A year ago, building an AI assistant that could manage your inbox, automate workflows or turn documents into podcasts was a good business opportunity.

Today, many of those capabilities are becoming standard features in products millions of people already use. And some of the companies that pioneered them are disappearing.

Relay built an AI automation platform, raised millions of dollars and attracted customers who wanted software that could do work on their behalf. It shut down in September. Huxe, founded by former Google NotebookLM developers, raised $4.6 million to turn written information into personalized audio. It closed in May. Yupp attracted roughly 1.3 million users and raised $33 million to help people compare AI models. It closed in March.

They're not isolated examples.

Across 2026, dozens of AI startups, applications and developer tools have shut down, discontinued products or been absorbed into larger companies. Some were small experiments. Others had serious funding, established customers and experienced founders.

We're witnessing AI's first extinction event.

Not because people are abandoning AI or the technology has stopped improving. AI is clearly here to stay, and it has become spectacularly good in 2026.

AI is advancing so quickly that entire categories of products are losing their original reason to exist. Features that once justified a standalone business can now be replicated by a frontier model, built into an existing application or assembled by a competent developer in a matter of hours.

It's a strange dynamic: The market for AI is expanding, but the window of opportunity for certain AI businesses is shrinking.

And I suspect we're only seeing the beginning.

The First Wave of Extinctions

Before getting into the list, an important distinction: Not every discontinued product represents a failed company. Some businesses ran out of money or couldn't find enough paying customers. Others were acquired, pivoted or deliberately retired products that no longer fit their strategies. Open-source projects can also stop development while remaining available for people to use.

Those are different outcomes, and they shouldn't be treated interchangeably. But collectively, they reveal how quickly the market is reorganizing.

Standalone Products Squeezed by Larger Platforms

One of AI's biggest promises is a personal assistant that understands your life, remembers what matters and handles work on your behalf. It's also becoming one of the most competitive categories in technology.

Relay | Shut down September 2026

Relay let users build AI-powered workflows across applications, including automations that incorporated human approval steps. The product addressed a real need: People wanted AI that could act rather than simply answer questions. But as larger companies integrated agent capabilities into their platforms, the competitive landscape changed. Relay closed in September, and founder Jacob Bank joined Google to work on Chrome.

The demand for automation obviously didn't disappear, but the opportunity to own that demand through an independent product became more difficult.

Ollie | Shutdown scheduled for November 6, 2026

Ollie was a privacy-focused AI household assistant designed to help families manage calendars, meals, appointments and other responsibilities. It raised $7.5 million in seed funding led by Khosla Ventures. But amid increasing competition and difficulties raising additional capital, the company announced it would shut down.

A useful reminder that even a differentiated product with credible investors isn't guaranteed enough distribution to survive.

Vellum Personal Assistant | Closed November 2026

Vellum developed a privacy-focused assistant built around security, model independence and user control. Its hosted service closed on November 7 after the company struggled to grow distribution quickly enough to support the business. The code remains available for self-hosting, but the commercial service is gone.

Mindy | Discontinued June 2026

Mindy tried to make AI delegation as simple as sending an email. Rather than opening a separate app, users could ask the assistant to research information, manage tasks and complete work. The team joined WorkWhile, and the standalone product was discontinued.

The underlying lesson across these examples isn't that personal AI assistants are a bad idea. I think assistants will become one of the most important interfaces in computing. But that doesn't mean there will be hundreds of sustainable standalone assistant companies. The more capable the general-purpose platforms become, the harder it may be to justify using a separate product for an overlapping set of tasks.

When AI Capabilities Become Commodities

This is where the extinction event becomes particularly interesting.

A company can create a genuinely useful AI product, attract customers and establish a market, only to discover that the underlying capability has become widely available elsewhere.

Huxe | Closed May 2026

Founded by former Google NotebookLM developers, Huxe transformed documents and written content into personalized audio experiences. It raised $4.6 million, but the category became increasingly crowded as larger AI platforms developed similar capabilities.

The product closed in May. The use case still makes sense. People want to consume information through audio. But as that functionality becomes more accessible, the value of a separate application becomes harder to defend.

Reforged Labs | Closed July 2026

Reforged Labs built AI tools for generating video advertisements for mobile games. It raised $3.9 million and reportedly worked with companies including Supercell, King and Ubisoft. Unlike a startup struggling to identify a customer problem, Reforged had a defined use case and real commercial relationships. But foundation models were improving rapidly, making it easier for customers to generate similar content themselves.

The capability became more accessible. The specialized business became less defensible.

There's a common thread here. The market is increasingly rewarding the ownership of workflows, customers and distribution rather than individual AI capabilities. A feature can be incredibly useful without being able to support a standalone business. And as the foundation models improve, that distinction matters more every month.

Products Absorbed, Consolidated or Rebuilt Elsewhere

It's possible for a product to disappear even when its underlying technology remains useful.

Businesses consolidate their offerings, teams move, and standalone tools get rebuilt inside something larger.

Notion Mail | Discontinued September 2026

Notion introduced its AI-powered email application in April 2025, promising a more intelligent way to organize and manage the inbox. About 17 months later, it discontinued the standalone product. Notion explained that users were increasingly managing email through agents without opening a dedicated inbox.

This is a subtle but important change. The work still needs to happen, but the interface where it happens is becoming less important.

Reve | Closed September 2026

Reve developed an AI image generator known for its prompt adherence and ability to render text. Following an investment from OpenAI and the movement of researchers to the company, Reve discontinued its standalone service. The underlying talent and technology retained value. The independent product disappeared.

GitHub Spark | Discontinued August 2026

GitHub Spark allowed developers to build and deploy applications using natural-language instructions. GitHub discontinued the standalone product as its AI development offerings evolved around Copilot.

Roo Code | Original extension archived May 2026

Roo Code developed an AI coding extension with specialized agents inside VS Code. The team shifted its focus toward a cloud-based agent product called Roomote, while the original extension was archived. The open-source code remained available.

OpenAI Sora | Consumer app closed April 2026

Sora generated enormous attention for its ability to create realistic AI video. But impressive generation capabilities didn't automatically translate into a sustainable consumer platform. OpenAI closed the standalone consumer application and subsequently discontinued its API.

ChatGPT Atlas | Closed August 2026

OpenAI introduced Atlas as an AI-native browser, bringing ChatGPT capabilities directly into web browsing. Less than a year later, the standalone browser was discontinued, with browsing functionality moving into other ChatGPT experiences.

Google Pixel Studio | Discontinued June 2026

Google retired its standalone AI image-generation application for Pixel devices and directed users toward Gemini.

Google Assistant | Replaced by Gemini on mobile beginning September 2026

Beginning in September 2026, Google transitioned mobile devices and paired accessories from Google Assistant to Gemini. Some other Assistant-enabled device categories remained on Google Assistant.

Microsoft Copilot for Gaming | Wind-down announced May 2026

Microsoft moved away from its dedicated gaming assistant initiative as its broader Xbox and AI strategies evolved.

OpenAI DALL·E API | Retired May 2026

OpenAI discontinued the older DALL·E API as newer image-generation models became available.

I wouldn't call most of these failures in the traditional sense. Large technology companies routinely consolidate products and replace older technologies. But they demonstrate something important: Even the companies developing frontier AI models are still figuring out which interfaces and experiences are worth maintaining.

The capability might survive while the product disappears.

The Businesses That Couldn't Make the Economics Work

Not every closure is a story about commoditization or platform competition.

Some companies simply couldn't build sustainable businesses.

Yupp | Closed March 2026

Yupp let users compare responses from competing AI models and provide feedback about their quality. It raised $33 million and attracted roughly 1.3 million users. Despite that traction, the founders concluded the company didn't have a sustainable path forward.

LOVO AI | Filed for Chapter 7 bankruptcy in May 2026

LOVO developed AI voice-generation technology but faced litigation involving allegations of unauthorized voice cloning. The company filed for Chapter 7 liquidation.

Character.AI | Filed for Chapter 11 in October 2026

Character.AI attracted significant investment and user interest through customizable AI companions. But legal disputes, safety concerns and operational pressures contributed to its Chapter 11 filing. The service continues operating during restructuring, so this is not a completed shutdown.

Soul Machines | Entered receivership and was sold in 2026

Soul Machines built AI-powered digital humans and avatars. The company entered receivership and was subsequently sold to AppDirect. Its technology continued under new ownership, but the original independent business did not.

And while they predate the latest wave, two earlier failures deserve mention.

Humane raised roughly $230 million for its AI Pin, a wearable device intended to reduce people's reliance on smartphones. It shut down in February 2025 after failing to deliver a compelling enough product experience.

Builder.ai raised more than $500 million around the promise of AI-powered software development before collapsing in May 2025 amid financial problems and questions about its business representations.

These are different stories with different underlying causes. But they represent an uncomfortable reality: Raising capital, attracting users and demonstrating impressive technology don't guarantee a durable business.

Why This Extinction Event Is Happening

Looking across these companies, the shutdowns are obviously not all the same. Some ran out of money, some were acquired, some products were folded into larger platforms, and others simply became unnecessary as the technology around them improved. But taken together, I think four forces explain why so many AI businesses are suddenly under pressure, and which ones may be vulnerable next.

1. Intelligence is becoming a commodity

logo

Available to paid subscribers

Get the Frontier Pass to access Forward Future's private Slack community, member-only videos and content, insider interviews, and monthly AMAs with Matthew Berman.

Join The Frontier Today

Already a subscriber? Sign in

Reply

Avatar

or to participate