Why you can’t vibe-code a factory with Julian Counihan (Schematic Ventures)

by Alpha Partners Editorial

What happens when you invest in a category everyone calls uninvestable - and you turn out to be right?

That's the story behind Schematic Ventures and its founder Julian Counihan, who started writing pre-seed checks into supply chain technology and industrial automation in 2017, when the category was largely ignored by institutional venture capital. His background as a software developer building warehouse control systems gave him a read on the market that most investors simply didn't have: e-commerce growth required more automation, robotics was advancing fast enough to matter, and the founders entering the space had genuine operator credibility. The result was a first fund built around companies like Altana, Flock Freight, Platform Science, and Harbinger - a Class 6 electric truck manufacturer that may be the most capital-intensive pre-seed bet in Schematic's history, and one that Julian backed precisely because the founders walked in with a sophisticated multi-year financial plan.

Now, with Physical AI generating more investor attention than ever, Julian is navigating a new challenge: separating durable value from hype in a category that suddenly has no shortage of tourists. His view is clear-eyed. Most new entrants aren't coming in because they love industrial technology - they're escaping uncertainty in software, where it's no longer obvious what AI will or won't replace. You can't vibe-code a factory or an actuator. But that doesn't automatically mean every robot startup deserves a check. Julian's edge is two decades of factory visits, industry relationships, and lived experience through multiple hype cycles in hard tech - and that's the kind of credibility that doesn't transfer with a thesis pivot.

Beyond the portfolio, this episode goes deep on Julian's personal return to daily coding via AI agents and why he believes the real magic of tools like Claude Code isn't the model - it's the harness. The software architecture around the model is what creates that feeling of indistinguishable-from-magic. He's using that insight to automate Schematic's own operations, enabling a small team to move with the focus of a much larger firm. His take on humanoid robotics is equally grounded: the breakthrough won't come from a leap in model intelligence or hardware, but from finding the right application that matches current performance to the right use case. It's a conversation that rewards anyone thinking seriously about where the physical world and intelligent systems actually meet.