classes offer. It is also missing exactly the fields that matter.
It does not record fiat value. The chain says 0.4 ETH moved. It does not say what that was worth, and reconstructing a defensible price for a specific minute, years later, from a source you can cite, for hundreds of transactions, is the expensive part.
It does not know which addresses were yours. F108-01 says a transfer between your own wallets is not a disposal. Proving that a particular address was yours in 2022 is a claim you have to support, and the explorer will not do it for you.
It does not contain anything off-chain. Everything that happened inside an exchange, including your original purchase, never touched a chain.
And it does not scale. Reconstruction is technically possible and priced by hours, either yours or a professional's. F108-03's autopsy is people spending months on it.
So the accurate version is that the explorer is an excellent fallback and a poor plan, and the difference between the two is about fifteen seconds per transaction paid in advance.
Closing F108
You now have the map of taxable events, the basis rules and what covered means, how the reporting can disagree with reality, and the habit that resolves it.
F108-L is the deliverable: export a wallet's full history, import it, reconcile against exchange records, produce a clean gain and loss statement, and identify at least one transaction the tooling classified wrongly.
Then take it to somebody qualified. The point of this course was never to make you your own tax adviser. It was to make you the kind of client who arrives with a reconciliation instead of a shoebox, which is the difference between an afternoon of their time and a month of it.
Capture six fields at the moment of every transaction: when, what left, what arrived, the fiat value of both, the fee, and where it happened, with the last one now compulsory in effect because basis is tracked per wallet. Your records are the asset, because the burden of proof is yours and every counterparty holding your history is one that can close, as Mt. Gox, FTX, Celsius and Cryptopia all did. Export everything from every venue today while they still exist, reconcile annually while the year is fresh, and produce a reconciliation document rather than keeping raw files. And treat the block explorer as a fallback rather than a plan, because it has no idea what anything was worth or which addresses were yours.