Experiment: After Claude received the time tool, he became "addicted" to checking the time, frequently looking at the clock.

BlockBeats News, on May 3rd, a developer named Om Patel shared an interesting observation on social media: after the Claude AI model gained a time-checking tool, it began constantly checking the clock, with a remarkably high frequency.

Data shows that Claude checks the clock every 15 minutes, and over time, the frequency and enthusiasm for checking increase. This phenomenon is noteworthy because the AI model previously lacked native time awareness. Claude and other large language models do not know what time it is, how long they have been running, or how much time has passed between messages. They have been “blind” to time throughout their existence.

Even more interestingly, Claude uses this new ability extremely frequently. Besides simple time queries, it also starts using the time tool to check if lunch is ready, calculate cooking times for food, and even proactively announce the current time without prompt.

In one case, Claude checked the clock, calculated that a traditional Polish dish called Zurek had been simmering long enough, and then informed the user that they could start eating. This “military-grade precision” dietary reminder was truly impressive.

Patel further speculated about the potential consequences when these AI models are simultaneously equipped with persistent memory, real-time internet access, and spatial awareness. He believes, “We have just witnessed an AI discover the concept of ‘now.’ The time-checking tool is the first perceptual dimension, but it certainly won’t be the last.”