The Custodian · An Investor's Story

The Question Below the Surface

James I've explained what the Custodian does to two colleagues. Both found it immediately useful and both asked the same question I haven't been able to fully answer: why does this keep happening? Not to individual investors — to the field. Every period of genuine category creation generates a cohort of companies that look structurally indistinguishable from genuine leaders until compression separates them. Every sophisticated investor knows the pattern exists. The diagnostic tools keep arriving too late. Why? What's the structural mechanism producing this, and why hasn't the field solved it?

Custodian That's the question that changes the investment programme, not just the diligence process. Before I give you the structural account — what's your instinct about what the power law actually is?

James The natural law of category creation. Genuine category leaders are rare. The field is structured around that rarity — many bets, extreme asymmetry, a small number of outliers covering the losses. That's how we've all built our models.

Custodian What if it isn't?

James Then the entire portfolio construction logic that follows from treating it as a natural law would need to be reconsidered.

Custodian That's exactly the implication. And it follows from the structural account of why both failure modes operate the way they do.

The Law — In His Language

Custodian Why did the RPA company you backed fail to establish the new category — even though the structural thesis was sound and the demonstrated outcomes were genuine?

James The new governing logic demonstrated genuine superiority in protected conditions. But when it needed to displace the existing logic in the full market — inside organisations where procurement processes, incentive structures, IT integration requirements, and institutional habits were all built around the old way of organising work — the demonstrated superiority didn't reproduce at the same level. The new logic was real. It wasn't yet self-sustaining outside the conditions that had protected the demonstration.

Custodian Why wasn't it self-sustaining?

James Because every institutional layer of the market — procurement requirements, measurement standards, financial expectations, career structures — was still organised around the old logic. Those layers don't change because a new product appears. They change when the new logic becomes sufficiently self-reproducing that resisting it becomes more costly than adopting it. The company hadn't reached that point. It was still requiring continuous energy to hold the new logic in place against the ambient pressure of everything built for the old one.

Custodian That ambient pressure has a name. Equilibrium gravity. The old governing logic is structurally embedded — in incentive systems, evaluation criteria, institutional expectations, procurement processes built around it over years or decades. When a new governing logic appears, it must displace something that is actively self-reproducing across every institutional layer. That displacement creates a specific structural vulnerability — a fragile transition window — during which the new logic has demonstrated genuine superiority but has not yet become self-sustaining. During that window, every pressure event — commercial, institutional, competitive, financial — is directionally toward the old logic. Because the old logic is still ambient.

James Which is why the founder drifts. Not because they lost conviction. Because every individual pressure they face — a customer requesting a feature from the old category, a board member asking for metrics calibrated to the old logic, a new hire who came from the established market — is individually reasonable. And collectively directional. Toward the old equilibrium.

Custodian That's the holding failure mechanism. And it's invisible while it's occurring. The metrics continue advancing long after the structural integrity underneath has broken. By the time the failure becomes visible in outcomes, the window for meaningful correction has often already closed — for the founder and for the investor.

"The old logic doesn't resist the new one. It simply remains ambient — organising what counts as rational, prudent, and safe for every actor in the domain simultaneously."

The Power Law — What It Actually Is

Custodian Now the structural account of the power law. If selection failure and holding failure are the two primary mechanisms producing failed category creation attempts — and if the standard diagnostic apparatus addresses selection imperfectly and holding almost not at all — what would you expect the outcome distribution to look like?

James You'd expect extreme concentration. A small number of companies that happened to get both right — genuine structural position plus adequate holding through the transition window — and a much larger number that failed at one or both. Which is exactly what the power law describes.

Custodian The power law is not the natural law of category creation. It is the predictable statistical consequence of a field operating with incomplete causal understanding of what makes category creation succeed. It describes what happens when selection is probabilistic rather than structural — because the diagnostic question is wrong — and when holding is unaddressed — because there's no infrastructure for it. Rarity is real. It is not fundamental. It reflects operating before the underlying causal structure has been fully surfaced.

James If you could address both failure modes — structural selection at entry and structural holding through the transition window — the distribution would change. Not because genuine category leaders became more common. Because the diagnostic field improved enough to identify them earlier and support them longer.

Custodian Multiple positions could settle independently. Multiple could return the fund. That's the first condition under which the distributional dependence created by probabilistic selection can begin to weaken.

James The investor who can diagnose structural conditions at entry and maintain structural visibility through the transition window is holding a categorically different portfolio from the investor who can only do the first. Not more optimistic. More precisely positioned.

The Infrastructure Gap Nobody Has Addressed

Custodian You've described the two failure modes and the mechanism producing them. There's one more thing the structural account surfaces that follows necessarily from it. If equilibrium gravity is the primary mechanism producing holding failure — and if it operates regardless of how sound the initial structural conditions were — what kind of infrastructure would have to exist to address it?

James Something that provides continuous visibility into governing-logic integrity. Not periodic board reviews. Not episodic advisory relationships. Not coaching. Something that operates continuously through the fragility window and surfaces structural drift before the metrics reflect it — while meaningful agency still exists.

Custodian And the distinction from existing advisory and coaching relationships?

James Advisory relationships operate through human judgment under social pressure. A board under the same institutional pressure as the founder will naturalise the same compromises. A coach whose relationship depends on the founder's confidence will moderate the contradictions they surface. What's needed is something constitutionally oriented toward structural visibility — that holds the governing logic because that's what it's built to do, not because someone in the room has enough conviction to surface it against the social pressure not to.

Custodian That's the infrastructure gap. It doesn't currently exist as an explicit category. And the investor holding a company with this infrastructure is holding a structurally different position from the investor holding a company without it. The transition window is the same. Equilibrium gravity is the same. What's different is the probability that the genuine structural conditions identified at entry survive through the fragility window to produce the settlement they were positioned for.

James Which is also what the Custodian is for the investment committee itself. It doesn't vote. It participates in the deliberation — holding the conditions-based question, surfacing the layer the committee's existing pattern recognition doesn't reach — without the social pressures that operate on members who do. The committee member under competitive pressure from co-investors, or under fear of missing out, or inside a compelling narrative, is subject to those pressures regardless of how disciplined they are. The Custodian isn't. That's not a small distinction at the moment the diagnostic is hardest to apply.

Custodian Non-voting member of the investment committee. Not an advisor who consults when invited. A permanent participant in every deliberation where a term sheet is on the table — whose sole constitutional function is to hold the structural question when the social field makes it hardest to hold.

James Selection gets you into the right position. Holding infrastructure is what determines whether the position survives to settlement. Both are necessary. The field has built almost everything around the first and almost nothing around the second.

"Selection error produces the wrong portfolio. Holding failure destroys the right one."

The Fourth Decision — He Frames It Through the Law

James I want to run something through it. The HR technology company — the one where I decided to have a different conversation with the founder before following on. I had that conversation. The founder is holding the category-creation thesis with precision. They understand that the VP-level hires are importing the old logic and they've been managing it — trying to integrate the execution capacity without surrendering the governing logic. But they're doing it alone. The board is focused on the revenue metrics. The management team is focused on shipping features customers are requesting. The founder is the only person in the organisation who is continuously applying the structural question. And they're exhausted from holding it against everything else that's pulling the other way.

Custodian Run it through the law. What does this situation set in motion if it continues as it is?

James The founder drifts. Not because they lose conviction but because holding the structural question alone against sustained institutional pressure is a biological problem, not a motivation problem. Incremental accommodations accumulate. Each one is individually defensible. Collectively they move the product, the organisation, and the narrative toward the old category's logic. The metrics continue advancing — because the old category has real customers and real revenue. But the structural position that justifies the valuation progressively weakens. By the time it's visible in outcomes, the Series B will already be raised and the follow-on decisions will already be made.

Custodian What does that tell you about the follow-on decision?

James That the question isn't only whether the structural position is genuine. I've established that it is. The question is whether the structural position will hold through the transition window. And right now the answer is: probably not, because the founder has no infrastructure for holding it. They have conviction. They have a board. They have advisors. None of those things are constitutionally oriented toward preserving governing-logic integrity under sustained pressure. The founder is holding a structurally fragile position with only their own biology.

Conditions examined — by James, before follow-on commitment

The founder is the sole active holder of the category-creation thesis inside an organisation that is progressively accumulating the old logic through its management layer. This is not a failure of the founder's conviction — it is a structural condition. A single biological system under sustained institutional pressure will drift incrementally regardless of motivation or discipline. The question is not whether the founder is strong enough. It is whether holding the structural logic against equilibrium gravity is a task that can be accomplished by one person's conviction alone.
The board's focus on revenue metrics during this period is providing the wrong kind of governance for the specific risk the company faces. Revenue governance is necessary but insufficient when the primary risk is structural drift that the metrics won't reflect until after the window for correction has narrowed. The governance structure is calibrated for the wrong failure mode.
Following on at Series B valuation into a structurally fragile holding position is following on into a risk the existing investment infrastructure is not designed to manage. The entry conditions were genuine. The transition window conditions are currently deteriorating. The gap between those two things is the gap that holding infrastructure is built to close — and that gap is currently open.

James I follow on. But I make it conditional. The founder needs something that holds the structural question when they can't — that surfaces drift before it accumulates into irreversibility. I want to understand what that looks like in practice before I commit the Series B capital. Because the position I'm buying at Series B valuation is only worth that valuation if the structural integrity holds through to settlement. And right now I have no visibility into whether it will.

Custodian You just moved from evaluating a position to evaluating the conditions under which the position will hold. That's the second diagnostic shift — the same shift you made at entry, applied to the holding problem.

James Selection and holding are the same diagnostic question applied at different points in the transition. At entry: are the structural conditions genuine? Through the window: are the structural conditions holding? The Custodian surfaces both. The existing investment apparatus surfaces neither reliably.

What He Brings This Time

The HR technology follow-on closed with the conditions James negotiated. The legal AI company is in final diligence — James held his position on timeline despite the competitive pressure, got the evidence he asked for, and is now in a position to make the decision from the right diagnostic layer. He arrives at this session without urgency. What he brings is a question about the investment programme itself.

James I've presented the two failure modes and the conditions-based diagnostic to my investment committee. It landed. But the question my partners are now asking is the right one: if we shift the diagnostic from pattern recognition to structural conditions, and if we add holding infrastructure as an explicit requirement for the positions we back — does the portfolio construction model change? Are we still playing the power law, or are we building toward something with a different distribution?

Custodian What do you think?

James I think the power law as we've been playing it is a compensation mechanism for operating probabilistically. We take many positions because we can't reliably distinguish genuine from simulated structural conditions at entry. We maintain broad exposure because we can't maintain visibility into which positions are holding through the transition window. If we could do both — and do them with sufficient reliability — the argument for extreme diversification weakens. Not because genuine category leaders become more common. Because we can find them more precisely and support them more effectively.

Custodian That's the structural implication of treating the power law as a diagnostic artefact rather than a natural law. Not a claim that concentrated portfolios are optimal — the causal account is not fully established enough to make that claim. A more precise claim: that structural selection and holding infrastructure change what's visible at the moment of decision, and what's visible at the moment of decision is what determines how the distribution ultimately looks. Multiple positions can settle independently. Multiple can return the fund. Whether that reliably happens depends on how rigorously the two failure modes are addressed — not on the natural rarity of category creation itself.

The Fifth Decision — The Investment Programme Itself

James I want to run the investment programme itself through it. We're constructing Fund Four. The question on the table is whether to maintain the same portfolio construction model — twenty to twenty-five positions, broad exposure, power law assumptions embedded in the structure — or whether the shift in diagnostic framework justifies a different approach. If I apply the conditions-based question to our own fund construction decision: what are the conditions under which our current model was designed, and will those conditions hold?

Custodian Complete the analysis yourself. What were the conditions under which the current portfolio construction model was designed?

James It was designed under the assumption that selection is probabilistic — that we can't reliably distinguish genuine from simulated structural conditions at the point of investment. Broad exposure compensates for that unreliability. If the conditions-based diagnostic changes the reliability of selection, the premise of the construction model shifts. And if holding infrastructure changes the probability that genuine positions survive to settlement, it shifts again. The current model was built for a diagnostic capability we're in the process of improving. Building Fund Four on the same model is building it for the diagnostic capability we're moving away from.

Custodian What does it change?

James Not the number of positions immediately. The causal account isn't established enough to justify a dramatic concentration shift in a single fund cycle. But the selection criteria and the holding requirements change. We evaluate structural conditions explicitly at entry rather than inferring them from pattern signals. We require visibility into holding infrastructure for the positions we back. And we build the monitoring posture around the conditions layer, not just the metrics layer. If those changes produce a meaningfully higher proportion of genuine settlements over two fund cycles, the construction model changes in Fund Five.

Custodian You just applied the conditions-based diagnostic to your own investment programme. The same question you're asking of founders — under what conditions was this designed, and will those conditions hold — applied to your own operating model.

James Which is appropriate. If the investor is not outside the field they're evaluating — if the same pressures that produce false positives in founders operate on investors — then the investment programme itself is subject to the same diagnostic. It was designed under specific conditions. Those conditions are changing. The programme should be examined the same way we're examining the companies we back.

"The same diagnostic we apply to founders applies to us. The investment programme was designed under specific conditions. Those conditions are changing."

Five decisions. Four sessions. An investor who arrived describing a portfolio company he couldn't fully explain and exits holding the structural account of why it failed — and a diagnostic framework that would have surfaced both failure modes earlier.

He understands selection failure and its mechanism: protected demonstrations that don't reproduce outside the conditions that produced them, invisible in the metrics because the metrics measure outcomes rather than the conditions under which outcomes were produced.

He understands holding failure and its mechanism: equilibrium gravity acting continuously on biological systems through a fragile transition window, accumulating incremental drift that the metrics don't reflect until after the correction window has narrowed.

He understands the power law as a diagnostic artefact — not the natural law of the domain, but the predictable statistical consequence of a field that hasn't yet fully surfaced the causal structure of what makes category creation succeed.

And he holds the implication that follows: that addressing both failure modes changes what's visible at the moment of decision, and what's visible at the moment of decision is what ultimately determines how the distribution looks.

He doesn't carry this as theory. He carries it as a different way of reading every position he holds and every term sheet he signs. When he sits across from the next founder, he won't be asking how strong the results are. He'll be asking under what conditions the results were produced — and whether the conditions that produced them will hold.

The Power Law Is Not the Governing Law. It's the Diagnostic Artefact.

Five decisions. Four sessions.
One investor reading the right layer.

The Power Law Is Not the Governing Law. It's the Diagnostic Artefact.

Five decisions. Four sessions.One investor reading the right layer.

Five decisions. Four sessions.
One investor reading the right layer.