We drew up the GPT-5 Day starting lineup on the whiteboard. OpenAI on offense, timeline on D.

FROM TODAY: Live from the Ultradome for GPT-5 Day kickoff. Starting lineup is on the whiteboard. OpenAI on offense, timeline on D.

We asked @clairevo (Founder, ChatPRD) about the future of product management and how AI tools could reshape the role entirely. Vo predicts a significant shift, suggesting traditional PM roles won't last in their current form: “Product managers are either going to develop the hard skills to do design, go-to-market, and engineering, or engineers and designers are going to get tools... and be able to actually do the product management job.” “We're going to see a new type of role emerge, where people augment their specialist capability with AI. I don't think there's going to be product managers as they were five or ten years ago for much longer.”

We asked our @benhylak (Co-founder at Raindrop) about the challenges of automating software engineering and why AI model launches may feel less exciting over time. "We passed the Turing test, and now the next test is superintelligence, self-replicating, smarter than every single person, knows everything. The bar has really moved - we’ve shifted the goalposts 100%." "We might be 70–75% of the way through automating software engineering. The tough part is that last 25%, it’s the hardest and the least universal. It’s difficult to explain to people." "As we close that gap, launches will probably get less interesting. The last 5% of software engineering? It’s probably not going to be that exciting to most people, even to me."

.@ScottWu46 (Co-Founder & CEO @cognition) on being an application layer company in AI: "The core of being an applied lab is focusing on a very particular use case, delivering direct results." "We want to work on a lot of particular capabilities that apply to software engineering and run the whole stack from there (product, UX, bringing to market, selling)." "The base IQ is very much already there in the models, and you can see the raw problem-solving ability."

We asked @max_a_schwarzer (Researcher @OpenAI) about post-training improvements in GPT-5: "The thing that is most impressive about GPT-5 is how much getting all of the details right matters." "The team went off and did a deep dive of completely rebuilding the post-training stack, and it turns out that when you do that, you get results that would have taken another order of magnitude in pre-training improvements to produce."

.@harjotsgill (Co-Founder & CEO @CodeRabbitAI) on testing GPT-5 and using it within their product: "It's a generational leap. We've been using @OpenAI models for years, and our product is reasoning-heavy with PhD-style problems." "GPT-5 has the highest score on some of the hardest pull requests in our dataset, almost 2x better than others."

We asked @guygr (Co-Founder @augmentcode) about the limitations of AI in making design and architecture decisions. "All the decisions you make around how the code is structured still require close supervision. It's high stakes because if you let it go autonomously for long enough, it can really slow your project down." "Once you get to low tens of thousands of lines, the bad decisions made around the design and architecture start to show up and development slows down."

The fastest way to make a company “AI-native” isn’t to redraw the org chart, it’s to remove unnecessary handoffs. If AI allows one person to draft designs, copy, analysis, and code, the bottleneck shifts from “who can I get to help?” to “how fast can I iterate a V1?” @bradlightcap (COO of OpenAI) put it this way: “AI just accelerates people to get to a great V1 of everything. If you’re a high-agency individual… you’re no longer gated on people." "We find those two or three… AI superstars and champions, and then use them as touch points for the rest of the org.”

We spoke to the CEO of Vercel (@rauchg) about the future of vibe coding in enterprises: "The most surprising emergent thing I've learned is just how much demand there is in enterprises for vibe coding. It's bridging the gap between business and engineering." "Even the CEO of a successful company told me that it felt like asking for a feature from his own engineers was like petitioning the government." "Vibe coding solves that problem. All of the non-technical employees weren't able to ship software, and now they can."