Good morning! Most startups spend years chasing an exit. Moltbook built an AI social network, went viral, and got scooped up by Meta faster than founders can draft a pitch deck. Now the team is headed straight into Meta’s Superintelligence Labs to craft social networks for AI agents and explore agentic commerce. Not bad for a startup that didn’t even exist 2 months ago.
Yann LeCun just raised $1 billion to build physical AI. The ex-Meta chief scientist believes humanlike intelligence starts in the physical world. His new startup, Advanced Machine Intelligence, is building world models that pull messy data straight from sensors like cameras and microphones. The AI distills this noisy information into abstract concepts that it uses to plan actions and predict outcomes. With persistent memory he believes it will actually learn from experience. It’s like using a simulation to have a complete experience of the real world. Just like college. [WSJ]
Mira Murati locked in 1 gigawatt of Nvidia’s next-gen Vera Rubin chips. The former OpenAI CTO secured tens of billions in compute, plus an undisclosed Nvidia investment into her startup, Thinking Machines. Right now, their only product is Tinker, an API that lets you fine tune AI models without massive hardware costs. But Murati’s aggressive hiring of elite AI researchers hints at bigger ambitions. Her real goal is creating AI that anyone can customize and control. With compute power on this scale, she’s clearly planning something massive. Exactly what, Murati’s not saying (yet). [Inc]
Most AI productivity tools are junk. Wispr Flow is one that will actually save you time. Just talk normally and it turns your voice into clean text in any app. Zero edits. 4x faster than typing.
Millions use it daily, including teams at OpenAI, Vercel, and Clay.
– Speak to Claude, Cursor, ChatGPT, and all your other tools
– Flow understands code and context
– Talk and you’re ready to send as-is
Now on Android, and free during launch. Try Wispr Flow.
A judge banned AI agents from freely surfing the web. A US district judge ruled AI agents can’t access websites without explicit permission from platform owners. The decision stems from Amazon’s lawsuit accusing Perplexity’s Comet AI browser of logging into user accounts without Amazon’s consent. Perplexity argued the agents had user’s permission to access the marketplace, same as if people shopped. The judge disagreed, deciding platform owners, not users, have final say. Using agents could now violate federal computer access laws. Perplexity plans to appeal, calling the ruling harmful to consumer choice and innovation. If it stands, OpenClaw might become obsolete overnight. [Forbes]
Reach your audience
Get in front of 70,000+ tech decision makers with budget and technical influence. 95% US based. 66% open rate. Read daily by AI leaders and builders at growing companies.
Google and Tesla think they’ve found a “smarter” way to run the power grid. They formed a group called Utilize to reclaim the grid’s idle capacity. Today’s electrical grid is built for short spikes in demand, leaving huge amounts of potential energy unused. Utilize says sustainable tech like battery storage, virtual power plants, and smart panels already exist to unlock all that untapped energy. But if smart systems depend on AI to allocate energy, do we really trust it to decide who gets power? [TechCrunch]
Silicon Valley engineers are now asking to get paid in AI tokens. Top candidates are negotiating for dedicated compute and token budgets in their compensation packages. Investors like Tomasz Tunguz predict these inference packages could soon make up over 20% of total pay, potentially adding $100k per engineer annually. For engineers, access to GPU-hours allows them to run more agents at once, boosting their productivity. Skip the equity and go for the GPU, because in 2026 your trajectory depends on how much compute you control. [Business Insider]
fun stats
💔 1 in 3. AI apps are ditched within a month. They make 41% more per user, but users churn 30% faster than traditional apps.
💧 2030. The year AI data centers could drink up 1.45 billion gallons of water daily, equal to all of New York City. And that’s the “very conservative” estimate, according to a recent study.
🤖 93%. Companies planning to deploy their own agentic AI within 2 to 3 years, rather than outsourcing to developers or using SaaS tools.