Khosla Ventures

Today, PrismML is emerging from stealth. We are an AI lab with Caltech origins centered on building the most concentrated form of intelligence. At PrismML, we believe that the next major leaps in AI will be driven by order-of-magnitude improvements in intelligence density, not just sheer parameter count. Our first proof point is 1-bit Bonsai 8B, a 1-bit model that fits into 1.15 GBs of memory and delivers over 10x the intelligence density of its full-precision counterparts. It is 14x smaller, 8x faster, and 5x more energy efficient on edge hardware while remaining competitive with other models in its parameter-class. We are open-sourcing the model under the Apache 2.0 license, along with 1-bit Bonsai 4B and 1.7B. When advanced models become small, fast, and efficient enough to run locally, the design space for AI changes immediately. We believe in a future of on-device agents, real-time robotics, offline intelligence, and entirely new products that were previously impossible. This is the future we’re building toward. We’re fortunate to be backed by Khosla Ventures, Cerberus Ventures, Caltech, and supported by Google.

It’s in the name… MAGNETIC Confinement Fusion. To suspend a superhot plasma (many times the temperature of the sun) within a fusion machine, we at Realta Fusion need extremely powerful magnets. High temperature superconducting technology has made this possible, and Commonwealth Fusion Systems (CFS) has made it a reality. Today we announced our strategic partnership with CFS in which we will work together to design, develop and manufacture HTS magnets for our magnetic mirror fusion machines. With this agreement we are securing access to the world’s best developed HTS supply chain and the manufacturing capacity at scale needed for magnets in our demonstration devices and future commercial fusion power plants. Not only does this remove a massive amount of risk on our path ahead but it also accelerates the arrival of commercial fusion in Realta’s power plants. For CFS, this is a validation of their HTS magnet capabilities and business model. We appreciate the huge effort CFS has already put into supplying magnets for the WHAM device and for the design we are deep into for our next machine. Thanks to Darren Woulfe, Rick Needham, Bob Mumgaard, Elle Allen, Grant Kristofek and the whole magnet team at CFS for the support and for seeing the potential in our magnetic mirror fusion technology.

The next wave of robotics will be trained on the real world. Rhoda AI is using millions of internet videos to teach robots how to operate in complex industrial environments. “The real world is messy, complex and being able to actually work on production lines is much, much harder than doing the demo.” - Vinod Khosla Excited to welcome Jagdeep Singh back to the portfolio following two decades of working together (QuantumScape, Raxium, Infinera, Lightera and more!) https://lnkd.in/gqe39tSC

“Today’s large language models can write sonnets and debug code. But ask one to fold a towel and you’ll discover the gulf between knowing about the world and knowing how to act within it. Language is but a compression of human experience. Text captures only a thin slice of what we know.” KV Partner Nicole Fraenkel on world labs and why we bet on General Intuition last year https://lnkd.in/ggZyJjPs

Congrats to Guild.ai on their Series A — universal by design and neutral across models, vendors, and frameworks. Excited to support the team turning agents into shared production infrastructure

What Actually Scales in Precision Medicine? Great science is necessary. It is not sufficient. Yesterday at PMWC - Precision Medicine World Conference 2026, I gave a keynote on a question that sits underneath much of precision medicine: How do we decide which innovations actually scale? Not which ones are scientifically elegant. Not which ones are technically possible. But which ones survive real-world economics and become standard of care. After working as a physician-scientist, health system CEO, executive at CVS/Aetna, and now at Khosla Ventures, I’ve seen the same pattern repeat: In healthcare, scaling is not primarily a scientific problem. It’s a systems problem. And ultimately, it’s an economic problem. Here’s why great science often stalls at adoption: 1️⃣ Clinically interesting ≠ economically complete Payers ask: Does this change total cost of care? Is the effect durable? Does it hold up in real-world populations? 2️⃣ Operational burden is underestimated If innovation adds workflow friction in fragile systems, it gets rejected over time. 3️⃣ Incentives are misaligned Fee-for-service rewards activity. Precision medicine often reduces activity. Without aligned payment models, adoption slows. From an investment standpoint, durable precision medicine companies share three traits: • They change a clinical decision — not just generate data • They improve outcomes and unit economics • They have distribution embedded in real pathways Product matters. Evidence matters. But distribution compounds. When evaluating a company, I ask five questions: 1. Who is the true economic buyer? 2. What decision changes on Monday morning? 3. Is economic impact measurable within 1–3 years? 4. Does this reduce or add operational complexity? 5. Is distribution defensible? If those align, scaling becomes possible. If they don’t, no amount of scientific elegance compensates. The constraint in precision medicine is rarely technology. It’s alignment. Full remarks here: https://lnkd.in/ebr4jvJE #PrecisionMedicine #HealthEconomics #HealthcareInnovation #VentureCapital #DigitalHealth

Huge congrats to the team at Science! Proud to back this mission to restore vision and push the frontier of neural engineering and BCI innovation. Onward. 🚀

Sales teams, this is for you. https://lnkd.in/ejtuGPyV

🚀 We're Hiring: Enterprise Product Manager (DC or SF) 🏛️ State Affairs, backed by Khosla Ventures is building the AI-powered "operating system" for government affairs. We need a visionary B2B SaaS PM (5+ years) to own: 🔌 Ecosystem Integrations: Connecting our data to Fortune 500 tech stacks. 🤖 AI Engineering: Leveraging LLMs to surface critical policy signals. ⚙️ Complex Workflows: Solving governance and collaboration for massive teams. CC: Jamie Roberts Seltzer, Tamar Blanc, Andrew Dorsey Help us reshape how America decodes state policy! 🇺🇸 👇 Apply directly here: https://lnkd.in/gY7EMd4D