Impossible Startup Ideas: Why Disruptive Beats Incremental

Key Takeaways

Two modes of progress: Incremental improvement copies what works (1 to n). Disruptive improvement invents what does not exist yet (0 to 1).
Your brain rejects the disruptive: Humans extrapolate from the present, so genuinely new ideas feel impossible before they feel obvious.
The biggest unicorns sounded absurd first: Airbnb, SpaceX, Uber and Starcloud each began as an idea most serious investors dismissed.
Outliers pay the rent: Per Horsley Bridge data, roughly 10% of VC deals generate over 90% of returns, and those deals are almost never incremental.

Most investors pass on the startups that later define a decade. The reason is not bad diligence. It is a wiring problem. The human brain is an extrapolation engine: it predicts the future by stretching the present. So when a founder describes something that has no precedent, the brain has nothing to stretch from, and it defaults to "impossible."

The startups that actually move the curve, from Airbnb to SpaceX, started as ideas that sounded un-serious or un-workable to the people who heard them first. That gap, between how an idea sounds and how it ends, is exactly where the largest returns hide.

Disruptive improvement is progress that creates a new category by solving a problem no one has solved yet, often because no one believed it could be solved. It is not a better version of an existing product. It is a version that did not exist before you built it.

Incremental vs Disruptive: The Two Modes of Progress

In Zero to One (2014), Peter Thiel, co-founder of PayPal and Palantir, draws a clean line between two kinds of progress. Horizontal progress, which he calls going from "1 to n," is copying what already works: taking a working model and scaling it, tweaking it, marginal-cost-improving it. Vertical progress, going from "0 to 1," is doing something entirely new. Thiel argues that real technological advance comes almost entirely from the second kind.

The distinction matters because the two modes produce completely different return distributions.

Dimension	Incremental (1 to n)	Disruptive (0 to 1)
Risk	Low, predictable	High, hard to underwrite
Returns	Linear, capped	Power law, uncapped
Idea stage	Sounds sensible	Sounds impossible
Competition	Crowded, fast followers	Empty, you define the lane
Market	Known TAM	Unknown, you create it
Capital efficiency at exit	Lower multiples	The 50x outliers

Incremental startups are easier to diligence because the comparables already exist. That is exactly why their returns are capped. Disruptive startups are hard to diligence because the comparables do not exist yet. That is exactly why a single win can return the rest of a fund combined.

Thiel frames the whole thing as one question: "What important truth do very few people agree with you on?" The contrarian answer, the one most people dismiss, is where vertical progress hides.

Why Your Brain Rejects Disruptive Startup Ideas

The 0-to-1 problem is not just a market problem. It is a cognitive one.

A brain that evolved to predict the next few seconds from the last few seconds is not built to model things that have no precedent. When you hear "people will pay to sleep in a stranger's spare room" or "private rockets will land themselves back on a pad," the prediction engine has no prior to draw from. It does not output "this is hard." It outputs "this is impossible."

That reflex keeps you alive on the savanna. It costs you the next Airbnb.

This is why the same idea can look insane in 2007 and obvious in 2017. Once the thing exists, your brain retro-fits it into the model and calls it inevitable. Before it exists, the model throws a 404. According to CB Insights, only 0.07% of venture-backed startups ever reach a $1B valuation. Most of those survivors are the ideas the prediction engine refused to render.

Worse, the safer the idea feels, the more competition it has, and the lower the return. Felt-safety and felt-impossibility are inversely correlated with the size of the outcome.

The Startups That Looked Impossible (And Won)

The pattern is consistent enough to be a screen. Four companies that sounded absurd at first pitch, and what they became.

Airbnb: paying to sleep in a stranger's airbed

In 2008, the idea that strangers would rent each other's homes, sleep in each other's beds, and trust a website to mediate was broadly mocked. Investors passed because the model had no comparable and "people just won't do that." Airbnb reached a valuation of over $80B at its December 2020 IPO. The thing the brain refused to render became one of the highest-valued marketplace businesses ever built. It is also a textbook network effects company: every new host made the platform more valuable for every guest.

SpaceX: a private company that lands rockets

In 2002, the consensus was that rockets were for nation-states and that reusing an orbital-class booster was engineering fantasy. SpaceX made orbital-class reusable rocketry real, landing its first booster on December 21, 2015. That single 0-to-1 move collapsed launch cost per kilogram and reopened a market, small-satellite constellations and orbital infrastructure, that depended on cheap, repeatable access. SpaceX now sits near the top of the private-company valuation table.

Uber: black-car rides from a phone, against the taxi cartel

"Strangers will get into a stranger's car at 2 a.m. because an app says it is fine" sounded un-serious in 2009. The taxi industry was a regulated century, an obvious fortress of fixed supply. Uber priced at a $45 share and roughly an $82B valuation at its May 2019 IPO. The disruptive move was not the app: apps existed. It was building a two-sided real-world market on top of phones, payments, and trust that no incumbent had assembled.

Starcloud: data centers in orbit

Putting gigawatts of AI compute in space sounds like a science fiction pitch. That is the point. Starcloud, originally founded as Lumen Orbit and now operating as Starcloud, argues that as launch costs fall, data centers in space unlock continuous solar energy, radiative cooling, and gigawatt-scale deployment without Earth-bound permitting constraints. According to TechCrunch, roughly 200 VCs tried to get into its $11M seed round in December 2024, with backing from NFX, Nvidia, Sequoia, Benchmark, and Andreessen Horowitz. Whether it works is still unknown. The signal is that the category is being taken seriously at all.

The headline is the same in every case: the idea that felt impossible to imagine was the idea that created a market.

Why Disruptive Bets Pay the Most

This is where the cognitive problem meets the math problem.

Data from Horsley Bridge Partners, one of the largest LPs in VC funds, shows that across 7,000+ investments from 1985 to 2014, roughly 65% of individual deals returned less than 1x. The top 20% generated over 90% of total returns. Only about 4% of deals returned more than 10x.

That shape is the power law. And it has a brutal consequence: a portfolio optimized for sensible, incremental bets is structurally built to miss the whole game. One 50x return outweighs ten 2x returns. The cost of a miss is small. The cost of missing the winner is fatal to the fund.

Because incremental ideas are easy to underwrite, they are easy to compete for, and the competition compresses the return. Because disruptive ideas are hard to underwrite, they are easy to pass on, and the passing is what leaves the upside on the table for the few investors who can sit with the discomfort of a 404 in their own head.

How to Put This Into Practice

None of this means every impossible-seeming idea is a winner. Most are not. The job is to separate "sounds impossible because it is hard" from "sounds impossible because the brain cannot render it yet." That is exactly the gap a research-backed evaluation tool is built to close.

Unicorn Screener scores startups across the dimensions that research shows predict outlier outcomes: founder quality, market structure, traction velocity, and competitive defensibility. The point is not to love impossible ideas blindly. It is to discipline the gut reaction that rejects them on reflex. The public Top 50 leaderboard is a live view of where the highest-scoring companies currently sit, several of which began as pitches that sounded un-serious.

One honest caveat: no scoring model can guarantee an outcome, and many genuinely impossible-seeming ideas are in fact impossible. Disruptive is not the same as correct. The framework just keeps you from auto-rejecting the ones that are merely un-renderable.

What This Means for You

Default to skepticism about felt-impossibility. When an idea sounds absurd, ask whether it is wrong or merely un-renderable by your brain.
Hunt for the contrarian truth. Thiel's question is a screen for vertical progress. If everyone agrees, it is 1 to n.
Look for the empty lane. Real 0-to-1 categories have no comparables. The absence of a comp is the signal, not the bug.
Score the founder, not the precedent. When there is no market to model, founder traits carry most of the predictive weight.
Screen systematically, not reflexively. Try Unicorn Screener to score a startup on the dimensions that actually matter before your gut casts the deciding vote.

For more on how the math punishes safe bets and rewards the un-renderable, see what makes a startup a unicorn and the power law of VC returns. The biggest winners do not sound like winners on day one.

Want to screen startups like a top-tier VC? Score any startup for free with our research-backed evaluation model.