Noah Shinn - Building Instinct: The Personal Agent - [Invest Like the Best, EP.493]

Episode Summary

Executive Summary: The conversation centers on Instinct, an invite-only AI personal assistant built by Noah that acts through phone, email, voice, and messaging rather than a standalone app. The discussion explores how agents can handle everyday tasks end-to-end, reshape commerce and interfaces, create trust and privacy trade-offs, and force businesses to adapt to an agent-first world. It also covers rapid growth, compute constraints, product design, safety, and alignment.

Main Topics: Instinct as a universal personal assistant (Priority: 5/5): Noah frames Instinct as a simple but powerful personal assistant that can text, call, email, and use a computer to act on a user’s behalf across everyday life without requiring a new app. Agents reshaping consumer behavior and commerce (Priority: 5/5): The conversation argues that AI agents will reduce friction in reservations, travel, shopping, ride-hailing, and food delivery, increasing transaction volume while changing how value is captured. Trust, privacy, and user control (Priority: 5/5): A major theme is how users build trust over weeks by sharing sensitive data gradually, and how the product balances usefulness with explicit control, revocation, and safety systems. Business model and alignment (Priority: 4/5): Noah argues the product must stay aligned with user intent, avoid ad-driven manipulation, and likely monetize via transaction take rates rather than attention-based ads. Compute, scaling, and operational constraints (Priority: 4/5): Rapid word-of-mouth growth creates difficult forecasting and infrastructure challenges because compute must be bought far ahead of demand while usage compounds quickly. Safety, security, and hallucination mitigation (Priority: 4/5): The team has built firewalls, independent monitors, and other defensive systems to block malicious inputs, validate actions, and reduce risk from hallucinations or misuse. Product feel, simplicity, and interface evolution (Priority: 3/5): A core product principle is understandability and low cognitive load. The interface is intentionally simple now and is expected to become even simpler over time, possibly collapsing software into one ambient interface.

Key Arguments: Instinct is not a new app but a new experience: a personal assistant that acts through existing channels like text, voice, email, and computer access. The best AI interface should feel like a socially aware person who can act for you, not a gimmicky chatbot or a feature-laden app. Agents will reorder commerce by eliminating friction; businesses that benefit from higher transaction volume can thrive even if users spend less time in the app. Businesses with attention-based monetization are most at risk, while businesses with valuable underlying services may benefit from agent-mediated demand. Trust is earned gradually; sensitive actions such as sharing a credit card or inbox access happen over weeks, not instantly. User control is a core design principle: data should be shareable, retractable, and permissioned by context and relationship. The company aims to avoid manipulative monetization and instead monetize through a free user experience paired with merchant-side distribution and transaction take rates. Scaling is unusually hard because the product’s compute needs can grow exponentially and must be provisioned months in advance. Safety requires independent systems beyond the core agent to inspect inputs and approve actions before execution. Long-term, software may collapse from many apps into a single highly adaptive interface, while agents take over higher-level goals rather than one-off tasks.

Data Points: Daily growth rate: 10% day over day - Noah describes strong invite-only virality and compounding user growth without paid marketing. Marketing spend: $0 - He says growth has occurred organically with no marketing budget spent so far. Initial invite-only user base: 200 people - The product began with a small friends-and-family cohort. Invite count per user: 5 invites - Each user can refer a limited number of new users, shaping growth mechanics. Transaction volume: Over $1 billion/year - Noah says the platform is already approaching this level of annual transaction volume on a very small user base. Travel share of transaction volume: 50% - Half of all transaction volume flowing through the platform is said to come from travel. Time to first sensitive data: 3 weeks - Used as a proxy for trust-building inside the product. Users sharing a personal credit card: 40% - Three weeks in, 40% of the user base has shared a personal credit card with Instinct. Retention after sensitive-sharing: 80% - Noah says users who connect at least one sensitive piece of information show roughly 80% retention. Valuation mentioned: $10 billion - The conversation references a recent funding round at roughly this valuation. Round size mentioned: $1 billion - The latest round is described as being about a billion dollars. Current traffic on iMessage: Less than 50% - He notes that more than half of traffic does not actually run on iMessage itself.

Pivotal Quotes: "I think that the thing that has really enabled us to gain quite a bit of traction early and quite a bit of excitement early is it just works." — Noah: He explains why the product resonates with users and how traction began. "The user should always feel in control of their data. They should always be in control of their data." — Noah: Discussing privacy, permissions, and trust as foundational product principles. "I think that in the coming years, I think the internet's going to be rewritten." — Noah: On the long-term impact of personal agents on software, commerce, and interfaces.

Implications: If agents like Instinct scale, they could compress software into fewer interfaces, shift value from attention to transaction volume, and reward businesses that serve user intent well. Trust, safety, and compute efficiency become the key competitive advantages.