choworks.dev

A Dead Airline's Emails Just Sold for $10 Million

AI training datarecord keepingdata as assetsolo businessGoogle
The moment a pile of records becomes an asset Older records lie faded at the bottom and grow sharper toward the top. A price tag hangs off the topmost one. Data as Asset Email · Chat · Source code · And the years nobody deleted Data as Asset Email · Chat · Source codeAnd the years nobody deleted

A story from last week has been sitting in my head for days. Google bought a dead airline's internal email, all of it. Ten million dollars.

What Google actually bought

Spirit Airlines was a US low cost carrier. It shut down in May 2026, and the bankruptcy court auctioned off what was left. One of the lots was every byte of data the company had generated while it was alive.

Here is the inventory.

What How much
Internal emails 100 million
Teams messages 500 million
Lines of source code 30 million
Plus flight operations records, inventory, HR and marketing workflows, employee productivity data

Bidding opened around $5 million. Mercor, an AI training data company, went to $7.5 million. A third anonymous bidder was in the room. Google took it at $10 million. Passenger profiles and loyalty records were excluded, and the terms require a third party to scrub personally identifiable information before Google receives anything.

So here is where we are. A few years of coworkers emailing each other about work turned into a ten million dollar asset once the company died.

A few years ago this was garbage

I sat with that one for a while.

When a company folds, you sell what has value. Aircraft. Office furniture. The brand. And the mail server, the chat logs, the shared drives? Historically the responsible thing was to delete all of it before it became a privacy liability. It was overhead. It cost money to keep.

If someone had told me a few years back that a bankrupt airline's internal small talk would fetch eight figures, I would not have believed them.

Same records, different price The internal email, chat logs and source code a company piles up would have been quietly deleted a few years ago. In August 2026 the same data sold to Google for ten million dollars. The data did not change. The buyer did. A few years ago August 2026 100M internal emails 500M Teams messages 30M lines of code 100M internal emails 500M Teams messages 30M lines of code Only time passed The data is identical Deleted $10 million A storage cost, nothing more Google, at a bankruptcy auction What changed was not the record. It was who showed up to buy it. A few years ago 100M internal emails 500M Teams messages 30M lines of code Deleted A storage cost, nothing more Only time passed The data is identical August 2026 100M internal emails 500M Teams messages 30M lines of code $10 million Google, at a bankruptcy auction What changed was not the record. It was who showed up to buy it.

Why dead companies specifically

The simple reason: the public internet is already scraped. That job finished sometime around late 2024. Wikipedia, news, blogs, public GitHub, all of it went in. What is left is what was never public.

But the more interesting reason is about the kind of data. Models have seen plenty of well written prose. What they have not seen is how work actually happens.

  • Who asked whom what, at 11pm, when something broke
  • Why a decision that was already made got reversed the next morning
  • Why the fourth version of a quote got approved when three were rejected
  • What the unwritten rule was that everyone in that company knew and nobody documented

None of that is in a blog post or a paper. It only exists inside companies. Which is why a company has to die before it reaches the market.

This market was already running

Spirit made headlines because of the size. Buying and selling the data of dead companies has been a business for a while.

  • SimpleClosure has brokered close to 100 shutdowns in the past year, typically $10,000 to $100,000 per company
  • Sunset does the same thing. Pricing depends on company size, age, and "data richness," a measure of how much the internal records connect to each other
  • Mercor goes after individuals. It has reportedly offered up to $300,000 for the work product someone produced at a previous job

Read that last one again. Not companies. Individuals.

So I opened my own hard drive

I have run a small company for ten years. Things accumulate.

Quotes, contracts, client email threads, invoices, work chat groups, notes scattered through project folders, one cell in a spreadsheet explaining why we priced that job the way we did. I have always treated this as a mess I should clean up someday. Storage I was wasting.

Looking at it now, it is not a mess. It is a complete record of what one person went through for ten years and how they decided things. There is exactly one copy of it in the world.

To be honest about it, I do not think my archive is worth ten million dollars. That is not the argument. The argument is this. I just watched something go from trash to asset, and the price was not set by the people who created it. It was set by a buyer who showed up years later.

The paradox of automation

I hand most of my work to AI now. Moving my website went that way. Generating quotes went that way. More of it will.

So what is left that is mine? I ask myself that a lot lately.

An answer is coming into focus. What is left is not execution, it is context. AI can do my work for me. But what I lived through over ten years and why I judged things the way I did does not exist anywhere unless I write it down. You cannot search for it. I am the only one holding it.

And what is happening right now is that exact category, the stuff that only existed inside companies, is getting a price tag.

What to keep, and how

Start with something grand and you quit in three days. I hold to about five things.

Stop deleting. Archive instead of delete. Do not throw out old project folders in a tidying mood. Storage is one of the cheapest things in the world right now.

Pull it into one place. Chat apps will export a conversation. Gmail comes out whole through Google Takeout. There is no guarantee the service you use today exists in ten years. A record that lives only on someone else's server is not yet yours.

Write the reason, not the result. This is the one that matters. "Moved to n8n" is worth nothing later. Why I moved a perfectly working Apps Script over to n8n is worth something. Results are available everywhere. Reasons are only known to the person who lived them.

Keep the format boring. I write markdown files and commit them to git. Not because it is clever, but because it will still open in ten years. A record locked inside a tool dies with the tool.

Your AI conversations are records too. I learn by interrogating Claude. Those transcripts contain exactly what I did not know and the order in which I figured it out. I keep those too.

This blog is that record. Some of it is written for other people. Half of it is written for me to read later.

Nobody can see the future

Who would have guessed that 500 million Teams messages from a failing airline would be worth ten million dollars? The people who wrote them were just doing their jobs that day. Not one of them typed a message thinking it would become anything.

Which means you can run the logic in reverse. What the small note I write today turns into a few years from now is not knowable. The unknowability is the reason to keep it.

Whether it ends up having value is not my call. Exactly one thing is my call.

Whether it still exists.

Write it down. Write it down again. Keep writing it down.

In a similar spot? Feel free to ask. [email protected]

← All posts