Founder of a Shuttered Crypto Startup Earned $25 000 by Selling Archives for AI

Keelan Miskell began winding down his crypto startup this year and did not expect to make any money from its digital archives. The users were anonymous, and the company was only a few years old. But MIT Technology Review reports that he has already received about $25 000 for the codebase. Another payment has not yet been made. Miskell's story shows how shuttered startups are becoming suppliers of raw material for model training.

What exactly was sold

As the article makes clear, Miskell did not expect any revenue because he considered the assets to be of little value. The user base is anonymized, and the brand never became well known. He therefore was unlikely to think of the archives as a product. But the company still had its code: the result of several years of development, with a history of decisions, edits, and mistakes. That is what the first $25 000 paid for.

The second part of the deal has not closed. Its size is not given in the available portion of the article, and the word "probably" in the original indicates that the final figure is still an estimate.

Why the archives of dead companies are suddenly in demand

The reasons can only be reconstructed, and what follows is analysis, not a retelling of the source. AI labs have already combed through the open internet. Public code, articles, and forums are enough for basic training, but there is a shortage of closed data from real development work: working repositories, internal correspondence, documentation, tickets. Such data does not sit in the open, and at living companies it is guarded by lawyers and security policies.

A company that is shutting down is in a different position. The founder controls all the assets, customer obligations are dwindling, and there is usually no time for lengthy legal approvals. For a lab, this is a chance to buy something that is not for sale on the open market.

Key facts

• Seller: Keelan Miskell, founder of a crypto startup that is closing in 2026.
• Company age: a few years.
• Users: anonymous, so he valued the user base low.
• Already received: about $25 000 for the codebase.
• Expected: one more payment, the size of which has not been confirmed.
• Buyers: AI labs, as the article's headline indicates.

Where the risks lie

The main question concerns the rights to what is being sold. Code may contain third-party licensed components, keys, internal comments, and fragments of personal information. User anonymity lowers the risk but does not eliminate it entirely: logs and databases often retain identifiers that, when matched against other sources, can be used to identify a person. For the labs, this is a question of data provenance. If a model is trained on material of unclear legal status, that may later surface in lawsuits and corporate audits.

There is also a reputational side. Employees and customers of a closed company do not always know that its archives ended up in a training set.

What founders who are closing a company should do

Before shutting down, it helps to take an inventory of assets: repositories, documentation, datasets, anonymized logs. Then check what you are actually entitled to hand over: the terms of user agreements, the licenses of dependencies, contracts with contractors. Secrets, access keys, and any personal data should be scrubbed before the deal, not after.

It also helps to decide in advance what you are willing to sell and to get written guarantees on how the data will be used. As Miskell's case shows, payment may arrive in installments, so the payment terms should be fixed in the contract.

For investors and lawyers, this is a signal: the asset list of a closing company now has a separate line for data. Three years ago it was neglected; today people pay for it.