The AI graveyard: a running list of projects and startups that didn’t make it

Introduction

The rapid rise of artificial intelligence has sparked a flood of ambitious products, startups, and corporate initiatives. Yet not every venture survives the intense competition, technical hurdles, or shifting market demand. This article surveys a selection of high‑profile AI efforts that have shut down, been absorbed, or dramatically missed expectations, offering a snapshot of the growing "AI graveyard" and the insights it provides for founders, investors, and engineers.

Stand‑alone tools that couldn’t compete with platform giants

• Relay – Launched as an AI‑driven alternative to Zapier, Relay let users automate email and task workflows with intelligent agents. After five years, the product was eclipsed by native automation features in OpenAI, Google, and Microsoft tools, leading to its complete shutdown in August 2026.
• Character.ai – Once a popular chatbot platform, Character.ai filed for bankruptcy just five years after its debut, unable to sustain a user base as larger models became freely accessible through major providers.
• Notion Mail – Introduced in April 2025 to bring AI‑enhanced email organization, Notion Mail saw users gravitate toward external AI agents for inbox management. The service is slated for closure in September 2025.
• Huxe – An AI audio app that transformed text into podcast‑style narration, Huxe folded in May 2026 after platforms like Spotify rolled out comparable features to massive audiences.
• Yupp – A playground for comparing responses from hundreds of AI models, Yupp attracted a niche community but never achieved sufficient product‑market fit, leading to its shutdown in March 2026.

Corporate experiments that never left the lab

• OpenAI’s super‑app experiment – In July 2026 OpenAI merged Chat, Codex, and Work modes into a single interface, only to roll back after user backlash over complexity. Similar attempts to spin off standalone apps—ChatGPT Atlas, Operator, and the video‑sharing platform Sora—were short‑lived, with features ultimately folded back into the core ChatGPT experience.
• Apple Intelligence & Siri – Apple promised a context‑aware Siri in 2024, but repeated delays and engineering setbacks postponed its release until the iOS 27 beta. The protracted rollout contributed to a $250 million settlement over alleged mis‑marketing of AI capabilities.
• Microsoft Recall – Announced at Build 2024 as an AI‑powered “photographic memory” for Windows, Recall faced immediate privacy concerns. After a year‑long delay and redesign, security researchers still demonstrated data‑extraction vulnerabilities, keeping the feature under scrutiny.
• Figgs AI – This storytelling platform let users create custom AI characters and attracted over a million users in its first year. However, the cost of keeping the service free proved unsustainable, prompting its closure in 2024.

Hardware ventures that stumbled

• Humane AI Pin – Backed by $230 million in funding, the wearable aimed to deliver AI without a smartphone. Performance issues and a battery‑fire warning led to the business shutting down in February 2025, with assets later sold to HP.
• Rabbit R1 – Marketed as an AI companion at CES 2024, the device sold 100,000 units but suffered from unreliable performance and limited integrations. The company continues to iterate, now focusing on a “Project Cyberdeck” portable coding device.

What the graveyard teaches us

1. Differentiation matters – Stand‑alone AI tools must offer capabilities that larger platforms cannot easily replicate. When giants embed similar features natively, niche products lose relevance quickly.
2. User experience is king – OpenAI’s super‑app redesign illustrates how even well‑intentioned UI changes can alienate a loyal user base if they add friction.
3. Privacy and trust are non‑negotiable – Microsoft’s Recall controversy underscores the lasting reputational damage that privacy missteps can cause, even after technical fixes.
4. Hardware risk is amplified – Devices like the Humane AI Pin highlight the challenges of delivering cutting‑edge AI performance in constrained form factors, especially when safety concerns arise.
5. Sustainable economics are essential – Many AI services, from Notion Mail to Figgs AI, faltered because the cost of providing free or heavily subsidized AI outweighed revenue potential.

Closing thoughts

The AI graveyard is not merely a list of failures; it is a repository of hard‑earned lessons. As the field matures, founders and investors can look to these cases to gauge market saturation, prioritize privacy, and design products that truly complement—rather than compete with—the ecosystems built by the industry’s biggest players. By studying what didn’t work, the next generation of AI innovators can build more resilient, user‑focused solutions that stand the test of time.