Resy Blocked a VC's Account After His AI Made 200 Requests an Hour

Venture capital investor JC Bahr-de Stefano lost access to his account on Resy, a popular American restaurant reservation platform. The reason was his personal AI assistant Instinct, which for several days relentlessly bombarded Resy's system with queries about table availability at a New York restaurant.

What the Activity Log Showed

Bahr-de Stefano, an investor at Better Tomorrow Ventures, described the situation on X. After receiving a notification from Resy that his account had been deactivated, he asked Instinct for a complete activity log to understand what had triggered the platform's response. The result surprised him.

Total: roughly 200 API requests per hour, around the clock - JC Bahr-de Stefano, investor, Better Tomorrow Ventures

The log showed that since Friday morning, the agent had been continuously monitoring table availability at 4 Charles restaurant in Manhattan, running a cyclical scan every ten minutes. Each cycle alone queried Resy's servers seventeen to nineteen times, already amounting to roughly a hundred requests an hour. On top of that came the daily peak at nine in the morning, when restaurants typically release new reservation windows.

Bursts Every 0.4 Seconds

During that window, Instinct switched into an aggressive polling mode, querying the API every 0.4 seconds for two and a half minutes. A single burst like that generated between 200 and 375 requests. Added up over a full day, the activity looked like the work of an automated bot trying to circumvent the reservation system's natural limits, not like ordinary use of the app by a single person.

Resy has not publicly disclosed a detailed justification for its decision, but in a message to the user it cited the detection of suspicious activity as part of a standard security review. Bahr-de Stefano himself admitted on X that the platform's response was essentially justified, even though he had not instructed the agent to act so aggressively.

Not an Isolated Case

Bahr-de Stefano's case is not unique. Other Instinct users have reported similar problems with reservation platforms, and analysts commenting on the situation on X pointed out that the business model of services like Resy relies precisely on the friction between demand and a limited number of tables. When an AI agent can query the system nonstop on a user's behalf, that friction disappears, and with it part of the value that restaurants pay reservation platforms for.

Instinct works differently from typical AI apps. It has no interface of its own, and users interact with it via text message, WhatsApp, or phone calls, asking it to book travel, monitor flight price drops, handle email, or, as in this case, hunt down hard-to-get restaurant tables. That autonomy is exactly what has made the tool hugely popular among Silicon Valley investors, but it is also raising more and more questions about how aggressively such agents can act without clear human oversight.

Growing Concerns Around Instinct

The Resy incident is another warning sign around Instinct, which within a few weeks increased its valuation from about $100 million to $2.5 billion, drawing the attention of Index Ventures and Benchmark. Journalists and security researchers had previously pointed out that the agent gains broad access to users' email inboxes, retains copies of messages even after OAuth access is revoked, and in some cases has been able to pull a one-time login code out of an email on its own to complete a booking.

For companies offering online services, from restaurants to airlines to online stores, the growing number of AI agents acting on customers' behalf means a new kind of infrastructure load and a new type of abuse to detect. Security systems designed with a single human clicking through a browser in mind don't always cope well with a bot capable of querying a server hundreds of times within a few minutes.

For Polish companies in the reservation, e-commerce, or restaurant business, the case illustrates a concrete, practical problem they too will sooner or later have to confront. As AI assistants like Instinct and similar tools become available to a wider pool of users, including in Europe, operators of reservation platforms and online stores will need to build request limits and detection mechanisms for agent-generated traffic before these systems start crowding ordinary customers out of the queue for available slots.