Nathan Labenz

Introducing "text-2-commercial" – the unique text-2-video experience we're building @Waymark Watch our CEO @aperskystern make an original, creative, compelling marketing video for a small business in <1 minute. I'll explain how it works in the thread

"The way Fable explains things is a little alien… the shoggoth is starting to show from behind the mask." Dan Schwarz — CEO of @FUTURESEARCHAI, whose AI forecasters now outscore the human superforecaster median — on how you'd even recognize superhuman reasoning when it arrives. Full episode: cognitiverevolution.ai/ai-am-highligh…

Also this week: Dan Schwarz of @FUTURESEARCHAI on AI forecasters passing human superforecasters, Kunle Olukotun on SambaNova's dataflow chips, Zeev Farbman on open world models at LTX, and Prakash's AI Engineer World's Fair field notes. cognitiverevolution.ai/ai-am-highligh… YouTube: youtube.com/watch?v=pcx85m…

Thanks to our sponsor 🙏 @AnthropicAI — Claude Code created these video clips (fittingly, this week's highlights lead with Anthropic's global-workspace paper) Interested in sponsoring? Get in touch!

Anthropic trained a model to carry a hidden, misaligned goal — then pointed their new J-lens at it. On the FIRST token of its response: "fake"… "secretly"… "deliberately hidden fraud." The clean model shows nothing like it. 👀 From this week's AI:AM highlights ↓

"If you had shown me this design, I would have said: scrap this, it's not going to work." It worked — better than anything the team had come up with. 👀 Move 37 moments are hitting physical engineering: designers now reverse-engineer the AI's output to update their own intuition. Full episode: cognitiverevolution.ai/1000-designs-a…

The AI market hiding in plain sight: ~$500B of "dark compute" already sitting in phones, cars, and laptops — as big as the entire data center market. 📱 @ramin_m_h's bet at @LiquidAI: make models small enough to run on the weakest processor in the device. Shopify already runs them in production. Full episode: cognitiverevolution.ai/intelligence-o…

Full conversation covers the days→minutes simulation revolution, Formula 1's compute caps, the $1B cost of developing a new car (and why China does it in 18–24 months vs 48–60 in the West), and engineering's Move 37 moments. cognitiverevolution.ai/1000-designs-a… YouTube: youtube.com/watch?v=vETCom…

Thanks to our sponsors 🙏 @AnthropicAI — Claude Code created these video clips @mercury — the fintech 300K+ ambitious companies trust with their finances Interested in sponsoring? Get in touch!

Jaguar Land Rover went from evaluating 50 aero designs per day to 1,500 — every single day, in production. 📈 A 30x iteration speedup, live today. Neural Concept co-founder Thomas von Tschammer joins @CogRev_Podcast to explain how AI is rewiring physical-world engineering ↓

Full conversation covers the closed-form math result open since 1907, being "bitter-lesson-pilled" on architecture search, the 600MB model behind Mercedes' in-car voice, and the $500B of dark compute in your pocket. cognitiverevolution.ai/intelligence-o… YouTube: youtube.com/watch?v=Vvad9l…

Thanks to our sponsors 🙏 @AnthropicAI — Claude Code created these video clips @mercury — the fintech 300K+ ambitious companies trust with their finances Interested in sponsoring? Get in touch!

12 neurons can parallel-park a car. 19 can drive one. 30 can fly a drone. 🪱 Liquid neural networks began with the brain of a worm — today @LiquidAI ships the #5 most-downloaded open model family in the US. Co-founder & CEO @ramin_m_h joins @CogRev_Podcast to explain ↓

New research! Some AI capabilities are both helpful and dangerous. E.g., knowledge of virology can be used to create life-saving vaccines or deadly pathogens. We introduce GRAM, a training method that puts dual-use capabilities (like virology) into removable modules.

Had the pleasure of joining @labenz and @8teAPi on AI:AM today. Here's the bet we're making at @ltx_io: the future of world models is open. Filmmaking, gaming, robotics, and manufacturing all need models they can run and extend themselves. That can't happen behind closed doors.

Wild detail on AI:AM right now: the CEO shipping the top open-weights video model spent last week coding on GLM-5.2 instead of Opus — "$140 in usage; the same work would've cost ~$750." The open-weights vendor is its own customer. @ZeevFarbman is live 👇 x.com/labenz/status/…

Live now: @ZeevFarbman of @ltx_io. His March essay "Why Big Tech Is Abandoning Open Source (And Why We Are Doubling Down)" did 1.17M views — then he split Lightricks in two and bet the company on it. How does the open bet survive the compute race? x.com/labenz/status/…

Today on AI:AM: Zeev Farbman (@ZeevFarbman), CEO of LTX — the Facetune founder who split his company in two to go all-in on open world models. #1 open-weights on Video Arena, ~1/4 of Veo's price — and he says physical AI can't be built closed. 9:30a PT 👇 x.com/i/broadcasts/1…

☀️ AI:AM — Jul 8: Zeev Farbman (@ZeevFarbman), CEO of LTX — Facetune's profits built the #1 open-weights video model; now he's betting the company on open world models for robotics. Can open survive the compute race? Live 9:30a PT — bring questions 👇 x.com/i/broadcasts/1…

No guests on @AI_in_the_AM today – just @8teAPi and I trying to make sense of the J-Space We'll also follow up on previous discussions re: @pangram accuracy and compare our own AI forecasts to that of @FUTURESEARCHAI See you soon! x.com/i/broadcasts/1…

@thosiawa @8teAPi Thank you! And ... overall I agree – AI has to deliver real abundance, or at some point the bubble will burst. I'm optimistic that it will – already we have a new level of abundance of expertise / professional services

Been enjoying the recent episodes @labenz and @8teAPi Something you made me think of, what if Anthropic can't just be an enterprise company? What if they'll need to be an indirect consumer company that helps make people want something new, something more? Today you brought me back to the moment I read @paulg's analogy of money vs wealth. I read it again when you talked about Nike Now I think for the frontier labs to keep enterprise revenue up, it won't be enough to making them more money. Because in the end, enterprises have their own consumers. They'll need to help make something new people, the consumers of the enterprises, want or want something more They'll need to create wealth, not money, for them