Kris K.

When machines learned to talk, everything changed. Not because of the technology itself, but because the last barrier between analysis and action disappeared. HappyRobot just announced a $44M Series B led by Base10 Partners. This isn't just a funding round. It's a declaration about what becomes possible when automation goes from partial to total. No more waiting for a human to call the supplier in Shenzhen or handle the angry customer in Dallas. Systems no longer suggest what needs to be done—they do it. Voice was the trigger that enabled total automation across every channel, turning suggestion boxes into engines of pure execution. But the real revolution is compound intelligence. When one AI worker discovers what works, every instance learns instantly. Your logistics AI masters your exact routes. Your negotiation AI learns your suppliers' psychology. These aren't generic models anymore, they're extensions of your business that can't be replicated. The winners aren't waiting for better technology, they're building their future now. Every day of compound learning, every customer interaction, every optimization becomes irreplaceable advantage. By the time others start, you'll have months of evolution they can never catch. The gap isn't growing. It's exponentiating. Read our full manifesto: https://lnkd.in/emb8qeXh Thank you to our incredible team, our visionary investors, and our customers who believe in a new form of economic life.

200

60 Comments

I'm excited to share that Goodfire is partnering with Radical AI to bring interpretability to materials science AI. Our scientists will build generative tools together to directly create new materials based on desired function, delivering simulated predictions in minutes that would take traditional methods like density functional theory (DFT) days or weeks! Scientific discovery is a profound application area for interpretability, and it’s one of our core focuses. Using interpretability techniques on scientific AI models with super-human performance helps us discover scientific insights while also serving as a testbed for our interpretability techniques. I’m excited to expand this work from biological models, like Evo 2, to materials science and engineering. I’m even more excited to work with Joseph F. Krause, Jorge Colindres, and the Radical AI team! They are experts in materials, move unbelievably quickly, and have a bold vision of the future. I think we have a good shot of unlocking more efficient and finely controllable approaches to materials design together. Read the full announcement: https://lnkd.in/eAdckcxN

194

10 Comments

Today the Pulse team published a deep dive on why a single “accuracy” score doesn’t tell you if a document extraction system will survive in production. The goal here is to lay out an introductory but still rigorous evaluation methodology - we have an exciting open-source benchmark building on this research coming out very soon. Let’s do the math: take 1,000 pages, each with 200 data elements. A model that’s 98% “accurate” on paper still produces 4,000 incorrect values. Now make some of those: 1/ Broken reading order that scrambles multi-column layouts 2/ Tables with shifted columns or missing headers 3/ Cross-page context lost entirely That’s enough to silently corrupt an entire dataset without throwing a single error. We’ve processed hundreds of millions of pages and built a multi-axis evaluation framework to measure what actually matters: reading order validation, region-level ANLS, reading order accuracy, TEDS for table structure, and continuity checks across page boundaries. The result? Fewer silent data corruptions, more predictable performance, and pipelines that keep working on the next million documents you haven’t seen yet. Full technical write up in the comments!

36

2 Comments