Coatue Pt2. Open AI’s Kevin Weil Dives into All Things ChatGPT | BG2 w/ Bill Gurley & Brad Gerstner

Today's model is the worst you'll ever use

Weil emphasizes that current AI capabilities represent the baseline, with models improving every 2-3 months. This allows product teams to build features that 'kind of work' today, knowing they will perform excellently within months.

Measuring daily habits, not monthly visits

OpenAI tracks weekly active users (WAU) rather than monthly, with the goal of becoming a daily habit used multiple times per day. User engagement data shows time spent in the product is increasing as people discover broader utilities beyond writing.

Tight loop between research and product

Success comes from building products on the 'edge of what models can do,' creating a feedback cycle where user interactions inform research improvements. This leads to step-function capability jumps that solve problems computers have never solved before.

AI transitioning from reactive to proactive

The roadmap focuses on moving beyond question-answering to AI that anticipates needs and takes actions in the real world—scheduling appointments, booking travel, and handling tasks before users even request them.

Personalization through memory

ChatGPT is being trained to remember user preferences, family details, and personal context to provide tailored recommendations (e.g., family-friendly hotels) rather than generic responses, making the utility deeply specific to individual users.

Deep ecosystem integration

Future capabilities require connecting to users' complete digital lives—including contacts, email, calendars, and documents—allowing AI to operate across platforms with user permission and data portability.

The 'gentle singularity' is already here

Referencing Sam Altman's blog, Weil argues AI change is happening gradually without societal disruption. This makes premature regulation dangerous, comparing early AI restrictions to regulating the auto industry before seat belts were invented.

Iterative deployment over restriction

OpenAI advocates for releasing early and often to allow society to 'co-evolve' with AI, understanding capabilities and limitations through real-world experience rather than hypothetical fear or top-down mandates.

Open source and data ownership

OpenAI plans to launch an open-source model and maintains that users should control their own data, with the ability to take their information to whatever AI service they choose.

Path to $100B+ revenue is visible

Weil confirms the opportunity is larger than initially imagined, with a clear potential path to $100 billion in revenue and becoming one of the world's most valuable companies through both consumer and enterprise growth.

Operating speed exceeds previous unicorns

OpenAI moves faster than Twitter or Instagram did at similar stages, pushing decision-making down to teams and releasing frequently to discover model capabilities through real-world use rather than internal analysis.

Hardware reinvention on the horizon

With Jony Ive's involvement, OpenAI sees opportunities to reinvent physical devices and services, as AI requires redesigning every product category—creating opportunities for startups and fast-moving incumbents alike.