Lightspeed

AI just engineered 16 functional viral genomes from scratch, built to kill drug-resistant bacteria. Crafted by researchers at Stanford and the Arc Institute, the genomes are engineered to target E.coli, including strains that have always been resistant to the natural phage. The faster AI crafts biological design, the more manufacturing and clinical validation become the real bottleneck to patient delivery. More on Lightwork: → As foundation models get more capable and established players build out their own AI stacks, the moat for healthcare startups increasingly appears to sit in distribution and trust. → Healthcare, initially a slower adopter of AI, is having a leapfrog moment. Some physicians are embracing tools like scribing and voice AI, setting a new tone for healthcare adoption. Full video below. CC: Claire Zau, Brenton Fargnoli, M.D.

"If you're going to spend your life doing something, do something that matters." Lightspeed Partner Raviraj Jain's advice to founders building hard things: "It's a phenomenal time to be alive and a phenomenal time to be in technology and to be a founder." "Make the boldest and bravest bet to change the world as you would like to see it." "No idea is too ambitious. In fact, that's what we get attracted to: the most ambitious ideas." "The journey is not straightforward. You will have a lot of roadblocks... and that's fine, because that's really what building in these hard industries is about." From Lightwork

There’s a category-defining solution in cybersecurity. Joe Diamond, CEO of Axonius, explores asset intelligence and how it goes beyond inventory visibility, focusing on how assets interact within an organization’s ecosystem. Asset intelligence helps enable a deeper understanding of existing vulnerabilities. This week on Lightwork, Joe dives into the evolution of this API-based approach and its connection to the AI security capabilities of Axonius’s new MCP server. He even says that this shift could be “as big, if not bigger, than the invention of the internet itself.” Full video link in the comments below.

“The most valuable position in healthcare AI, we’d argue, belongs to whoever becomes the front door.” Lightspeed Partners Brenton Fargnoli, M.D., Galym Imanbayev, MD, and Faraz Fatemi write about the future of healthcare AI and where they see the venture opportunity moving: “Clinical AI has moved from interpretive work, like reading an image and offering a diagnosis, toward consultative.“ “The frontier from here is procedural and eventually surgical, which will be the point where AI stops living entirely in software and starts physically helping the human body.” Link to the full piece in the comments.

LLMs are increasingly becoming local. Unsloth Desktop centralizes AI training and execution all in a single, local space. Congratulations to Daniel Han, Michael Han (Unsloth), and the Unsloth AI team on the launch.

How Base Power Company turns thousands of homes into a virtual power plant: "Instead of having one giant power plant, you have thousands of homes and nodes that connect together." "They network together. They add up to a utility." "Power is often very cheap at night. Base charges up all their batteries cheaply then." "When the grid is slammed, like a very hot summer afternoon in Texas, prices spike." "It's basically buying low and selling high, at scale, thousands of times a day across thousands of homes." From Lightwork

Everyone I know says the same thing about AI: it’s impossible to keep up. I always assumed that was just a feeling. Turns out you can measure it. Every week on Lightwork, we struggle to cut the AI news down to three stories. This week, I finally understood why there’s so much of it. The internet took roughly five years to reach its first 16 million people. AI reached a billion people in roughly the same window. We sat down with Lightspeed’s Raviraj Jain this week, and he pulled up a chart spanning 2,000 years of world economic output. For most of it, the line barely moves. Then the steam engine arrives. Electricity follows. Then cars. Computing. The internet. AI. Each revolution arrives faster than the one before it. Raviraj’s explanation was simple: every technological revolution gets to build on top of the last one. Without electricity, you don’t get computing. Without computing, you don’t get the internet. Without all of it, you don’t get AI. But the part that really stuck with me is that each wave doesn’t just inherit the technology. It inherits the distribution and infrastructure, too. The internet had to wait for people to buy computers and get connected. AI showed up on devices billions of people already had in their pockets. That’s why this moment feels so fast. It actually is. And Raviraj’s bet is that the curve gets steeper from here, not flatter. Full conversation with Raviraj below.