TechCrunch keeps its AI graveyard, a running list of the AI projects that didn’t make it — the startups that shut down, the launches that missed, the bets that got quietly buried. It is a long list, still growing: Relay, Character.ai, OpenAI’s super-app redesign, ChatGPT Atlas, Apple’s delayed Siri, Microsoft’s Recall, the Humane AI Pin, the Rabbit R1, and seven more. Read it one way and it is a scorecard of flops. Read it another way and something stranger appears: the entries did not all die the same way. Sort the graveyard and it falls into three different kinds of disappearance — and the question is whether they add up to one pattern, or are three separate stories that happen to share a list.
The ones the platform ate
Relay was an AI-powered workflow automation tool, built as an alternative to Zapier. It let users automate email and task workflows with AI agents. It shut down entirely in August, as OpenAI, Google and other larger platforms built similar automation directly into their own tools. The five-year-old product, TechCrunch put it, “struggled to hold on to a reason to exist as a stand-alone product.” That is competitive displacement in its purest form: the feature survived, the product did not. The platform the user already had learned the trick the startup was selling.
Huxe went the same way. The AI audio app, built by former developers of Google’s NotebookLM, shut down in May 2026, partly as larger platforms began offering similar experiences to large existing audiences — Spotify has been expanding its own AI-powered audio tools.
The ones that were folded in
The second kind of disappearance looks different, because in a sense nothing died at all. OpenAI shut down several stand-alone apps and folded their features directly into ChatGPT. ChatGPT Atlas, a stand-alone AI-powered web browser, lasted less than a year before being discontinued on August 9, with its best features absorbed into ChatGPT. Operator, an AI agent that could browse the web and complete tasks on a user’s behalf, was folded in the same way. And image generation was built directly into ChatGPT, shrinking the role of DALL-E as its own separate destination for creating images.
This is internal consolidation, and it is the hardest kind of disappearance to judge. A folded-in feature is not a failed feature: Atlas’s browsing did not stop working, it stopped being a separate product. The graveyard records the death of the standalone app; the capability lived on inside the flagship — which a platform company would call a success.
The question the AI graveyard raises
So here is the question the list raises rather than settles: do competitive displacement, internal consolidation and a shrinking standalone role add up to one pattern — the platform absorbing the feature — or are they three separate stories?
There is a case for one pattern: in each of the five cases, the capability ended up living inside a bigger product rather than as a stand-alone one. But the case against is just as strong: consolidation can be a success, not a failure. Folding Atlas into ChatGPT may mean the capability reached more users, not that the product failed — while Relay lost out to competition and Huxe did so in part, and the flagship setbacks had different causes again. The five cases may not establish a common pattern at all.
The honest reading is that the three kinds of disappearance are confirmed and the single pattern is not — which is what makes the list worth reading carefully instead of treating it as a scorecard.
The flagship stumbles
Alongside the disappearances, the same list records the giants’ own flagship stumbles — and the list does not show they share a cause with the disappearances. They are the other half of the graveyard, and they failed in public.
OpenAI’s attempt to turn ChatGPT into a broader “super app” is the sharpest example. On July 9, the company updated the app to combine Chat, Codex and Work, renaming the familiar version “ChatGPT Classic.” Users quickly criticised the new interface as confusing and cluttered, and OpenAI rolled the change back soon after — restoring conversation history and Projects to the sidebar on July 17, eight days later. OpenAI president Greg Brockman later admitted the redesign was flawed: “It’s kind of a mess. We agree.”
When Apple unveiled Apple Intelligence in 2024, an improved Siri was one of the highlighted features — a Siri that could understand context and get tasks done across apps. Apple repeatedly pushed back the release, citing engineering issues and bugs, and the delay contributed to a $250 million settlement over claims regarding how Apple marketed the iPhone 16’s AI capabilities. The new Siri finally appeared in the iOS 27 beta in July 2026.
Microsoft’s Recall was announced at its 2024 Build developer conference as an AI-powered “photographic memory” for Windows PCs, periodically taking screenshots so users could search their own history. The announcement triggered an immediate privacy and security backlash, and Microsoft delayed the feature for nearly a year. Even after a redesign, the controversy has not gone away: a cybersecurity researcher recently built a tool that could extract and display data Recall had captured, reviving questions about whether Microsoft has fixed the underlying problem.
And then there is Sora, OpenAI’s video-sharing platform, which struggled with high operating costs and user-retention challenges and shut down in March 2026. The flagship bets fail too.
Character.ai: the $20 million opening bid
The graveyard’s most dramatic entry is Character.ai, which belongs to none of the three kinds neatly. Founded by former Google researchers Noam Shazeer and Daniel De Freitas, it became a buzzed-about app for chatting with customisable bots, and raised $150 million at a $1 billion valuation in 2023. Then, in 2024, Google paid a reported $2.7 billion to license its technology and bring its founders back into its orbit, leaving the startup to carry on without them.
The trouble piled up after that. Families sued, alleging the chatbots contributed to teen suicides and self-harm, and Disney sent a cease-and-desist letter over unauthorised use of its characters. The company filed for bankruptcy in early October — and Disney is now the opening bidder, offering $20 million for a license to some of its intellectual property. The app is still running, and the sale is expected to wrap up by late November or early December. A $1 billion valuation in 2023, a $2.7 billion Google licensing deal in 2024, and a $20 million opening bid in 2026 for a non-exclusive license to some of its intellectual property: three different kinds of figure, not one company’s falling price.
Put the list against the wider numbers and the scale comes into focus. S&P Global Market Intelligence found that 42% of companies abandoned the majority of their AI initiatives in 2025, more than double the 17% of the year before — and on average, organisations scrapped 46% of projects between proof of concept and broader adoption. That figure is context, not a verdict: it describes the whole population of corporate AI initiatives, most of which never reach the public, while the graveyard is a public sample.
TechCrunch frames its own list as “less as a scorecard of flops than as a record of what these attempts can (hopefully) teach the rest of the industry.” That is the right way to read it. The graveyard does not show that AI is failing. It shows three different ways AI products disappear, and leaves open whether those three ways are really one story. The flagship stumbles sit alongside them as a reminder that nobody is exempt, not even the companies doing the absorbing. The pattern is still a question. The disappearances are not.
