The Future of AI:
Building at OpenAI

Kevin Weil

Chief Product Officer, OpenAI

2024

The Paradigm Shift

AI Models: Always Getting Better

"The AI models that you're using today is the worst AI model you will ever use for the rest of your life. Every two months, computers can do something they've never been able to do before and you need to completely think differently about what you're doing."

Every 60 days brings a step-function improvement in model capability
The pace of change requires complete rethinking of product strategy
Traditional tech companies work with fixed foundations, AI is the opposite
Building on shifting foundations demands a different product mindset

The Core Problem

Fuzzy Inputs, Fuzzy Outputs, New UX

Not deterministic: Same input ≠ same output (fundamentally different from traditional software)
Nuance matters: LLMs excel at subtle human language and context
Reliability varies: 60% vs 95% vs 99.5% accuracy requires completely different product designs
Evals are essential: Understanding your specific use case accuracy is the foundation of good AI products

The chat interface insightChat is the ideal medium for fuzzy, general-purpose AI because it allows the full expressive power of human language without artificial constraints. It's the lowest common denominator for talking to superintelligence.

Use-case specificityHigh-volume, prescribed use cases can be faster and more specific. But chat remains the catch-all baseline for anything that doesn't fit your vertical.

How OpenAI Works

Research + Product as One Team

Evolution: OpenAI began as pure research; ChatGPT was a "low-key research preview"
The problem: Separate research and product teams = just consuming your own API
The solution: Unified teams of research, engineering, and product iterating together on evals and use cases
The mandate: Every new product starts as a cross-functional research-eng-product collaboration

ImageGen launch moment

Internal usage exploded before launch. Team using it to ghibli-ify their photos, constantly visiting the shared gallery, a clear signal of product-market fit internally.

The org principle

OpenAI is intentionally PM-light (~25 PMs). High-agency PMs guide product-focused engineers rather than micromanage. Empowered teams move faster.

Hiring Mindset

What OpenAI PMs Need

High agency: See a problem, solve it. Don't wait for permission or clear direction
Comfort with massive ambiguity: No one will hand you a roadmap; you discover problems as you go
EQ and influence: Lead through relationships, not authority. Especially critical with research teams
Execution speed: Move quickly; iterate; learn from shipping, not planning

Why junior PMs struggleThe role is self-directed and ill-formed by design. Early-career PMs expecting clear scope or structure will find it frustrating. This role requires leadership maturity.

Contrarian Views

AI Product Myths Kevin Rejects

✗AI models will inevitably become commoditiesINSTEAD →✓ Better products come from integrating research, eng, and product to fine-tune models for specific use cases, creating defensible moats.

✗It's impossible to build AI products as fast as traditional softwareINSTEAD →✓ The pace at OpenAI proves that organizations can move at unprecedented speed when research and product work as one team.

✗AGI is a distant theoretical problem, not a near-term product realityINSTEAD →✓ Build your product roadmap assuming capabilities will leap forward every 60 days. Design for the models of tomorrow, not today.

✗Startups can't compete with OpenAI in the foundation model spaceINSTEAD →✓ Startups win by building on APIs and fine-tuning for specific verticals where OpenAI won't go deep. Find your wedge and own it.

Based on Kevin Weil's episode on Lenny's Podcast. All ideas on this page are from the episode.Watch on YouTube Follow @kevinweil on X

GO DEEPER IN THE EPISODE

05:16 OpenAI’s new image model 08:13 The role of chief product officer at OpenAI 11:42 His recruitment story and joining OpenAI 15:59 Working at OpenAI 18:44 The importance of evals in AI 24:40 Opportunities in the space 26:34 Shipping quickly and consistently 29:47 Product reviews and iterative deployment 32:53 Winning consumer awareness 36:03 Designing thoughtful experiences 40:56 Chat as an interface for AI 45:21 Collaboration between researchers and product teams 48:05 Hiring product managers at OpenAI 53:06 How OpenAI uses AI: vibe coding, AI prototyping, and more 01:04:34 Raising kids in an increasingly intelligent AI world 01:08:07 Why Kevin feels optimistic about our AI future 01:14:20 The AI model you're using today is the worst AI model you'll ever use 01:17:58 Reflections on the Libra project 01:21:51 Lightning round and final thoughts

Every sketch here began as a transcript.

Send us yours and we’ll draw it. $1 a transcript.

Sketch mine →

0:00