Instinct Redefines Personal AI: 10% Daily Growth

Utility and functionality take precedence over complex app-based interfaces in the Instinct design, which operates through SMS and calls to replace standalone applications, as the founder notes, "We're building a personal assistant."

The founder emphasizes that Instinct prioritizes utility and functionality over complex app-based interfaces, aiming for a system that "just works."

The assistant interacts primarily via SMS and calls, removing the need for users to download or manage a traditional standalone application.

This design choice allows the AI to integrate seamlessly into daily routines without adding to app fatigue.

This unique architecture enables network effects, as agents can efficiently coordinate complex tasks such as carpooling arrangements or shared event planning.

The 'trusted person network' ensures that interactions remain private and controlled, limiting agent access only to approved contacts.

The founder believes that the internet will be rewritten in the coming years, fundamentally changing how most people interact with software.

The goal is to shift from manual site navigation to a more efficient, agent-based interaction model where AI performs tasks proactively.

Agents could, for example, check for reservations every five seconds across all platforms simultaneously, automating previously tedious processes.

Traditional reservation systems operate on a first-come, first-serve basis, often overlooking the significance of a user's occasion.

Instinct's agents can communicate specific contexts, such as a 30th birthday celebration, directly to the restaurant's system.

This enables restaurants to prioritize special events over routine table requests, fostering a more personalized service.

The system creates a better match between a user's intent and a business's capacity to cater to specific needs, enhancing the overall experience.

A 40% share of the user base provides credit card details within three weeks, an act of voluntary data-sharing that correlates with an 80% retention rate as users maintain control over their information.

Sharing sensitive information, such as credit card details, serves as a proxy for the level of trust a user places in the AI agent.

Users who connect at least one piece of sensitive information to Instinct demonstrate an impressive 80% retention rate.

Data sharing remains voluntary and reversible, ensuring users can revoke access at any time.

Instinct's philosophy is to avoid business models where user data is the product, rejecting practices common among major platforms like Google, TikTok, and Instagram.

The company aims to prevent its AI from influencing user behavior in ways that do not align with the user's explicit desires.

Prioritizing user goals over simply completing tasks without consent is central to maintaining trust and ethical interaction.

This focus on higher-level objectives ensures that the AI acts in the user's best interest, fostering a more reliable and trustworthy assistant.

Instinct evaluates its take rate strategy by comparing it to various industry benchmarks, such as Shopify's 2.5-3%, Amazon's 10%+, and Apple's 30%.

The company's goal is not merely to skim small percentages, like payment rails, but to provide substantial value that justifies a higher fee.

This strategy suggests a potential to capture higher fees, especially in high-value industries like travel, by optimizing overall value.

The founder aims to position Instinct at the upper end of the value scale by demonstrating significant user benefits.

While conflict may arise if agents reduce friction to near zero, proactive agent behavior could ultimately increase total transaction volume for services like Uber.

The hypothesis is that reducing friction for essential goods and services will lead to an overall increase in demand for those services.

This shift signifies a change in how existing services interact with users, making them more accessible and efficient.

The industry is transitioning from high-friction app usage to low-friction, agent-led interactions, as observed by the Instinct founder.

Traditional business models monetize user attention and time spent on apps, whereas new models focus on monetizing transaction volume.

Instinct aims to liberate users from forced infinite scrolling and other engagement tactics designed to maximize time on application.

This shift redefines the business model towards value-added utility, emphasizing efficient task completion rather than prolonged user engagement.

Instinct shifts its focus from raw 'AI capabilities' to 'user understandability,' emphasizing how well users comprehend the AI's actions.

The goal is to reduce cognitive load, including carefully formatting text for optimal scanning and quick comprehension.

Understandability is identified as a primary driver of Instinct's high engagement and viral growth, indicating its importance for user adoption.

The "product feel" represents a new "muscle" that AI developers need to cultivate, focusing on intuitive user experience rather than just technical prowess.

This growth is entirely driven by an invite-only referral system, where each user receives five valuable invites to share.

The company has spent $0 on marketing, demonstrating that its growth is purely word-of-mouth and product-driven.

The invite-only model also helps manage compute load, ensuring high quality of service for existing users as the platform scales.

Instinct achieves performance comparable to frontier models like Opus 5 while maintaining a significantly low-cost structure for its AI inference.

The company avoids "blanket costs" associated with generic APIs by specifically optimizing deployments for various types of batch work.

Batch tasks are engineered to be three to eight times more efficient on the same underlying compute resources.

This optimization drastically reduces serving costs, supporting the goal of keeping the product free for users.

Instinct implemented an early invite-only access program from the start as a safety mechanism, preventing scaling issues before infrastructure safeguards were fully in place.

The company opted to build entirely new systems to address initial mistakes, rather than simply applying patches to existing solutions.

A fundamental principle guiding this approach is that users must always retain full control over their data, ensuring privacy and security.

This proactive strategy allowed Instinct to harden its firewalls and ensure data integrity from its earliest stages.

Instinct employs a 'watchdog' layer to scrutinize tool calls before they are executed, ensuring safety and preventing unintended actions.

This validation layer is intentionally decoupled from the main agent's incentive system to ensure neutral and unbiased oversight.

Safety measures focus on proactively identifying and catching sampling errors in the underlying AI models, preventing hallucinations and incorrect outputs.

By separating validation from the agent's core function, Instinct maintains a robust defense against potential AI misbehavior.

The current software development cycle, often measured in months, is too slow to incorporate rapid consumer feedback, leading to inefficiencies.

Users increasingly suffer from "application fatigue" due to the sheer volume and specialization of individual apps they must manage.

Instinct aims to consolidate software functionality into a single, easy-to-use interface that generates features dynamically.

Complex tasks like trip planning or project management will be handled through on-the-fly file generation, eliminating the need to build individual applications.

Instinct recently completed a funding round led by prominent investors such as Sequoia, Benchmark, and Coatue, signaling strong market confidence.

Building a product for billions of people at an affordable cost necessitates significant initial capital investment in compute infrastructure.

The founder views venture capital as a crucial tool for taking calculated risks and avoiding "local optima," such as restrictive subscription models.

Instinct's platform is not limited to a single messaging platform, with over 50% of its traffic occurring outside of iMessage, indicating broad accessibility.