Funding the Fog

Early-stage venture funding now shapes what founders build, and this can systematically push them away from deep innovation. Capital and consensus often do more than just distribute resources. By warping the space of imaginable ideas, they influence founders towards concepts that fit existing narratives. Imagination narrows, and novel ideas get traded down into fundable ones. This effect is uneven; venture capital is not monolithic, and seed capital has fueled remarkable progress. But like any powerful tool, its incentives shape behavior in ways worth examining.

Transformative ideas can fail to get off the ground not because they're bad, but because they are unintelligible to the structures that distribute attention and capital. Legibility dominates because capital demands signals amid uncertainty and fiduciaries prioritize defensibility over fortitude. Funding probability and technological uniqueness share an inverted U-shaped relationship with moderately novel ventures raising 2-3x more capital than ventures >2σ from the norm.1

Capital flows preferentially to ideas that fit easily into benchmarks or narrative heuristics. Fundability is easy to see—it can be measured, benchmarked, explained, and defended. True long-term value often emerges in ways that are not immediately visible. It accumulates in non-linear networks and long feedback loops. Early-stage funding tolerant of ambiguity can shift this dynamic and create space for breakthroughs that otherwise struggle to gain traction. Studies support the reciprocal of this pattern, showing that startups which resist herd behavior and pursue non-consensus paths are more likely to survive and succeed where consensus entrants falter.2

The pull toward familiarity is visible in the market. Cohorts of founders are converging on remarkably similar products. Recent accelerator batches show clusters of almost identical companies.3 Concerningly, founders who cater to familiarity are 19.3% less likely to receive patent allowances and show slower technological progress, despite securing terms sheets at a higher rate.4

The pattern is not new. Containerization was overlooked because its payoff depended on ecosystem-wide coordination, making a discrete return impossible to model. Malcolm McLean self-funded what became the largest cost collapse in trade history while capital concentrated in incremental logistics upgrades. Tesla was not obvious to investors because it was evaluated as a car company rather than a multi-layer reinvention of batteries, supply chains, and charging infrastructure. Small early-stage funds, otherwise known as micro VCs, appear repeatedly in the origin stories of outlier companies, not because they are unusually prescient, but because their incentive structure pushes them toward variance. They cannot outbid larger funds for consensus deals, so they spread thinner capital across idiosyncrasies.5

These examples are, of course, visible precisely because they succeeded; many radical ideas fail. However, they illustrate the type of misalignment that becomes much more likely under the dynamic described in this essay.

In an era of accelerating changes, this structural dynamic risks locking society into local optima, with value accruing unevenly to incumbents rather than originators.

The system trains the next generation of builders to prioritize perception over potential, clarity over depth. Incremental innovation is more easily funded, while radical ventures may struggle to attract capital. Venture media arms embed storytellers to craft viral pitches, amplifying signals that drown illegible edges. The Bay Area echo chamber fuels public narrative that outruns the slope of actual improvement. Returns will be thin where the market is the loudest.

A lot of the money pouring into AI is therefore being invested in the wrong places, and aside from a couple of lucky early investors, those who make money will be the ones with the foresight to get out early.

AI offers a contemporary illustration of this distortion. Frontier models and consumer applications attract the bulk of capital because they are easy to evaluate and benchmark. But these layers are structurally where value compresses the fastest—cost curves race downward, competitive moats erode, and distribution advantages collapse into whoever controls compute and data. Real value is more likely to accrue in slower, harder to navigate layers: domain-specific infrastructure, long-horizon scientific workflows, and composite systems where progress is measured in experiments rather than demos. Yet these ventures at the wave’s edges remain underfunded because they resist the narrative clarity that capital optimizes for. The next breakthrough will arrive where no one is looking.

The point is not that investors should hallucinate conviction where there is none. Capital will not suddenly abandon legibility; its constraints are structural. But those constraints can be shaped through fund structures that tolerate longer feedback cycles. Those willing to fund the weird, the early, and the difficult-to-explain historically unlock outsized impact. When capital is structured to tolerate ambiguity, it tends to surface a more meritocratic set of bets by default; the ideas that survive are the ones with genuine technical substance rather than those optimized for funding.

Capital warps attention, attention warps imagination, and imagination warps the future. Capital should invest in the unintelligible, reward the opaque, and fund the fog. Recognize the warp, and perhaps bend it toward broader horizons.