The number of AI projects and ventures that get announced and then shelved keeps growing into 2026. In a compiled failure list, Apple's years-long delayed Siri overhaul sits side by side with OpenAI's hastily announced and quickly withdrawn "super app" attempt. Neither reached its goal, but for opposite reasons: one was driven by excessive caution, the other by excessive haste.

What happened?

The list brings together AI projects that either launched and failed to meet expectations, or were shelved before ever shipping. Apple announced years ago that it would add personalized AI features to Siri capable of understanding on-screen content, then repeatedly pushed those features further into the future. OpenAI, meanwhile, quickly unveiled plans to turn ChatGPT into a one-stop "super app" for shopping, payments, and daily tasks — only to be forced to walk it back shortly after. Both saw a clear, public retreat.

Why it matters

Apple's habit of delaying isn't new; the company has historically preferred not to ship a feature before it's fully baked. But in the AI race, that cautious instinct came at a steep price. As Google embedded Gemini into phones and Samsung advertised Galaxy AI, Apple felt compelled to announce Siri features that weren't ready yet — and then failed to deliver on its promise. The result was a years-long "coming soon" cycle that reinforced the perception that the company was falling behind in AI.

OpenAI's story unfolded at the opposite speed. The company grew ChatGPT through frequent, fast waves of features — from plugins to the GPT Store, from agents to Sora — a pace that was a natural response to competitive pressure. But the super-app move suddenly expanded the product's scope: the idea of bundling payments, shopping, and daily tasks into a single interface triggered trust and regulatory concerns all at once. Here, speed wasn't read as an advantage — it was read as a sign of losing control.

Who does this affect, and how?

The cost of these two strategies falls on different groups. The table below summarizes the key differences:

StrategyApple (Siri)OpenAI (Super App)
ApproachAnnounce early, keep delayingAnnounce fast, pull back fast
Main riskErosion of expectation managementLoss of trust and scope control
User impactDisappointment, but limited loyalty lossConfusion, growing skepticism toward the brand
Investor/market impactPerception of "falling behind in AI"Perception of "can't do everything at once"

Apple's problem is long-term reputational erosion; OpenAI's problem is a short-term trust shock. For enterprise customers and developers, the two send different signals: working with Apple carries "delay risk," while working with OpenAI carries "direction-change risk."

What's next?

Whether Siri actually reaches the level Apple promised at its next major software event will largely determine the company's AI credibility going forward. On OpenAI's side, the real question is whether the super-app ambition has been abandoned entirely, or will be reintroduced through smaller, incremental steps. Other players in the industry — particularly Meta and Amazon, which share similar "do-everything assistant" goals — will be watching both examples closely for lessons.

Both types of mistakes come at a cost, just in different ways. Delaying and failing to deliver slowly erodes trust, while announcing in haste and then backpedaling breaks it all at once. The winner of the 2026 AI race will likely be whoever finds the middle ground: a company that makes sure a feature actually works before announcing it, but doesn't let it drag on for years afterward.