data-driven VC is over

deep research, the art of vc, and "wisdom"

Welcome. I’m Lawrence. A pleasure. I invest in computing hardware at pre-seed and seed stage. Traditional CMOS semiconductors are increasingly inadequate for AI workloads—both for energy-hungry training in data centers and low-power inference at the edge. To sustain computational growth, we need breakthroughs in architectures, designs, and materials that maximize performance per watt. I focus on emerging paradigms like photonic, analog, neuromorphic, temporal, quantum, and reversible computing, which rethink computation at the physical level. lawrence@lunar.vc.

---

So it’s probably over. I had a good run, I suppose. Imagine Jean-Baptiste, a master silk weaver in Lyon, France in 1805. For twenty years, he had worked his way from apprentice to master, developing an intuitive understanding of how threads interacted, how tension affected patterns, and how to translate artistic designs into woven reality.

His workshop employed several apprentices who looked to him not just for instruction but as the living embodiment of a craft tradition stretching back generations. His identity was fundamentally intertwined with his expertise—he wasn’t just someone who wove; he was a weaver in the deepest sense.

Then came Jacquard's machine with its punched cards and automated pattern control. Suddenly, patterns of extraordinary complexity could be woven by operators with minimal training. The cognitive component—the very heart of his expertise—had been externalized into a mechanical system. Deus ex machina.

And here I sit. My identity fundamentally intertwined with my expertise, I am not someone who just researches; I’m a researcher. It’s a noble profession. We used to spend our days in libraries. Now we sit in coffee shops with 100 tabs open. But for how much longer? The answer: zero point zero years.

By now, those of us on the frontier are using Deep Research daily. Four years ago, when I started “State of the Future,” I spent about 20 hours researching each technology. VTOL: 20 hours. MPC: 20 hours. Xenobots: 20 hours. Space Elevators… you guessed it, 21 hours.

I previously wrote: “we are now on the cusp of the age of zero marginal costs of digital production. It’s text and images now, but obviously, it’s code, video, and then more complex productions. If you make digital things for a living, it’s probably already over for you.”

That’s me! I make digital things for a living. But when Devlin came for the software developers, I stayed silent. When it came for the filmmakers, I said nothing. It was easy for me to say, yeah, sorry lads, it’s over. A lack of empathy, I think.

But it’s more of an emotional gut punch to see at least a decade of competency collapsed into an agent. I am actually good at researching complex stuff, synthesizing it, and making it accessible. But as of 2025, AI is better at it. Sorry, lad.

So, speaking from the coalface of the augmentation versus automation debate, here’s my attempt to think about what happens when my core competency is automated in VC.

Here’s my argument:

1. DR has commoditized “thesis development” and “market research” in VC so larger AUM and more resources no longer an advantage

2. Everyone will rely on the same research and converge on the same opportunities (defencetech! agents!)

3. To outperform, VCs will make “irrational” decisions based on non-quantifiable “gut-feel” (crypto again?, rocket cargo? temporal computing?)

4. In a world of commoditized research, human wisdom is scarce and one of the remaining competitive advantage

5. Winners = solo GPs and emerging managers with tight focus and industry relationships + "wisdom”

---

Democratization of Research

First, this is classic democratization. Something that was a specialised skill-set and expensive is now cheap and fast. Previously, lots of the thesis-development (what should be invest in?) And ‘diligence’ (is this company good?) needed analysts and associates. Is rocket cargo a thing? Why is photonic memory so difficult? Etc. More AUM meant more resources to do this work OR those with less AUM could do fewer deals.

The structural barriers that protected information advantages have dissolved. Tiger Global's legendary research operation—once a differentiating force with hundreds of analysts producing proprietary market maps and competitive landscapes is over. Today, a solo GP with access to Deep Research can generate comparable analysis in hours, not months, and at a fraction of the cost. The moat of information superiority has evaporated.

The traditional knowledge diffusion timeline is also collapsing. When Sequoia identified the emerging API economy in 2015, it took nearly two years for this thesis to permeate the broader VC consciousness. With research agents, such insights propagate almost instantaneously, creating synchronized awareness across the industry about emerging opportunities like the current wave of agent-native applications.

So think about crypto, where resources and capital are allocated much more mimetically than traditional venture capital because of the always-online, 24-7 nature of the industry. DR is likely to result in the same for VC as trends/memes rise and fall faster. DeFi? DeSci? DePin? DeBeers? (lol)

DR has commoditized “thesis development” and “market research” in VC so larger AUM and more resources no longer an advantage

The Convergence Problem

As we rely on same agents trained on overlapping data, we will see the emergence of an efficient market theory of venture capital—a troubling prospect for an industry built on information asymmetry.

Investment theses are homogenizing at unprecedented speed. In 2022, when generative AI burst onto the scene, we watched as virtually every major venture firm published *remarkably* similar market maps and opportunity landscapes within weeks of each other. Research agents will accelerate this convergence, with algorithmic analysis driving firms toward identical conclusions about opportunities.

The competition for deals is intensifying as identification of promising startups becomes democratized. Founders report receiving nearly identical outreach messages from dozens of venture firms within hours of each other—all citing the same market data points, competitive analysis, and growth projections derived from increasingly standardized research processes. “Hey, I love your project, do yoy have 20 mins to jump on a call?”

I don’t know if Lindy works properly yet, but for sure, we have DR + Snov + zerobounce + personalised email agent. The pipeline from research to phonecall will collapse requiring no human input.

And why stop there? Everyone already uses Claude/ChatGPT to prep for calls, so why not send the prep questions in advance of the call. Why not prepare a memo in advance. Why not get your agent to speak to the founder’s agent to share all relevant information. What are we left with here? Everyone with the same analysis and information.

Everyone will rely on the same research and converge on the same opportunities.

"Irrational" Decision-Making

As data-driven decision-making becomes commoditized, we're entering a counterintuitive era where decisions that appear to contradict available evidence may generate the greatest returns.

Are we back to the good old days of actual contrarian thinking? When a16 invested in Airbnb, market analysis suggested the short-term rental market was too niche and regulatory barriers too high. Their willingness to override consensus analysis—precisely the kind that research agents excel at producing—led to one of venture capital's greatest returns. In tomorrow's landscape, such "irrational" bets may be even more valuable.

The most valuable investments increasingly come from domains where structured data is least available. Benchmark's early investment in Uber came not from market sizing data (which was pessimistic about the taxi industry) but from Matt Cohler's personal experience with the service and recognition of an emotional response that no market report could capture. Research agents struggle with exactly this type of experiential, emotion-based insight. (At least, right now and likely for 12 months at least)

What are we left with? Investing for emotional reasons. Or for reasons that cannot be supported by data. This is likely to be difficult for bigger funds with more institutional decision-making processes managing more AUM. Gut feel is easier for angels, micro-VCs and maybe solo-GPs. Emerging managers to tend to outperform peers in terms of returns, and this dyanmic makes me think this will become true for the next few vintages.

To outperform, VCs will have to make “irrational” decisions based on non-quantifiable “gut-feel”

Wisdom as the New Scarcity

As research becomes abundant, wisdom is emerging as the scarcest and most valuable resource in the venture decision-making process, creating a new basis for differentiation.

Wisdom operates as a meta-layer above conventional knowledge. While information tells us "what is" and knowledge explains "how things work," wisdom addresses the profound question of "what matters and why." It's the difference between knowing how to build something and understanding whether it should be built at all. Subtweet on AGI obviously. Claude tells me wisdom is:

• Pattern recognition across domains—the ability to see resonances between seemingly unrelated fields. A wise person notices how ecological principles apply to organizational dynamics, or how childhood development illuminates political movements.

• Contextual activation—knowing which knowledge applies when, and perhaps more importantly, which doesn't. The academically brilliant person who makes terrible life choices lacks this contextual intelligence.

• Emotional integration—the synthesis of cognitive understanding with emotional intelligence. Wisdom isn't cold calculation but embodied understanding that incorporates feeling as a form of data.

I conceptualize this as "lived experience as training data" where success and failure become encoded not just as memory but as embodied understanding. Wisdom isn't mere accumulation—you can live decades and remain unwise. The differentiator is reflective practice, the computational processing that transforms raw experience into structured understanding.

Externally, wisdom looks like as decisiveness without impulsivity—passing on perfect-on-paper deals because something feels misaligned, or backing technically flawed pitches because you sense exceptional founder adaptability. These aren't random emotional reactions but sophisticated processing systems integrating multiple streams: micro-expressions, narrative coherence, and pattern recognition from past encounters.

Consider Roelof Botha's 2004 investment in YouTube. Basically a company wrapped around copyright infringement with no obvious business model. The metrics screamed 'pass,' but supposedly Botha spotted an unusual user engagement pattern that overrode the more obvious risks. It’s hard to specifically say this was “wise”, but certainly it was courageous . His wisdom-driven decision to override analytical red flags resulted in a 60x return 18 months later for Sequoia. Decent.

This sounds esoteric, I know. But these "woo-woo" human elements remain computationally intractable in 2025. The alchemy of intuition, emotional processing, and contextual judgment resists neat algorithmic encapsulation.

Will AI eventually replicate wisdom? Absolutely. With expanded context windows, multimodal training, and embodied experience, systems will synthesize these capabilities. But today, human wisdom remains the computational function hardest to externalize.

In a world of commoditized research, human wisdom is scarce and a competitive advantage

The Art of VC

As Deep Research commoditizes VC’s analytical side, success will hinge more on intuition and contrarian conviction. The winners in this landscape won't be the mega-funds with massive operations, soz a16z (not that they care, they are playing the real-quiz of government takeover anyway) but the solo GPs and emerging managers who combine tight industry focus with relationship depth and that most elusive quality: wisdom.

These nimble investors operate at the intersection of specialized knowledge domains, moving with a velocity and conviction that institutional processes can't match. They're not constrained by consensus-seeking partner meetings or rigid investment theses, allowing them to act decisively on insights that larger firms process too slowly.

The solo GP with deep founder relationships in a specific sector doesn't just see different data—they see the same data differently. They recognize patterns others miss not because of superior analytical frameworks, but because their lived experience has calibrated their pattern-recognition systems to detect signals invisible to outsiders or algorithms. Their wisdom isn't just knowing more facts; it's understanding which facts actually matter in contexts where data alone cannot provide certainty.

This evolution doesn't diminish human investors but rather clarifies their unique value: the courage to be wrong for the right reasons, the vision to see potential where data indicates risk, and the conviction to back founders whose ideas challenge conventional wisdom. As we enter this new era, perhaps the greatest irony is that AI research tools, designed to make us more data-driven, may ultimately elevate the most distinctly human aspects of investment decision-making to new levels of importance.

It was the Art of the Deal after all, not the Data-Driven of the Deal.

Winners will be solo GPs and emerging managers with tight focus and industry relationships + "wisdom”

You will note, I am now a research-led investor not a researcher and investor. Got to keep the wolf from the door for a bit longer…