BENGALURU: A bowl of fruit, a home camera and a curious coder have ignited a debate about surveillance, trust and the creeping reach of artificial intelligence into domestic life.
Pankaj Tanwar, a Bengaluru technology professional, says he fired his cook after an AI-powered monitoring system in his kitchen caught her repeatedly helping herself to fruit from his refrigerator. The system, which Tanwar jokingly calls his “AI roommate”, did more than watch. It analysed, narrated and judged.
Instead of sifting through hours of CCTV footage, Tanwar’s setup produces written reports describing kitchen activity in near real time. Connected to a camera perched above the refrigerator, the AI tool monitors movement, logs behaviour and sends alerts whenever something unusual happens.
my cook was stealing fruits from my fridge so i deployed my ai roommate in the kitchen and it caught her red handed 👩🍳
— Pankaj (@the2ndfloorguy) March 1, 2026
it monitors the kitchen when she's cooking and pings me the moment she takes anything. sends me weekly report. i also trained it to check if she washes her… pic.twitter.com/wnyzXttkVm
One log shared online detailed the cook’s actions step by step. “She came in at 7:12 pm. First thing she did was open the fridge. Not to check what to cook. She just stood there for a bit, took two apples and put them in her bag. Then started cooking.”
According to Tanwar, the system flagged two similar incidents in the same week. After reviewing the footage and AI summaries, he dismissed the cook.
But the digital tattletale was not just counting apples. The tool also kept score on hygiene and cleaning habits. In one update it noted that the cook had touched the dustbin lid, scratched her nose and then rolled chapatis without washing her hands. Another weekly report observed frequent pre-cooking fridge visits and missing items such as apples, a banana and a handful of blueberries.
The AI even commented on housekeeping standards, noting that while the kitchen slab was wiped daily, less visible areas such as the space behind the stove had been ignored for days.
Then came the sting in the tail. A reminder embedded in the report that the cook charged Rs 4,800 a month.
Tanwar says the system relies on a vision model that analyses video and converts activity into text updates. Face detection runs locally and blurs identities before the data is processed by the language model. He plans to move the entire setup to a local device so it can operate offline. Future upgrades may include gas-leak detection and idle-time tracking.
The experiment quickly ricocheted across social media, drawing equal parts applause, laughter and alarm.
Some users marvelled at the ingenuity. Others wondered whether such scrutiny was fair, or humane, for a domestic worker earning modest wages. One critic remarked that paying Rs 4,800 a month while spying on a cook hardly counted as technological heroism.
The episode has opened a broader question. If AI can audit a kitchen with forensic zeal, how far will people let it peer into everyday life?
For now, Tanwar’s system has proved one thing with brutal efficiency. In the age of algorithmic housemates, even a missing apple can leave a digital trail.