Two Hundred Requests an Hour, One Canceled Reservation
An Instinct user got his Resy account nuked after his agent polled a restaurant's API every 0.4 seconds at drop time. Half of X called it inevitable, the other half kept posting flight bookings and voice-note hotel bookings.
What people were saying
The day belonged to one canceled dinner. @jbahrdestefano asked Instinct for its own activity log after Resy deactivated his account and voided all future reservations, and the log was, in his words, holy hell: an availability sweep of one West Village restaurant every ten minutes around the clock since Friday, roughly 17 to 19 API calls per run, plus a 2.5 minute burst at the 9am drop polling every 0.4 seconds. Total, about 200 requests an hour. He conceded Resy was in the right. @pitdesi carried it to a wider audience, and @buccocapital turned it into the day's biggest post by simply noting that Amex banning tech bros for wrecking Resy is pretty funny (5,000 likes agreed).
Also circulating: @chrisbarber drew a 2x2 of the assistant space and got the usual "you forgot mine" replies, and @varunram wondered aloud why Poke sold when this playbook was clearly still on the table.
The praise
@royrubin05 spent 48 hours running three hotel bookings, four restaurant reservations, excursions and assorted purchases through WhatsApp voice notes, and argued the breakthrough was distribution, not the model. @daksh0x had it watch a flight price, book it with a saved card, apply every offer and pull the OTP out of email. @vedhsaka had it guess a discount code by trying 37,100 combinations, which is impressive and also, arguably, exhibit B in the Resy case. @Slokh used location sharing to fill gaps in a Korea itinerary, @apoorv03 handed out 1,000+ invites, and @ProbyShandilya predicted the infrastructure required to survive this demand will end up being the team's real legacy.
The pushback
@BoBrainerd called it the told-you-so moment for agent infrastructure: browser plus credentials plus no rate controls equals liability, and the user eats the consequences. @hnshah reframed the whole thing as the early Uber playbook, with $350M subsidizing behavior change and every correction quietly becoming training data. @chowtato reported ten minutes of hotel research returning stale prices that no longer exist, plus failed forms. @Bfaviero said his OpenClaw and Mac Mini setup now matches it and he owns his data.
Notable anecdotes
@BShen7 has been directing it through Meta glasses by shouting at an announce-messages workaround. @DratchCap built a wardrobe profile from a year of Gmail receipts and handed it over as a personal shopper. @dalibali2 suspects his EA is in trouble.
Interest over time
Updated at 12:01pm ET from 52 posts. Posts published later appear on the Posts tab.