Every once in a while, you can feel the ground shift beneath your feet.

After reviewing the startup applications for our last 1752vc cohort, we noticed something unusual: a clear bifurcation. A line in the sand between two eras.

A moment in which the dominant paradigm is no longer “software is eating the world” - it’s that software itself is evolving into something entirely new.

The metaphor that kept surfacing in our diligence rooms?

Horse and buggy → Buggy with an engine → The modern car.

Let me explain.

“We’re building a better dashboard for…”

These startups are stuck in the last era of software. You’ve seen them: founders pitching new spins on workflows that already exist - and have already been saturated with tooling. They’re often chasing digitization for the sake of it, but the real question is: who’s still asking for this?

Common traits:

These aren’t bad businesses. But they’re built for an environment where the buyer still wants “software to help their team work better” - not systems that do the work.

Founders in this category tend to be strong domain experts or ex-consultants. But the tech vision often lags behind. The infrastructure and UI may look modern, but under the hood, it’s a repackaged service business with code.

In this world, AI isn’t a threat - it’s an existential question mark. Because once your customer realizes they don’t need a person clicking buttons anymore, your value prop vanishes.

“We use AI to…” (but it’s just glued onto legacy logic)

This is where most early-stage AI startups currently reside. They’ve bolted ChatGPT onto their product, added some summarization or automated replies, and called it a platform.

What’s happening here is transitional. These companies aren’t building AI-native businesses - they’re trying to sprinkle AI dust over traditional workflows. And while that can be good enough for early traction, it won’t be enough for long-term defensibility.

Common traits:

There’s often strong short-term utility, but long-term fragility. Why? Because anything a buggy-with-an-engine startup does, a real car startup will eventually do natively, cheaper, and faster.

This category is where the vast majority of pre-seed funding is currently flowing. It feels exciting. It demos well. But it could be a trap.

If the AI feature is a bolt-on, it will be commoditized.

If the workflow is unchanged, it will be replaced.

If the product can’t generate data moats or develop a unique feedback loop, it won’t survive.

“There’s no dashboard. There’s just a result.”

These are the next-gen companies. They don’t ask, “How can we improve this existing process?” They ask, “Why does this process exist in the first place?”

Car startups are AI-native from day one. They’re not trying to be more efficient; they’re trying to be unnecessary in the best way possible.

Common traits:

These are companies where the cost to operate approaches zero as usage scales. Think: no account managers, no onboarding calls, no training videos - just a prompt, an output, and a continuously improving engine underneath.

The founders in this category don’t pitch “AI features.” They’re often infrastructure-first, or they’ve rethought the category entirely. It’s not an AI writing tool - it’s a co-author. Not a dashboard - it’s an agent.

Investor lens: These startups look lean, strange, sometimes even “too early.” But once they click, their compounding advantages accelerate faster than the market can catch up.

They’re not just hard to build. They’re hard to compete with once built.

If we’re genuinely in the car-building era, we haven’t yet hit the Model T moment. But the roads are being laid, and the infrastructure is catching up.

Here are a few frontier areas we’re actively watching (and backing):

What This Means at 1752vc

We’ve shifted our lens accordingly. The best founders we’ve backed in the last 12 months aren’t asking “How do I build faster?” They’re asking, “What can be replaced altogether?”