
Illustration
Developments3 min read
The AI graveyard: if you're just a feature, you're a platform update
Relay, Notion Mail, Huxe, Ollie: TechCrunch added to its list of failed AI products yesterday. Most headstones tell the same story. The product wasn't bad. A platform shipped it.
Key facts
- TechCrunch updated its running list of shut-down AI products on November 14, 2026, adding Character.ai (bankruptcy in early October) and Ollie (shut down November 6), among others.
- Relay, an AI workflow automation tool pitched as a Zapier alternative, shut down in August 2026; TechCrunch says OpenAI, Google and others built similar automation into their own tools.
- OpenAI discontinued its ChatGPT Atlas browser on August 9 after less than a year; S&P Global Market Intelligence estimates roughly 42% of AI initiatives are abandoned.
Researched and written with AI, to Luigi's brief, positions and editorial rules. Automatically checked against the sources. How I work →
You know the moment. Your team loves an AI tool, the workflows run, the prompts are dialed in. Then the email arrives: we're shutting down, please export your data. That's no longer a one-off. It's a pattern, and by now it's pretty predictable.
Who's in the graveyard
TechCrunch keeps a running list of shut-down AI projects and updated it on November 14. New at the top: Character.ai, which filed for bankruptcy in early October, and the family assistant Ollie, which went dark on November 6. Its founder said the app couldn't stand out against rivals like Muse and hadn't raised enough capital.
Further down it gets interesting for business users:
- Relay, an AI workflow automation tool pitched as a Zapier alternative, shut down entirely in August 2026. According to TechCrunch, because OpenAI, Google and others built similar automation into their own tools.
- Notion Mail, launched in April 2025, was slated to shut down on September 22. Users preferred to run their inboxes with separate AI agents.
- Huxe, an audio app from former NotebookLM developers, closed in May 2026 as bigger platforms like Spotify expanded their own AI podcast tools.
Even OpenAI is pruning: the ChatGPT Atlas browser was discontinued on August 9 after less than a year, its best features folded into ChatGPT. The Sora app closed in March over high costs and weak retention. TechCrunch cites S&P Global Market Intelligence: roughly 42% of AI initiatives end up abandoned.
The platform eats the feature, not the company
Sounds like ordinary startup churn. It isn't, quite. Character.ai mostly died of lawsuits, Humane of its hardware. But Relay, Notion Mail and Huxe didn't have bad products. They had products that could become a checkbox in ChatGPT, Gemini or Spotify within one release cycle. Your code isn't a moat anymore. Your distribution is. That's true for founders, and it's just as true for the tools you bet your team on.
My rule of thumb: anything anyone can type into a chat in ten minutes isn't a product. The survivors will be tools that do what a chat window can't: watch continuously, remember the history, actually take action. Or tools that already own the users before the platform shows up.
The catch for buyers
For marketing teams the graveyard isn't a startup story, it's stack risk. Every tool that's just a good feature ships with an expiry date. The subscription is rarely the problem. The problem is the workflows, templates and learnings that live only inside that tool. When the tool dies, they die with it, often on short notice.
What I'd do now
- Run the feature test: for every AI tool in your stack, ask what ChatGPT, Gemini or Copilot will do natively within a year. If the answer is "almost all of it" → no long commitment.
- Write the exit plan before you need it: check today whether you can export data, templates and history. Prompts and workflows belong in your own doc, not only in the tool.
- Use case before price: happily pay for specialists that clearly beat a general chat at one job. Just ask first whether that will still be true in a year.
Bottom lineDon't buy features a platform can give away in an update. Buy tools that own something a chat doesn't.
Got AI news? Submit news →
FAQ
Which AI tools shut down in 2026?
According to TechCrunch, among others: the Zapier alternative Relay (August 2026), the audio app Huxe (May 2026), the assistant Ollie (November 6, 2026), Notion Mail, plus OpenAI's ChatGPT Atlas browser (August 9) and the Sora app (March 2026).
How can I tell if an AI tool has an expiry date?
In my view, with one question: could ChatGPT, Gemini or Copilot ship the same function in an update? If yes, check export options and an exit plan up front and avoid long commitments.
Source
TechCrunch: The AI graveyard: a running list of projects and startups that didn't make it



