Everything, everywhere, all at once

For the last half-century, the venture capital asset class has generated immense alpha through geographic and social arbitrage. If you were in the right circles on Sand Hill Road, you saw the deal; if you were not, you did not. "Proprietary deal flow" was the industry’s primary moat; an investment craft of information asymmetry where access was a core source of the alpha.

But as we settle into the market reality of 2026 in an AI world, that moat has not merely been eroded; it has been digitized, indexed, and distributed. The craft of VC has already changed, and we think that is a good thing.

The era of "finding" the deal is almost over (probably entirely over within the next 24 months). With the proliferation of LLM-driven scrapers, GitHub watchers, and automated talent tracking, the "time-to-discovery" for a new startup has collapsed from months to minutes. Every investor will soon see everything, everywhere, all at once. The one caveat being true inception stage companies with no online presence, but even signals for that exist by tracking talent networks.

A breakout open-source repository or a stealth-mode founder updating their LinkedIn is no longer a secret; they are data points broadcast instantly to the entire VC market. Not every investor is currently tuned in yet, but many are, and soon everyone who will stay relevant will have to be plugged into these feeds. The game has changed (and that is good).

Now that information has become ubiquitous, the investment market has more valuable tools to find great opportunities. And with all of the data signals out there, there are three observations here on how this access is affecting the venture market:

• AI sourcing is automating outreach: Traditionally, deal flow has relied on warm intros. However, AI tools are augmenting these sourcing channels by adding automated signals (GitHub repositories, product hunt launches, LinkedIn job changes, domain registrations) to identify founders before they fundraise.
• The cost to find a deal is collapsing: In 2010-2020, discovering a stealth founder required a lot of heavy lifting. Today, agentic workflows are not only adding new channels but also dramatically reducing costs and increasing flexibility for VC firms (fewer FTEs). This is driving the cost of deal discovery down considerably.
• Consensus trading is in full swing: Private market data is no longer private, and the signals of high-growth companies are now ubiquitous data points tracked by hundreds of firms at once. This has led to extreme consensus trading. While early-stage deal counts remain flat, global late-stage funding surged 66% year-over-year in Q3 2025, where just 18 companies captured one-third ($32B) of all global venture capital (Crunchbase).

Alpha will be in the Uncomputable

We are quickly moving towards a world where every investor sees almost everything. In this reality, the alpha generation will have to rely more on in-person, offline, and qualified networks. On the founder side, the best ones are also inundated with outreach and noise, leading them to want to retreat to high-trust, low-noise networks.

This is where the true opportunity lies: in the Uncomputable. While AI is exceptional at pattern matching against historical data, it is fundamentally incapable of underwriting many of the qualitative and human elements of investment decisions (critical at the early stage, still important at the late stage). Looking ahead, we believe it is likely that many of the next generational companies will defy many of the predictive models. The outliers break rules, and humans are experts at that. They will have been deemed uninvestable by any model, and investors who rely too heavily on such models will be left behind.

Specifically, the alpha is migrating toward sectors and traits that resist digitization:

• Physical product companies: AI has an uncanny ability to analyze the digital, yet analyzing and evaluating companies whose product is inherently physical in the real world is still a challenge. Additionally, many physical products lie in complicated industries that have regulatory, government, health, reputational, or existential human risk. This area of investing will struggle to rely on the digital signals that AI is so good at today.
• A founder’s "Grit": AI can measure code output, but it cannot measure a founder’s ability or likelihood to survive a supply chain collapse. Most startups pivot industries, product lines, and strategies, so an AI’s evaluation of a founder's current idea is not necessarily indicative of whether that founder will be successful with their company in 5-10 years. Who knows where their strategy might take them (a human can evaluate this better).
• Contextual nuance is a human advantage: AI is pattern matching data in digital form, but AI is not living with the weight of decisions that are made by founders on a daily basis as they steer their companies through uncertain waters.
• Speed does not outweigh insight: additionally, founders are placing higher value on investors who were not simply the “first to email them” but also came prepared, got up to speed, and brought their experience at other companies into their conversations with founders (and in the board rooms). This human element is an advantage that AI is not eroding; it is actually strengthening it by letting investors focus more on understanding rather than finding “access” to deals.

Consequently, deal sourcing is quickly becoming a sorting function between “in the flow” and “out of the flow” good deals. This erodes the benefit of personal networks as a source of access, yet does not erode the trust and human elements that still reside in those personal networks. This ubiquity of sourcing then leads to more emphasis on the human element of alpha generation by venture investors. This is what founders want anyway, so we see this as a net add to the founder/investor dynamic.

Takeaway: the more technology we apply to the investment process, the more emphasis it puts on the alpha generation being intensely human. The problem of information and “access” is being solved (quickly), yet we are not close to solving the problem of judgment (nor should we). As the venture market witnesses a flood of capital chasing the high-signal "consensus" bets, the greatest returns will be found in the offline, human, and uncomputable corners of the market that the algorithms simply cannot see.

‍