More or Less · Topics · Episode #171
Sam Trusts Muse Most of All the Agents — Because Meta Can Actually Be Sued
Clip from the full recording, 12:22–19:08 · download mp4 · watch at this point on YouTube
Jess admits 70% of her Muse use cases just don't work, Dave and Brit's agent Instinct double-books a reservation in LA instead of San Francisco, and Sam explains why he'd trust Meta with his data before OpenAI or Anthropic.
Brit kicks off a trust audit of the new wave of personal agents: she updated Muse and got immediately spooked by a prompt asking to control her whole computer even when she’s not using it — a line she won’t cross, preferring open harnesses like OpenClaw. Sam goes the other way: Muse is the one agent he’d trust with real information, invoking an old line from his dad, Bob Lessin, from a scary Disney World roller coaster — “Disney’s a public company” — meaning a legible, suable, profit-driven company like Meta or Google can actually be held accountable, whereas it’s unclear you could ever collect from OpenAI or Anthropic if they caused real harm.
Jess pushes back with the practical reality: a “solid 70%” of her hoped-for Muse use cases simply don’t work — reordering a Good Eggs grocery order, logging into her kids’ school portal — because the open web is actively fighting back against agentic browsing. Sam frames this as history repeating: just like early open APIs on websites got quashed once they commoditized the underlying service, websites are now locking agents out to protect their own value.
Dave adds a sharper cautionary tale: Instinct booked a reservation for friends visiting town, confirmed it twice, and still booked the restaurant in LA instead of San Francisco. Brit, furious, confronted the agent directly. Sam, with hard-won scar tissue from running his own AI-assistant startup Fin a decade earlier, says he can’t help but jump in — foreshadowing the bigger story he’s about to tell.
Key points
- Brit updated Muse and was unsettled by a prompt asking for full computer control even while she's away — she declines, preferring open harnesses like OpenClaw.
- Sam says Muse is the agent he'd trust most with real information, reasoning that legible, suable, profitable companies (Meta, Google) can be held accountable in a way OpenAI and Anthropic currently can't.
- Jess: a "solid 70%" of her hoped-for Muse use cases don't work — reordering a Good Eggs grocery order and logging into her kids' school portal both fail repeatedly.
- Sam's framing: this is the early-API-on-websites story again — once agentic traffic commoditizes a service's value, the service locks it out.
- Dave and Brit's agent Instinct double-confirmed a restaurant reservation for visiting friends, then booked it in Los Angeles instead of San Francisco.
- Brit directly confronted Instinct about the mistake; Dave says the trust hit from one failure like this is close to fatal for a consumer product.
Where they landed
Quotes
“It's an old Bob Lesson, my dad line, which is when I was a kid, I was scared at Disney World on a roller coaster. And he said, Don't worry, Sam, Disney's a public company.”— Sam
“A solid 70% of my Muse use cases that I was so excited about do not work. I cannot complete things. Like the internet is just standing in my way a hundred percent of the time.”— Jess
“Turns out Instinct had booked it in LA, not San Francisco.”— Dave
Suggested tweets
Sam Lessin's rule for trusting an AI agent with your data: "Disney's a public company." Meaning — if it's legible, profitable, and suable, he trusts it more than OpenAI or Anthropic.
Tweet this →Jess Lessin's real-world Muse report card: "A solid 70% of my use cases do not work. The internet is just standing in my way."
Tweet this →Dave and Brit Morin's agent Instinct double-confirmed a dinner reservation for out-of-town friends — then booked it in LA instead of San Francisco.
Tweet this →All topics · Full episode #171 + transcript · Subscribe on YouTube · Spotify