AI’s third era: the rise of persistent AI coworkers | Tara Seshan (OpenAI’s product lead)

Aug 30, 2026 Episode Page ↗
Overview

Tara Seshan, Product Lead for Codex and ChatGPT Work at OpenAI, discusses the evolving product management role in the AI era. She highlights the shift from "rowing" to "steering" with AI agents, the importance of ambition, and OpenAI's rapid iteration culture, emphasizing building for future model capabilities.

At a Glance
11 Insights
1h 21m Duration
15 Topics
7 Concepts

Deep Dive Analysis

OpenAI's Unique 'Founders-Led' Culture

The Evolving PM Role in the AI Era

Shift from 'Rowing' to 'Steering' with AI Agents

Ambition as a New Differentiator with AI

Building Products for Future AI Models (2-3 Months Out)

Understanding ChatGPT's Chat, Work, and Codex Modes

OpenAI's Rapid Shipping and Iteration Philosophy

The 'Vibe Shift' Around Codex and User Obsession

Blurring of Traditional Roles in AI-Driven Teams

Enduring Value of Human Brains in an AI Future

Practical AI Uses: Sites and Visualize Commands

Writing as Thinking vs. Writing as Reporting

Avoiding AI Brain Rot and Over-Reliance

Lessons Learned from Sutter Hill Ventures

Knowledge Work vs. Coding in the AI Context

Founders-Led Culture

A company culture, like OpenAI's, where everyone, especially in their product area, acts as a founder. There is limited top-down direction, and individuals are empowered to drive product-market fit akin to a startup founder.

Steering vs. Rowing

A model for the future of work where AI agents handle the 'rowing' (execution of tasks), and humans increasingly focus on 'steering' (setting direction, making opinionated calls, and providing feedback at higher levels of abstraction).

Ambition as a Differentiator

In an AI-powered world where easy and hard tasks become easier, the ability to be unreasonably ambitious and expand one's thinking about what's possible becomes the key differentiator for individuals and companies.

Building for 2-3 Months Ahead

A product strategy for AI, where development focuses on anticipating model capabilities two to three months in the future. Building for current models or models a year out is equally ineffective due to rapid advancement.

Writing as Thinking

A form of writing, such as drafting briefs or strategies, that is crucial for an individual to clarify and refine their own ideas. This process should be done by a human, not automated by AI.

Writing as Reporting

A form of writing, such as summarizing status updates or plans, that is primarily for conveying information. This type of writing is suitable for automation using AI models.

Product Marketing Fit

The concept that the way a product is talked about and marketed (its narrative and positioning) can and should be tested and refined even before the product is fully built, especially for B2B products.

?
What is unique about OpenAI's company culture?

OpenAI operates with a 'founders-led' culture where employees, especially in their product areas, act like founders, taking ownership and driving product-market fit with limited top-down direction.

?
How is the Product Manager (PM) role changing in the AI era?

The PM role is shifting from academic and theoretical work to being more prolific and empirical, focusing on defining core hypotheses, rapid testing, and elevating the ambitions of the team rather than extensive documentation or execution management.

?
How should product teams approach building for rapidly evolving AI models?

Teams should build for where AI models are expected to be in two to three months, as building for current capabilities or a year out will likely result in failure due to the exponential pace of model improvement.

?
What is the difference between ChatGPT's Chat, Work, and Codex modes?

Chat mode is for general conversations and search, Work mode (powered by Codex) provides agentic capabilities for knowledge work tasks like financial modeling without coding UI, and Codex mode offers the same power with a development-oriented UI for coding tasks.

?
Where will human brains continue to provide unique value in an AI-driven future?

Human brains will remain valuable for accountability (owning outcomes), expression (artistry and opinionation in creation), and fostering human connection (caring for colleagues, elevating ambitions, and collective enthusiasm).

?
How can one use AI without losing the ability to think critically?

To avoid 'brain rot,' one should consciously differentiate between 'writing as thinking' (which should be done by a human) and 'writing as reporting' (which can be automated by AI). The act of personally outlining, drafting, and refining ideas is crucial for maintaining sharp thinking.

?
What is the key difference between AI for coding and AI for knowledge work?

Coding tasks can be validated by output (e.g., tests), but knowledge work requires understanding the process, inputs, and reasoning. AI products for knowledge work need to focus more on collaboration, showing in-progress work, citations, and the model's chain of thought.

1. Build for Future AI Model Capabilities

When developing AI products, build for where models will be in two to three months, not where they are now or where they might be in a year, as both extremes lead to failure.

2. Prioritize Empirical and Prolific Work

In a fast-changing AI market, being prolific and empirical (testing quickly with users) is more effective than being academic or theoretical (writing long reasoning documents).

3. Focus on Core Hypothesis Definition

The most important aspect of the PM job is defining the core hypothesis for your product, understanding what will determine its success or failure, and designing fast, effective tests.

4. Elevate Others’ Ambitions

As a product manager, a key part of your role is to remind your team and colleagues of what’s possible with AI, pushing them to be more ambitious and consider 10x bigger scales and faster timelines.

5. Use AI to Expand Capabilities

Instead of merely automating rote tasks, leverage AI tools to significantly expand the range of things you are personally capable of doing, becoming more of an ‘auteur’ in your work.

6. Utilize AI for Dynamic Site Creation

Use AI to instantly build dynamic, shareable websites for various purposes, from presentational artifacts to internal tools or personal tracking, replacing traditional docs and slide decks.

7. Leverage AI for Data Visualization

Use the ‘/visualize’ command in AI tools (like Codex) to quickly generate insightful visualizations from your data, simplifying the process of presenting complex information.

8. Distinguish Writing as Thinking vs. Reporting

Automate ‘writing as reporting’ (summaries, status updates) with AI, but preserve ‘writing as thinking’ (briefs, strategy docs) as a human process to refine your ideas and critical thinking.

9. Iterate Docs with Feedback

When writing documents for collaboration or buy-in, aim for 70% completion before seeking feedback. This allows others to contribute and polish the ideas with you, making the final product stronger.

10. Start and End Writing Yourself

For ‘writing as thinking,’ begin and conclude the writing process yourself to maintain sharp thinking. AI can be used in the middle for research, data pulling, or pushing back on ideas.

11. Test Product Marketing Fit Early

Before committing to building a product, focus on refining its product marketing fit by pitching the narrative and positioning to many people to ensure the story is transformative and compelling.

You fail if you build for where the models are now. You fail if you build for where you think the models will be in a year. Both outcomes are equally wrong. Only way to build is two to three months.

Tara Seshan

Elevating others' ambitions or reminding them of what's possible here is a huge part of the product management role.

Tara Seshan

This is the worst the models will ever be.

Kevin Weill (quoted by Lenny Rachitsky)

Software is not like real estate. You don't like put money in and get value out. It is a little bit more like filmmaking, where you can put a lot of money into a film, but that doesn't guarantee that the film is successful or good.

Patrick Collison (quoted by Tara Seshan)

Done is better than perfect.

Tara Seshan

I have a hundred times the power and tools that he had making that film in my iPhone. And like, what's my excuse for not elevating my ambitions and making better stuff?

Tara Seshan

OpenAI Product Development Memes

Tara Seshan
  1. Ask: 'Is this maximally accelerated?' (Are we moving as fast as possible?)
  2. Ask: 'Are you mainlining it yet?' (Are you using this product all day, every day?)
  3. Ask: 'Are we being as ambitious as possible?' (Are we pushing the scope and scale of what we're trying to do?)

Toni Morrison's Principles on Work

Toni Morrison (quoted by Tara Seshan)
  1. Whatever the work is, do it well, not for the boss, but for yourself.
  2. You make the job; it doesn't make you.
  3. Your real life is with your family.
  4. You are not the work you do; you are the person that you are.
two to three months
Optimal time horizon for building AI products Refers to anticipating future model capabilities, as building for current or year-out models is ineffective.
70%
Document completion level before seeking feedback Allows collaborators to contribute and polish ideas, fostering a more interactive process.