ElevenLabs just doubled its valuation to 2 billion — in five months. The London-based voice AI startup is exploring a secondary share sale that would let employees cash out at roughly 2B, up from 1B in February. The tender could close by September. Here's what caught my attention: → Founded in 2022, now serving Cisco, Meta, Salesforce, The New Yorker, The Washington Post → 30M+ ARR at end of 2025 — among the fastest-growing private AI companies → Went from .1B → 1B → 2B in ~2 years. That's a 20x jump. Secondary sales like this are becoming the norm in AI. Employees get liquidity without waiting for IPO. Companies retain talent in a brutal hiring market. Investors get access to proven revenue without the IPO premium. But the real signal? Voice AI is no longer a novelty. It's infrastructure. When Meta, Salesforce, and major publishers all embed your tech into their products, you're not a startup anymore. You're a platform layer. The 2B number also tells us something about private market appetite. Public markets may be wobbly, but the best AI companies are still commanding premium valuations — and finding buyers. My take: ElevenLabs won't be the last. We're going to see a wave of secondary sales and tender offers from AI startups with 00M+ ARR. The IPO window is narrow. Secondaries are the exit path of 2026. Link in first comment 👇 — Robbie Jeo

A $35,000 animated website. Built in one AI session. For $12. No web studio. No GSAP hand-coding. No Lenis setup. No frame extraction by hand. Here's the exact stack: → Claude Code writes every line of the code → Higgsfield renders the video assets → One agentic session ties it all together The output: smooth scroll animations, video frame extraction, production-ready UI — the kind of site agencies charge $6K–$35K to build. This isn't just "AI saves time." This is the price floor of creative work collapsing in real time. Web studios aren't competing with cheaper developers anymore. They're competing with a $12 API call. The question isn't whether this changes the industry. It already has. The question is: are you building with these tools, or waiting to see what happens? Source: @zeuuss\01 showed the full build — worth studying. — Robbie Jeo https://t.co/F0TFGOyItc

ElevenLabs just doubled its valuation to $22 billion — in five months. The London-based voice AI startup is exploring a secondary share sale that would let employees cash out at roughly $22B, up from $11B in February. The tender could close by September. Here's what caught my attention: → Founded in 2022, now serving Cisco, Meta, Salesforce, The New Yorker, The Washington Post → $330M+ ARR at end of 2025 — among the fastest-growing private AI companies → Went from $1.1B → $11B → $22B in ~2 years. That's a 20x jump. Secondary sales like this are becoming the norm in AI. Employees get liquidity without waiting for IPO. Companies retain talent in a brutal hiring market. Investors get access to proven revenue without the IPO premium. But the real signal? Voice AI is no longer a novelty. It's infrastructure. When Meta, Salesforce, and major publishers all embed your tech into their products, you're not a startup anymore. You're a platform layer. The $22B number also tells us something about private market appetite. Public markets may be wobbly, but the best AI companies are still commanding premium valuations — and finding buyers. My take: ElevenLabs won't be the last. We're going to see a wave of secondary sales and tender offers from AI startups with $100M+ ARR. The IPO window is narrow. Secondaries are the exit path of 2026. Link in first comment 👇 — Robbie Jeo

Unreal Engine just launched an MCP server. That means you can now build a full video game by talking to Claude. No wrestling with the engine. No memorizing node graphs. Just describe what you want, and it builds. I've been watching this shift happen slowly — AI writing code, AI designing UI, AI generating assets. But this is different. This is AI building worlds. The last barrier for indie game development was always the engine complexity. Unreal is notoriously steep. Most people give up before they ship anything. That barrier is gone now. If you've had a game idea sitting in your notes for years but never started because "I don't know Unreal" — that excuse no longer holds. The tools are ready. The question is whether you are. — Robbie Jeo h/t @per_simmons_ https://t.co/ozltrPiXez

AI chips are hitting a wall. And the solution is 4 billion years old. NVIDIA's next-gen GPUs are pushing past 1,000W of thermal design power. That's not a typo — one single chip now consumes more power than an entire high-end gaming PC. Traditional cooling methods are breaking down. Data centers are sweating. Literally. Enter synthetic diamond. China produces 95% of the world's synthetic diamonds. While the West focused on silicon, Chinese labs quietly built diamond wafer technology that conducts heat 5x better than copper. The result? Cooler chips. Denser data centers. And a complete rewrite of who controls the AI hardware stack. Diamond isn't just jewelry anymore. It's infrastructure. — Robbie Jeo

WeChat just quietly shipped AI to 1 billion people. No keynote. No viral demo. No "ChatGPT killer" headlines. Just a small icon in the search bar. Tap it, and you're talking to Xiaowei — Tencent's AI assistant now embedded in the app 1.3 billion people already use daily. This is how AI actually wins. Not by launching a standalone app nobody downloads. Not by forcing users to learn a new interface. But by becoming invisible infrastructure inside the tools people already can't live without. The West builds AI products. China builds AI plumbing. And plumbing always wins. — Robbie Jeo

At Anthropic, 90% of engineers build self-improving agents using one core principle: Close the loop. Not just "give the AI a task." Give it a way to check its own work. Here's the pattern they use: 1. Agent generates output 2. Agent runs a verification step against its own result 3. If it fails, it retries with new context 4. Loop until it passes — or escalates This is why modern AI agents feel smarter than older ones. It's not a better model. It's better architecture. The model hasn't changed. The loop has. If you're building agents today, stop thinking about prompts first. Think about feedback loops first. A mediocre model with a tight feedback loop will outperform a great model with no loop — every single time. — Robbie Jeo

Anthropic engineers just demoed how they build a full app from scratch using a loop of 3 AI agents. One plans. One builds. One judges. The loop keeps running until the app actually works. Most people are focused on which model is the smartest. But the real advantage? Knowing how to orchestrate agents together. The winners won't be the ones with the best AI. They'll be the ones who design the best systems around it. Full 40-min walkthrough from the Claude Code team: https://t.co/c8itOIzDbP — Robbie Jeo

A 21-year-old just ran 300 AI agents simultaneously. But here is the part most people miss: He did not just scale up. He built in truth. 300 Kimi K2.6 agents fire in parallel as workers. 1 Claude Opus 4.8 agent checks every single output. The checker cannot be fooled by the workers. The workers cannot collude. They do not talk to each other. This is not just a speed trick. This is accountability at machine scale. Most people building with AI think about output volume. The smart ones think about output integrity. A system that produces 300 answers fast is useless if 30 of them are wrong and you cannot tell which. His architecture solves this. If you are building AI systems in 2026, ask yourself: Where is your checker? — Robbie Jeo https://t.co/wsy3DL8hW6

@OpenAI Launches Scheduled Tasks in ChatGPT — A Step Toward Proactive AI OpenAI has rolled out a new feature called Scheduled Tasks in ChatGPT. Users can now ask ChatGPT to automatically handle recurring tasks, send reminders, or monitor information on their behalf. Key highlights: • Create tasks that run at specific times or during set windows (e.g., morning, afternoon, or evening) • ChatGPT can monitor the web and connected apps, and only notify you when there are relevant updates • New dedicated Scheduled page in the sidebar to view, edit, pause, resume, or delete active tasks • Tasks run at most once per hour and will auto-pause after a period of inactivity The feature is currently rolling out to Plus, Pro, Business, and Enterprise users. Why it matters: This is one of the clearest moves yet toward making ChatGPT truly proactive and agentic. Instead of users having to manually check or prompt ChatGPT every time, the AI can now handle routine monitoring and recurring work in the background — potentially changing how people use AI in their daily workflow.