Those two positions aren’t in tension. In fact, holding both at once is the only sane place to stand right now, and I think a lot of people flatten it into one or the other because nuance is exhausting to maintain in a market moving this fast.
The technology is real. I use it every day, it’s changed how I work, and I don’t think we’re anywhere close to the ceiling of what’s possible with the current generation of models, let alone the next one. That’s not the part I’m skeptical about. What I’m skeptical about is the number of businesses that have decided “we use AI” is a strategy, rather than a feature — the number of pitches I hear that describe a wrapper around an existing model as a moat, when the actual defensibility is somewhere between thin and nonexistent.
Here’s the tell I use now, and it’s blunt on purpose: if you can explain your company’s advantage without mentioning which model you’re built on, you probably have a real business. If the model is the first thing you mention, and the second, and the pitch collapses without it, you’re renting someone else’s capability and calling it a company. That’s not necessarily worthless — plenty of good businesses have been built on top of platforms they don’t own. But it’s a fundamentally different risk profile than the founder is usually presenting, and I think a lot of capital is being deployed without anyone doing that translation honestly.
The pattern that worries me most isn’t the obviously thin pitches — those get filtered out eventually. It’s the genuinely good teams building genuinely useful products on a foundation that could be commoditized or replicated by the next model release. I’ve sat across from founders who built something clever, something that clearly saves their customers real time and money today, and the honest answer to “what happens if the underlying model gets 30% better next quarter” is a shrug. That shrug is the whole risk, and it’s the question I now ask before anything else.
None of this makes me bearish on the space. It makes me more selective inside it. The AI companies I actually want to back are the ones where the value isn’t the model call — it’s the data flywheel they’ve built around it, the workflow integration that gets stickier over time, the trust relationship with a customer base that would be genuinely painful to rebuild elsewhere. That’s a much smaller set of companies than the current fundraising environment would suggest, and I think the gap between “AI company” and “durable AI-enabled business” is where the next few years of capital destruction and capital creation both get decided.
