Stop planning for 2027: how OpenAI builds product 90 days at a time | Tara Sesha and Nan Yu (OpenAI) (Lenny & Friends Summit, San Francisco, 10 September 2026)

Tara Seshan, Nan Yu · 29m · youtube.com ↗

18 korrents from this recording

Tara Seshan, Nan Yu did not write this page.

Every claim below is a statement made in this recording, quoted word for word and linked to the second it was said, so you can hear it rather than take our word for it. The wording comes from the transcript published alongside the recording; the sentence above each quote is our reading of the claim, not their wording.

1:27 · watch on youtube.com↗ · Tara Seshan

Our reading
However much theorizing is done before shipping, nothing compares to the evidence of users trying a product and iterating from there.

Their wordsthere's a lot of like theorizing that one can do before you ship something but nothing compares to like the actual empirical evidence of seeing users try it, use it, and then iterating from there.
3 min later
4:41 · watch on youtube.com↗ · Tara Seshan

Our reading
AI products should probably aim at where the models will be in two to three months, neither anchored in the present nor futuristic and unusable.

Their wordsprobably most importantly in this era is this aiming for 2 to 3 months of like where the models will be rather than being overly anchored in the present or being overly like futuristic and unusable.
1 min later
5:15 · watch on youtube.com↗ · Nan Yu

Our reading
The biggest constraint on building AI products is people's ability to understand and absorb what they are given.

Their wordsThe the biggest constraint uh is honestly like people's understanding and ability to sort of absorb what you're what you're giving them, right?
3 min later
8:34 · watch on youtube.com↗ · Nan Yu

Our reading
The most successful agent designs have probably been the ones that map onto human nature.

Their wordsI I I think when designing this like the the most successful um efforts have really tried to map on to human nature, right?
9:03 · watch on youtube.com↗ · Nan Yu

Our reading
Agents probably need to be grouped into bundles of activity rather than managed one by one.

Their wordsI think there there's there's something in there where we need to group these things into sort of bundles of activity
1 min later
10:23 · watch on youtube.com↗ · Tara Seshan

Our reading
Whether to build one agent or many depends far more on the use case and its practical considerations than on philosophy.

Their wordsthere there are philosophical things one could have and opine on in this topic but to me there are just so many practical product considerations like the use case matters so much in deciding this.
2 min later
11:56 · watch on youtube.com↗ · Nan Yu

Our reading
Systems thinking and user empathy in agent design are the old product principles in disguise.

Their wordsI I I think those are very important but ultimately they're just the old principles in disguise.
12:11 · watch on youtube.com↗ · Tara Seshan

Our reading
Beyond user empathy and systems thinking, building agents needs a relentlessness to try things and keep iterating.

Their wordsthe third thing that you just really need in addition to user empathy and systems thinking is like a relentlessness to like try it out and keep iterating
7 min later
18:46 · watch on youtube.com↗ · Tara Seshan

Our reading
For a product person working with AI research, learning to write good evals is the most important thing.

Their wordslearning how to write good evals and writing evals as much as possible is like the most important thing
2 min later
20:22 · watch on youtube.com↗ · Tara Seshan

Our reading
Building exactly for today means being left behind; the ideal horizon is two to three months.

Their wordsand then at the same time if you build exactly for today you're going to get left behind. So it's like I constantly repeat this but it's like two to three months out.
1 min later
21:22 · watch on youtube.com↗ · Tara Seshan

Our reading
How far ahead to plan depends on the pace of the market a business is in; in some markets it may be appropriate to plan many months out.

Their wordsmaybe it is appropriate to plan you know many many months out but I think it involves understanding like the pace of the market you're in
1 min later
22:14 · watch on youtube.com↗ · Nan Yu

Our reading
Onboarding probably matters much more now than it did before.

Their wordsI think something that probably matters way more now than it's done before is is onboarding.
1 min later
22:58 · watch on youtube.com↗ · Tara Seshan

Our reading
When an agent acts semi-autonomously, a user must be able to predict at any point what a feature will probably do.

Their wordsnow if the you know agent is doing things semi-autonomously like you actually really need to make sure that at any given point a user can like predict what is probably going to happen with a feature or a primitive that you're introducing.
23:19 · watch on youtube.com↗ · Tara Seshan

Our reading
The usual PM craft still applies, but testing things empirically and moving faster matter more than ever.

Their wordsall the usual stuff still applies except the importance of testing things empirically and moving faster has only increased.
1 min later
24:47 · watch on youtube.com↗ · Tara Seshan

Our reading
Product people should be in direct contact with their users.

Their wordsof course you should be in direct contact with your users.
1 min later
25:48 · watch on youtube.com↗ · Nan Yu

Our reading
With agentic, natural-language products, a more direct and deeper relationship with users probably bears a lot of fruit.

Their wordsSo I think the the having a a more kind of direct uh deeper relationship bears a lot of fruit there, right?
1 min later
27:16 · watch on youtube.com↗ · Tara Seshan

Our reading
Voice is such an intuitive interface for so many people that it changes the game.

Their wordsthe existence of voice like changed the game like it is such an intuitive interface for so many people.
1 min later
27:51 · watch on youtube.com↗ · Nan Yu

Our reading
Products that in effect drive themselves will probably become much more important and prominent in the next year.

Their wordssomething that looks a lot like self-driving is probably going to be the the thing that really you know becomes much more important and prominent in the next year.