Agent Defined Networking built from first principles doxx.net/blog/welcome-to-the… @BarrettLyon

Big launch from Coworker: OM2, 9x cheaper enterprise AI. Congrats Alex and team!

Congrats to @coworkerapp and team on OM2, proof that memory beats bigger models.

Big news today. @coworkerapp just dropped Opus-quality enterprise AI for one-fifth the price. Same frontier models, 82% less cost.

Privileged to be joining Woods Hole Oceanographic Institution as a Corporate Board Member. I feel most at home in the water. From marine research diving in Africa to ocean swims along the Australian coastline, the ocean has shaped some of the most meaningful parts of my life. Joining WHOI feels like an opportunity to channel that lifelong connection into something with real purpose. I see us entering a remarkable era for ocean science. Despite covering most of our planet and being integral to life, the ocean remains one of the least understood frontiers on Earth. Over the next decade, breakthroughs beneath the surface could reshape how we think about medicine, food systems, climate, materials, energy, and biodiversity itself. We have so much to learn, I can't wait. A few things that make WHOI such an extraordinary institution: - The world’s largest non-profit dedicated to ocean research and engineering - $500M endowment supporting long-term scientific discovery - 1,000 employees, including ~500 scientists and engineers - Joint program with Massachusetts Institute of Technology training the next generation of ocean scientists - A fleet of world-class research vessels and deep-sea submersibles - Nearly a century of advancing humanity’s understanding of the ocean For 90 years, WHOI has been behind some of the most important discoveries in ocean science. It’s a privilege to now contribute, in some small way, to what comes next. Special thanks to David Scully and Paul Salem for the introduction, encouragement, and leadership. If you care deeply about the future of our oceans I’d love to connect. Let’s get to work.

Congratulations to Nima Aghaeepour & Joe Phongpreecha for bringing Takeoff41, Inc. out of stealth today. A Stanford-incubated startup modernizing care for pre-term babies in the NICU. Science published in Nature, first hospital live, grant funding secured and a big year ahead. I’ve written recently about two ideas: (1) category-defining missions, and (2) AI as the wedge for offline monetization. Takeoff embodies both. Their AI software guides nutritional recommendations for TPN drip bags, often a premature baby’s only source of intake. Yet what I admire most is their ambition from day one to build a full-stack neonatal care platform focused on outcomes, not just tools. Manufacturing and delivering next-generation TPN bags guided by their AI software — greater value to the customer, and greater value capture for Takeoff. Deliver the outcome, not the software. Proud for focal to lead their pre-seed alongside our friends at @BasisSet

Thesis-driven founders are winning. They’re trading at a premium, attracting more VC capital far earlier. These are the Seed rounds making headlines. Thesis-driven founders aren’t selling products. They’re not leading with metrics. They’re articulating a point of view about how a market will develop, and inviting investors to underwrite that future. And when the future is this in flux, that clarity feels like the calm in the eye of the storm. Everyone else is circling the outside, starting to look like versions of the same startup, competing and blurring into one mass. You can usually tell which side you’re on from the diligence questions you’re fielding: what about X competitor, how do margins scale, where’s the defensibility? When a company is truly thesis-driven, the conversation shifts. It compresses down to something much simpler: Do you believe? Because if the thesis is right, multi-stage funds believe they can mobilize the capital and ecosystem around it to make it real. After all, VCs are in the belief business. What really works for thesis-driven founders, especially those whose thesis comes from lived experience or deep technical insight, is that they create narrative momentum. In a world changing this quickly, narrative momentum becomes a magnet. It pulls in capital before it’s obvious, attracts talent who want to be part of something meaningful, and signals to customers that this company might define the space rather than follow it. Over time, that belief compounds. Iterative product building, on its own, can trap you in the present. This isn’t fake it till you make it. It’s hard. It’s building real conviction to call the future early, and then building toward it before the rest of the market catches up. At Focal, these are the founders we’re drawn to. Dream far out. Dream deeply. Create your narrative momentum.

Better AI products are not winning by default right now. I’m seeing weaker solutions raise at 5–10x the valuation of stronger ones at the same stage — $6M seed vs $50M for companies that shouldn’t be in the same conversation. Customers aren’t seeing the value yet, revenue isn’t there, and the tech is often underbaked. Founders feel it when those funding headlines hit. At the core, this is a market rewarding sizzle over steak. During seismic platform shifts, multi-stage funds become power-law maximalists. They’re not trying to be right often — they’re trying to be massively right once, and put as much capital behind that outcome as possible. So they back what could dominate — the big story, the founders who can sell the winner early — not the product that’s already better. And once that capital lands, it compounds. Customers behave the same way. They want to buy the winner — the product with momentum, that feels inevitable, that will still be around in 5 years. Startup IBM. So capital becomes momentum. Momentum becomes belief. Belief becomes reality. The uncomfortable implication is that having the better product isn’t always enough. You have to win on multiple fronts: persona, narrative, speed, distribution, hiring, capital. The best founders don’t just build. They outmanoeuvre.

Counterintuitive feeling right now as a pre-seed investor: the fast, iterative startup is the riskier bet. Shipping fast isn’t optional, it’s expected. AI founders are told winning means launching on day one, hitting $5M in 12 months, and marking your arrival with an iconic round… then sprinting uphill again. The problem is everyone is running that same race, and fast products don’t stay differentiated for long — they get absorbed, copied, and iterated to the norm quickly. The companies that matter most often take longer to emerge. They start from the end state, build what isn’t yet visible, and are willing to look foolishly early. Founders willing to boil the ocean have suddenly become the safer bet. As five-year moonshots collapse into months, those are the companies we want to build our portfolio around. The patient capital that has since become scarce in today's market. It’s uncomfortable for founders and VCs. But that’s kind of the point.

As Anthropic melts away entire categories, the bar for what makes an interesting pre-seed investment has shifted. I increasingly find myself backing three types of companies: 1. Mission impossible Technical or scientific founders going after problems long deemed out of reach — now within striking distance because of AI. AI isn’t the product, it’s the unlock. 2. Trojan horse Agentic workflows used as a free or low-cost wedge — with real value accruing at the transaction or financial layer beneath. 3. Data treasure hunters Businesses that unlock novel, hard-to-access data — then build the model and feedback loop around it. Own the data, own the outcome. We’ve invested across all three recently. In fact, we hit a Focal record : deploying $10M over the last quarter. 2026 energy is on 🔥

Fascinating to see AI business models morph. Seems the way AI is monetized may be more disruptive (and defensible) than the tech itself? Most discussions are SaaS to usage or Per seat to outcome. I'm seeing more novel models that are 10-100X the value of a seat. Startups where AI isn’t the product being charged for, but rather a foothold into much larger pools of value. A few recent examples: 1. AI monetized by physical goods Biotech offering AI diagnostics at cost, then capturing value selling the drug compounds their models identify. 2. AI monetized offline Free AI predictive maintenance on industrial equipment used as the wedge to sell long-term servicing agreements and maintenance contracts. 3. AI unlocking financial assets Insurance workflows where AI structures claims for easy remediation, which then become securitized financial assets for trading and lending. I love seeing this kind of thinking. If you’ve seen other interesting models - or are building one - please let me know!