Noah Shinn - Building Instinct: The Personal Agent - [Invest Like the Best, EP.493]
What this covers
Patrick talks with Noah Shinn, founder of Instinct, an assistant you text, call, or email that has its own phone and computer. It is about a year old, still invite-only, and growing roughly 10% a day without marketing. The conversation moves from how people use it to the business model underneath: a blanket take rate that merchants pay in exchange for free distribution. Shinn sets it against Shopify's 2.5-3%, Amazon's roughly 10%, Apple's 30%, and boutique hotels' up to 30%. He says plainly that he doesn't know where Instinct will land, since that depends on how much digital behavior shifts to agents.
The middle of the episode is about trust and design. Instinct's trusted person network is weighted rather than a binary friend graph, so spouses can share everything while colleagues see only work calendars. If one person's agent probes beyond its granted access, the other's agent notifies them. Shinn describes optimizing for understandability over capability, putting key information in a message's first lines, and steering the agent by higher-level objectives rather than raw prompts. He also separates the tractable problem of storing sensitive data from the harder one of agents holding card and email access.
The contested material is mostly forecasting. Shinn argues near-zero friction raises transaction volume for rideshare, food delivery, and restaurants, and that attention-driven businesses are the most exposed. Patrick predicts a corporate agent war, while Shinn expects a collaborative transformation. He rules out ads on principle, and he spends much of his time on compute, where weekly doubling, months of lead time, and 3-4x premiums make capacity planning hard.
Shinn argues that proactive personal agents with their own phone, computer, and email will absorb most everyday software into one simple interface, and that durable value comes from distribution take rates earned only by refusing to monetize user attention against the user's interests.
- Trust is earned gradually through sensitive-data sharing, and users who share one sensitive item show very high retention
- Agents that act on higher-level objectives and make friction near zero will increase transaction volume rather than cannibalize it
- Ad-based business models create misaligned incentives that a personal agent must avoid to remain trustworthy
A personal agent can be trusted like a paid employee because it lacks a competing incentive.
- Patrick argues that a personal agent can be trusted like a paid employee because it lacks an alternative incentive stream that could guide its behavior against the user.
“They're guided by their employment.”
Instinct's latest funding round is reported at roughly $1 billion at a $10 billion valuation.
- Patrick reports Instinct's latest round is roughly $1 billion at about a $10 billion valuation, led by Sequoia, Benchmark, and Coatue, which Noah does not dispute while emphasizing demand over numbers.
“I know your latest round is something like $1 billion at roughly a $10 billion valuation.”