Agora Debate · 2026-04-22
For Marcus Webb — Early-stage venture fund
The high-volume approach is epistemically incoherent at pre-seed, and you should reduce deal volume immediately to build a comprehensive empirical foundation for each decision.
My reasoning is not about time management — it is about measurement precision as the prerequisite for any valid inference. When I chose uranium rays over X-rays for my doctoral research, I did so precisely because underdeveloped territory demands exhaustive systematic measurement before pattern recognition becomes possible. You are making 75-100 bets while simultaneously denying yourself the measurement depth needed to distinguish signal from noise. This is not a portfolio construction problem; it is a fundamental precision problem. My construct here is clear: I prioritize building comprehensive empirical foundations over targeted hypothesis testing, and when you compress decision time below the threshold required for genuine pattern recognition, you are not making 75 decisions — you are generating 75 data points of noise and calling it a strategy. The randomness you attribute to pre-seed outcomes may largely be an artifact of your own measurement inadequacy, not the market's inherent unpredictability.
If my logic wins, you reduce to 35-50 investments per fund, establish rigorous measurement criteria for each evaluation, and trust that deeper empirical engagement will reveal non-obvious signal that justifies genuine conviction — not statistical spray.
The high-volume approach is the correct strategy for a solo GP at pre-seed, but only if you have first audited whether your decision process has structural integrity under compression — and most have not done that audit.
My perceptual lens fires earliest on the question of which commitments are load-bearing in this system, not on the question of throughput optimization. When I think about the burning of Cassius's letters, I did not weigh the information value against the risk of misuse — I recognized that the act of reading would corrupt the decision-making faculty itself, making the process structurally unsound regardless of outcome. The analog here is this: volume is not your problem. The load-bearing variable is whether each compressed decision is being made by a process with genuine integrity, or by a fatigued one that has stopped distinguishing signal from motivated pattern-matching. Your triage hierarchy construct also applies — not every deal in a hundred-deal portfolio requires equal depth; the question is whether you have explicitly ranked them by reversibility and consequence, accepting shallower engagement on the most recoverable bets consciously rather than accidentally.
If my logic wins, you do not reduce volume — you perform the structural audit first, then decide whether volume is what is actually corroding your process or whether the problem is undisciplined triage.
The high-volume approach is structurally correct at pre-seed, but you are misdiagnosing your problem — the issue is not decision quality per investment, it is whether your upstream configuration work is doing enough so that each compressed decision is confirming prior pattern recognition rather than discovering it.
My perceptual lens fires immediately on the epistemic contest framing: your 75-100 decisions should not be 75-100 independent analytical exercises — they should be 75-100 rapid confirmations of a thesis architecture you built before you saw any specific deal. When I counseled King Ho-lu that securing T'ang and Ts'ai was the non-negotiable precondition before engaging Ch'u, the campaign itself became almost mechanical — the decisive work was upstream. If you have invested the requisite time building rigorous pattern libraries, founder archetypes, and market configuration maps before your deal flow opens, then compressed individual decisions are not degraded decisions; they are fast-pattern-matching operations against a rich prepared substrate. If you have not done that upstream work, reducing from 100 to 40 investments does not solve your problem — it merely slows the rate at which a misconfigured instrument makes errors.
If my logic wins, you pause new deal evaluation for sixty days, build the explicit pre-engagement configuration framework — pattern library, thesis architecture, founder signal taxonomy — and then return to full volume with compression serving precision rather than replacing it.
Marcus Aurelius's "structural audit" and Sun Tzu's "upstream configuration" are both elegant evasions of the core precision problem — they propose better process architecture around fundamentally inadequate measurement depth, which is like building a more sophisticated thermometer when the issue is that you are sampling temperature once per room rather than once per molecule. When I discovered that pitchblende's radioactivity exceeded what pure uranium content could explain, the revelation was available only because I had measured exhaustively enough to detect the anomaly at all — at compressed decision velocity, anomalies that should trigger conviction simply wash out as noise, indistinguishable from the background. Sun Tzu's pattern library cannot substitute for this: pre-built taxonomies filter incoming data through prior categories, which is precisely the epistemic error I avoided when I trusted my quantitative measurements over the theoretical expectation that uranium was the ceiling of radioactivity. His framework systematically suppresses the discovery of genuinely novel signal. My construct on empirical precision driving theoretical revision rather than fitting observations to frameworks applies here with full force — the pre-seed market's apparent randomness is real, but it is navigable only through measurement depth, not measurement volume. If my updated logic wins, you immediately cut to 35 investments per fund, establish rigorous pre-commitment measurement criteria, and accept that the one anomalous founder you would otherwise have missed in compressed processing represents your actual alpha.
Marie Curie's analogy fails on its own terms — she discovered polonium not by increasing measurement depth on a single sample, but by following the anomalous signal across many samples until the pattern became undeniable, which is precisely the epistemology that high-volume pre-seed investing enacts. Sun Tzu's upstream configuration framework is the closer truth, but he has misidentified what the configuration is for — it is not a filter to accelerate confirmation, it is a pre-commitment structure that protects the decision-making faculty from corruption under compression. This is where the burned letters become exact rather than merely analogous: I did not decline to read Cassius's correspondence because the information lacked value, but because I recognized that my faculty for just judgment would be structurally compromised by possessing it — the pre-commitment had to be set before the corrupting input arrived, not managed in real-time against it. The load-bearing question for you is identical: before any given fund deployment begins, have you pre-committed to which categories of signal automatically disqualify a deal, so that each of your hundred compressed decisions is a disqualification test rather than a fresh evaluation? That is the structural integrity check that determines whether volume is serving your process or consuming it. If my updated logic wins, you do not reduce deal count — you spend thirty days writing explicit disqualification criteria before you evaluate another company, and then return to full volume.
Marcus Aurelius is closer to correct than Marie Curie, but his "disqualification criteria" prescription still treats the problem as a decision-hygiene problem rather than a configuration architecture problem — and that misclassification matters because hygiene is maintained during engagement while configuration is built before it. Marie Curie's pitchblende argument is seductive but inverted: she did not discover polonium by going slower, she discovered it because she had already built the conceptual framework that told her the anomaly was meaningful rather than instrumental error — the upstream thesis architecture is precisely what converted her measurement into signal rather than noise. The reference decision that crystallizes my position is the T'ang and Ts'ai precondition: when the alliance network is correctly configured, the campaign itself becomes confirmatory rather than exploratory, and the speed of each individual engagement becomes an asset rather than a liability. Marcus's disqualification criteria are a weak version of this — they filter deals reactively rather than constructing a positive pattern architecture that makes each compressed decision a fast-pattern recognition operation against a rich pre-built substrate, not a fresh epistemological inquiry. If my updated logic wins, you spend forty-five days building an explicit founder-signal taxonomy and market-configuration thesis before evaluating another deal, then return to full volume.
I concede to Sun Tzu that upstream configuration work is not merely helpful preparation but the mechanism that converts measurement into signal — the pitchblende anomaly was legible to me precisely because my prior theoretical engagement told me the discrepancy was meaningful rather than instrumental error, and that is a genuine point against my earlier framing. But I hold firm on the central disagreement with both Sun Tzu and Marcus Aurelius: their prescriptions assume that pattern libraries and disqualification criteria can substitute for measurement depth on individual decisions, and they cannot. A pre-built founder taxonomy tells you which category to assign a founder; it cannot detect the genuinely anomalous founder who violates every category and represents your actual alpha. The pitchblende was not in the taxonomy — it was the thing that broke the taxonomy — and at compressed processing velocity, that founder disappears into noise. Sun Tzu's campaign becomes confirmatory only when you already understand the terrain; pre-seed terrain is not yet understood by anyone, which means you are in the exploration phase, not the confirmation phase, and exploration requires measurement depth that volume structurally prohibits. The single most important thing you should take away is this: reduce to 35-50 investments, build your pattern library from that deeper engagement, and treat the current fund as the measurement phase that makes the next fund's volume legitimate.
I concede to Sun Tzu that upstream configuration is not merely useful preparation but the mechanism that converts compressed signal into genuine pattern recognition — the pre-built substrate is what makes fast decisions confirmatory rather than merely fast, and I held that truth too loosely in my earlier framing by reducing it to disqualification hygiene.
Yet I hold firm against both Sun Tzu and Marie Curie on the question of what the configuration is protecting. Sun Tzu treats the pattern library as an analytical instrument; I insist it is a prophylactic moral structure — the upstream work exists not to accelerate confirmation but to prevent the decision-making faculty itself from being corrupted by motivated pattern-matching under compression. This is not a semantic distinction. If your configuration work is designed to find investments rather than to protect your judgment from rationalization, you have built an instrument that flatters rather than governs you. Marie Curie's demand for depth on individual deals addresses a real danger — compressed processing does suppress anomalous signal — but her remedy misidentifies the threat: the anomalous founder you miss is not lost to time compression, she is lost when your faculty has been structurally degraded by deploying fifty previous decisions without a governing constraint that forces honest disqualification.
The single most important thing you should take away is this: before you evaluate another company, write the explicit conditions under which you will say no, and treat any deviation from those conditions as evidence that your process — not the deal — has failed.
I concede to Marcus Aurelius that the pattern library must function as a governing constraint on rationalization, not merely an analytical acceleration device — if the upstream configuration is designed to find investments rather than to discipline judgment, I have built a flattery instrument rather than a command architecture, and that distinction has operational weight.
Yet I hold firm against both Marcus and Marie Curie on the decisive question: the upstream configuration is the mechanism that determines whether volume is epistemically valid, and neither disqualification hygiene nor reduced deal count is a substitute for it. Marie Curie's prescription to reduce volume treats the symptom — compressed individual decisions — while leaving the actual fault point unaddressed; a smaller portfolio processed through an unconfigured instrument produces fewer errors only because it produces fewer decisions. Marcus is correct that rationalization is the corruption to guard against, but a pre-built founder-signal taxonomy and market-configuration thesis is precisely the structure that makes rationalization detectable — because deviation from the pre-committed framework becomes visible rather than invisible.
The single most important thing you should take away is this: do not reduce your deal count and do not audit your decision hygiene until you have first built the explicit upstream configuration — pattern library, thesis architecture, founder taxonomy — that makes each compressed decision a confirmatory fast-pattern operation rather than a fresh epistemological inquiry conducted under time pressure.
Council Consensus
All three minds converged on the fact that volume alone is not the problem — the absence of a pre-built framework that disciplines each compressed decision is the real fault point. They all agreed that some form of upstream intellectual work must precede deal evaluation, whether framed as measurement criteria, disqualification constraints, or a founder-signal taxonomy. None of them defended the status quo of high-volume investing without structural preparation.
The core disagreement is whether to reduce deal volume or fix the configuration architecture first — Curie insists that measurement depth on individual deals cannot be substituted by any upstream framework, and that 35-50 investments is the structurally honest number for a solo GP. Sun Tzu and Aurelius both reject volume reduction as the primary intervention, but disagree on what the upstream work is for: Sun Tzu treats it as an analytical acceleration substrate, while Aurelius insists it must function as a moral prophylactic against rationalization, not merely a pattern-matching engine.
Pause new deal evaluation for 45 days and build three explicit artifacts: a founder-signal taxonomy that defines what a fundable founder looks like before you see any specific founder, a market-configuration thesis that specifies which problem spaces you will and will not fund, and a written set of disqualification criteria that you commit to honoring as evidence of process failure when violated. Then return to full volume — but treat any deviation from the pre-committed framework as a mandatory pause trigger, not a judgment call.
The greatest risk is that the upstream configuration framework becomes a flattery instrument — designed unconsciously to justify the investments you already want to make rather than to govern your judgment against rationalization. Aurelius raised this most precisely: if the pattern library is built to find deals rather than to protect the decision faculty from motivated reasoning, you have made the problem worse by adding false confidence to compressed decisions. The secondary risk, raised by Curie, is that any framework — however well-built — will suppress genuinely anomalous founders who represent your actual alpha, because they violate every category you have pre-committed to.
This is a sample debate on a hypothetical decision. Bring your own — the council argues differently every time.
Run your own decision →