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

Instinct's core strategy is to be an "experience, not a tool." It avoids building a new application, instead integrating with users' existing habits like texting, calling, and emailing. This removes the friction of learning a new interface and meets users where they already are.

Instinct measures user trust by tracking the time it takes for them to share sensitive information, such as a credit card. Reaching a 40% share rate within three weeks is a powerful indicator of trust, which directly correlates to an 80% retention rate for those users.

An ad-based model misaligns an agent, incentivizing it to influence users against their own interests. Instinct is pursuing a "blanket transaction take rate," similar to Apple Pay. The agent remains free for the user, ensuring its actions are solely on their behalf, while merchants pay for the distribution.

When two Instinct users connect their agents for tasks like scheduling, they create a "Trusted Person Network." This is more advanced than a simple social graph because it involves configurable, weighted permissions (e.g., spouse has full access, colleague has calendar-only). This builds a defensible, nuanced network effect.

Instead of replacing services like Uber Eats, an agent drastically reduces purchase friction (e.g., proactive ordering). This could paradoxically increase the total transaction volume for the incumbent service, even while bypassing its attention-based app interface and associated ad revenue. This creates a complex co-opetition dynamic.

A 10% daily growth rate means compute needs double weekly. With hardware lead times of several months, a company like Instinct must predict demand far in advance. Buying for projected growth (e.g., 100M users) is a massive capital risk; under-buying means throttling a viral product. This is a unique scaling challenge.

Unlike on-demand tools like code generators, proactive agents like Instinct are constantly working in the background—waking up, scanning information, and anticipating user needs. This "always-on" nature means compute consumption is not tied to direct user interaction, creating a demand that is orders of magnitude larger than previous AI products.

Instinct prioritizes making the agent's behavior predictable and easy to understand over simply adding more features. This focus on the user's "feel"—reducing their cognitive load through subtle UX choices—is key to driving its off-the-charts engagement and viral growth.

Instead of users needing a separate app for every task, future agents will generate complex, purpose-built interfaces dynamically. Instinct's "files feature" can send a full web application for a trip itinerary. This suggests a future where a single agent interface replaces millions of static applications.

The true power of personal agents will be unlocked when they transition from simple task-doers ("order this") to long-term partners pursuing high-level user goals ("help me save X amount by year-end"). This involves continuous, proactive work over months, fundamentally changing the human-AI relationship.

To solve foundational issues like hallucinations, Instinct builds safety systems that are architecturally separate from the core agent. These "firewalls" and "monitors" act as independent watchdogs, scrutinizing the agent's thoughts and proposed actions before execution. This systematic approach is more robust than simply fine-tuning the model.

Businesses should analyze their revenue sources to prepare for an agent-rich world. Companies that primarily provide an underlying good or service will benefit from agents reducing transaction friction. Conversely, businesses reliant on monetizing user attention through in-app experiences (like social media) are highly vulnerable to being disintermediated.

Current reservation systems are inefficiently first-come, first-serve. An AI agent can communicate a user's context (e.g., "it's a 30th birthday") directly to the restaurant's system, allowing the restaurant to prioritize high-value events. This creates a more efficient, context-aware marketplace that benefits both sides.