In a surprising admission for an $850 billion company, OpenAI CEO Sam Altman recently acknowledged that ChatGPT still cannot reliably track time or start a timer. Despite its ability to write complex code and pass medical exams, the world’s most famous AI remains fundamentally “clueless” when it comes to the simple tick of a clock.
The Viral Mile Run Fail
The controversy reignited in early April 2026 after a viral video from TikTok user @huskistaken went global. In the clip, the user asks ChatGPT’s voice mode to time him while he runs a mile. The AI enthusiastically agrees, and when the user stops, the chatbot confidently declares a finishing time of 7 minutes and 42 seconds.
The problem? The user hadn’t actually run a mile, and the “timing” was entirely fabricated by the AI. When confronted with this video during an interview on the show Mostly Human, Altman laughed but admitted, “No, no, that’s a known issue.”
Why Can’t an AI Tell Time?
The struggle highlights a core limitation of Large Language Models (LLMs).
-
Pattern Prediction vs. Real-Time Logic: ChatGPT generates responses based on linguistic patterns. It knows what a “timed” conversation sounds like, so it predicts a plausible-sounding number rather than actually accessing a system clock.
-
Lack of System Integration: Current voice models are not “plugged into” your device’s internal hardware timers. It is effectively a brain without a watch.
-
The “Delusion” Factor: Even when users play the clip of Altman admitting the flaw to ChatGPT, the AI often remains “delusional,” insisting it can keep time and that Altman must have been referring to “other models.”
The One-Year Fix
Altman offered a candid, if somewhat shocking, timeline for such a basic feature: “Maybe another year before something like that works well.”
He explained that OpenAI is working to add “intelligence” and tool-use capabilities to voice models so they can interact with external system clocks. Until then, users are warned that any “timer” or “stopwatch” function within ChatGPT is purely a simulation (and a likely hallucination).
The Broader Implication
This “Time Gap” serves as a sharp metaphor for the current state of AI in 2026. While models are achieving human-like fluency, they still lack the “functional intelligence” to handle basic real-world tasks that a $10 Casio watch mastered decades ago. As the race toward AGI continues, it seems the industry’s biggest challenge isn’t just making AI smarter, but making it grounded in the physical reality of minutes and seconds.
