The AI you used this morning is still seeing the world through a keyboard and a screen. Nathan Xu @Nathan_Xu_Gao from @PLAUDAI on why conversation is the rawest form of intelligence and what that means for how we build AI hardware.

AI-native businesses don't just use software differently. They transact differently, sell differently, and scale differently. Abhi Tiwari from @stripe on what that means for how companies build and grow. 🎥:

AI agents are only as good as what they can see. Ariel Shulman from Bright Data on why the web was built for humans, and what has to change before agents can actually navigate it. 🎥:

Today's AI is trained on human text. It's fluent in what people have written but blind to the world around it. At SuperAI Singapore, Robbie @Schingler of @Planet explained why the next foundation model should be trained on the Earth itself:

Predicting a protein's structure once took years of lab work. AlphaFold does it in minutes. DeepMind calls this a root node problem – one that unlocks a whole umbrella of others beneath it. Watch Ramine Tinati of @GoogleDeepMind at SuperAI:

Once anyone can fake a flawless message at scale, you can no longer tell a real stranger from a synthetic one. @balajis took to the SuperAI stage to explain why that pushes us into verified tribes, and forces a new way of building apps:

Most people treat ChatGPT the way a 1979 lawyer treated the spreadsheet: very cool, but no idea what to do with it. At SuperAI Singapore, Benedict Evans broke down why the value sits above the models, in the boundless questions they finally let you ask. Watch here:

The more AI reasons, the more compute each answer demands and the slower it runs. Past a point, latency caps how much reasoning you can actually use. That's why Andy Hock of @cerebras calls speed a form of intelligence. Catch the full session:

2,000 applications. 200 builders. 36 hours. 5 finalists. VeriClaims AI. BUNKERGUARD. Bloom. Willow. Built, shipped, demoed. That's NEXT. Powered by @awscloud, @vercel, @stripe, @ExaAILabs @RazerAI

What makes an AI model "frontier" in 2026? And who actually captures the value? Hemant Mohapatra (@MohapatraHemant) took on those questions at SuperAI Singapore (@superai_conf), on a panel with Geoff Soon (Mistral) and Cherie Shi (MiniMax), moderated by Zixuan Li ( Two arguments he made on stage that builders should take note of: 1.⁠ ⁠We're still in AI's extractive phase. •⁠ ⁠Value sits with whoever extracts intelligence best. So today's frontier is benchmarks and tokens per dollar per watt. •⁠ ⁠But the phase will shift to distribution. Then the frontier splits: •⁠ ⁠Problems of scale (coding, reasoning) → won on cost and usability. •⁠ ⁠Problems of scope (cancer, materials science, robotics) → fractal. They get harder the deeper you go. Winning there takes testing null hypotheses, RL, creating de novo knowledge. (Hemant has written about this extractive vs. distributive framework in depth — link below.) 2.⁠ ⁠Open weights exist to commoditize the intelligence layer. That's a good thing. •⁠ ⁠If a handful of companies own intelligence, the gains pool the way oil wealth does. •⁠ ⁠Commoditize the base layer and value moves up to applications, where everyone can pay to play. •⁠ ⁠The trap for open-source labs: positioning as SOTA model companies locks them into selling tokens at market price, without owning the infra to serve them. •⁠ ⁠Owning a pond doesn't make you a water business. The money is in cleaning, piping, and delivering the water. Thanks to the SuperAI team, as well as Alex Fiskum (@AlexFiskum), for hosting. @MistralAI @MiniMax_AI @ZixuanLi_ @Zai_org

The 100x company isn't about headcount. It's about AI as the intelligence layer connecting every workflow, decision, and customer interaction. Hyunjin Kim (@INSEAD), Haydn Sallmann (@googlecloud), @WilliamBryk (@ExaAILabs ), Ang Li (@SimularAI), and @bernardleong (Dorje AI) on what that operating model actually looks like.

Design. Product. Engineering. AI is collapsing the distance between all three. @randyjhunt (@NotionHQ), Nathan Xu (@PLAUDAI), Jose Florido (@magnific) and Savannah from @ideo on what it means to build when the gap between idea and execution is almost gone.

Data center demand growing 3.5x by 2030. GPU lead times at 36–52 weeks. Grid connections taking 4–7 years. Sachin Hindupur from @AMD on the infrastructure wall AI is about to hit and what it takes to produce more with the same or less.

36 hours of building. 5 teams standing. The NEXT Hackathon finalists have been announced. They demo live on the Weka Stage at 5:40pm today. $200K in prizes powered by @awscloud , @Vercel, @ExaAILabs, @stripe, and @RazerAI .