@sytaylor - Instinct Shows Consumer AI

Open original

@sytaylor on X The author, Stephen Taylor, argues that Instinct, a personal AI agent you control by text or voice note with no app to learn, is the first agent that feels like a finished product rather than a hobbyist project. He thinks this shifts consumer interaction from people using apps and websites to people sending their own agents to use businesses' systems, and that companies who block or ignore those agents will lose out.

  • Instinct remembers personal context and completes real tasks like booking a pub lunch, moving the booking, booking cycling lessons, and negotiating a Comcast bill, which the author says makes hours of admin disappear.
  • The hard part of personal agents has been building memory and context, and Instinct packages that into a consumer experience, while earlier setups required a separate Mac Mini and weekends of configuration.
  • Businesses need to expose reliable actions to customers' agents, using approaches like WebMCP, and Stripe's Link wallet lets an agent pay with a one-time card authorized per purchase without seeing the user's card details.
  • Websites that block bots risk treating legitimate customer agents as crawlers, so the author advises companies to test trusted agents on their own services now.

Instinct is a personal AI agent you control by text or voice note, with no app to install. The author describes it as the first consumer agent that feels like a finished product rather than a weekend hobby project. What makes it work, in his telling, is memory. He asked it to book Sunday lunch at a local pub, and when he later said "move it back a day" and then "move it back 90 minutes," it understood which booking he meant. It also handled a clunky web form, booked cycling lessons for his daughter, and sorted his calendar better than a human assistant or ChatGPT with calendar access. Sheel Mohnot's five-day test is the other data point: a podiatrist appointment, a Comcast bill cut from $100 to $60, a DMV booking, and two United reservations linked together. Each task is minor, but together they strip out hours of admin. The bigger argument is about channels. Personal agents until now required a dedicated Mac Mini, OpenClaw or Hermes, and several weekends of setup, and the hard part was never model intelligence but building persistent context. Instinct packages that context into a consumer product. The author expects agents to change business interfaces more than mobile did, because the customer stops operating the website or app and instead sends software to do it. That leaves businesses with an awkward problem. Many still block bots, which the author says turns a customer's representative into a supposed crawler. He points to Salesforce's Claudeforce, which puts Salesforce data and workflows inside Claude, as an example of a company accepting that the customer's agent, not its own, should be the interface. He also flags WebMCP, a draft standard that lets sites expose actions like "change booking" so agents don't have to guess at buttons. The author is clear that neither Instinct nor Grok Bot is the consumer breakthrough yet. Instinct still breaks, and Grok Bot's users keep hitting usage limits. Callable actions are only the first layer, though. He says a fuller agent-acceptance stack also needs identity, authority, payment, liability and recourse: how a business knows which human stands behind an agent, what that agent may do, how it pays, and who sorts things out when it gets something wrong.

Ask about this article