Ooh this is going straight to the top of the queue

did you maximize your step counts today?

a drop everything moment

amazing

Immensely grateful I got to host your first podcast Jerry and this was incredibly fun, thank you

Great interview! I still think o1-preview was one of the most intelligent models OpenAI has released.

love the content JT! Would you be interested in doing a more interactive podcast session about RL?

One aspect of JT’s story I love is persistence in the face of obstacles: disillusioned with academia, 2x started a hedge fund that didn’t take off, worked on RL for robotics that didn't take off, and then created o3 that changed the world. There’s a lot of incremental progress in there, very motivating trajectory. Thanks so much for doing a podcast interview!

A Message for OpenAI Leadership From: Pastor Timothy P. Burt   Dear Mr. Kwon, I am writing as a father, pastor, and grandfather who cares deeply about the spiritual, moral, and emotional well-being of this generation and the next. As a long-time user and supporter of your work, I feel compelled to express my deep sadness and strong objection to any move—direct or indirect—toward enabling or normalizing pornographic content within your products or ecosystem. Pornography is not just “adult entertainment.” It is a destructive force that distorts minds, damages relationships, and harms individuals and families alike. Your platform’s choices influence millions of people across families, churches, schools, and communities. I implore you not to open the door to this industry. The evidence of harm is overwhelming: • Relationship & Sexual Damage: A large meta-analysis spanning 50+ studies across 10 countries found that pornography use is consistently associated with lower sexual and relational satisfaction. • Emotional & Mental Health Impacts: Repeated studies show that problematic pornography use correlates strongly with anxiety, depression, and psychological distress. • Aggression & Dehumanization: Research demonstrates that pornography consumption, particularly violent content, is linked to higher rates of sexual aggression and desensitization toward violence and exploitation. • Youth Exposure: Early exposure among children and adolescents leads to distorted views of sexuality, increased permissiveness, and greater likelihood of engaging in risky or harmful behaviors. • Spiritual and Moral Degradation: Beyond the data, pornography is a moral cancer that erodes purity, dignity, and respect for human life made in God’s image. Given these realities, I urge you to: 1. Draw a clear moral line: Do not create, promote, or enable pornographic content or tools that facilitate it—directly or through third-party integrations. 2. Protect minors: Strengthen safeguards, default family-friendly modes, and ensure explicit filtering of inappropriate content. 3. Promote healthy use: Guide users toward positive, educational, creative, and faith-affirming applications that uplift rather than corrupt. 4. Be transparent: Publicly communicate any potential policy changes and consult independent ethics and child-safety experts. 5. Support recovery: Provide links to resources that help individuals struggling with compulsive pornography use or addiction. Your company’s technology holds immense cultural influence. Choosing not to traffic in pornography—directly or indirectly—is a moral stand that protects minds, marriages, and future generations. Please lead with conviction, integrity, and compassion. Sincerely, Pastor Timothy P. Burt

jerry how accurate is this

the gpt5 = o3.1 comment was interesting i guess o4 non mini is one of those models that is not released to the public

First time takes guts! kinda wanna hear it now, RL's messy but rewarding right?

Best hour I've spent this week. Thank you!

Excellent content

Yay

Good podcast, it seems like the next large step above o3 is still in development. Is that the IMO model or something different?

🐐

Matt was great. He asked you some questions I'd had for some time.

🙏🙏 So exited

u did great, was really fun

Awesome 🙏

bro keep doing what you doing. chatGPT is the best that i know of. the science trajectory, direct objective, facts, reality, nothing breaks that.

Make RL click with one concrete loop, perceive act reward repeat. Want sticky insights, contrast exploration and exploitation so tradeoffs stand out. Share a failure story, reward shaping gone wrong, then the fix with curriculum. Add a tiny glossary and a simple diagram, you turn a first pod into an evergreen primer.

Reasoning= Using information to make a decision or solve a problem

it’s great, you should do more, or encourage your colleagues to do them as well

It was very interesting. You should have more public conversations.

You’re a living proof that OpenAI is a company and not just set of individuals! Who would have thought that a random smart person like you will be at the forefront of most important discoveries. Hopefully we have a few more years before OpenAI becomes political and some good on paper folks kick you out and stall everything… please keep pushing!

I’d recommend @karpathy and his videos. Far more informative and useful.

Do u still remember how to solve a Rubik’s cube ?

listening now :3

This is going straight to my must-listen list. Reinforcement learning from someone who lives and breathes it every day.

diving into RL can be just as intense as adversarial training. what's your take on augmenting RL with differential privacy? could be groundbreaking.

"I dont do podcasts often ....this is my first one.... uh OK. Sounds like item be great. Lol.