The Copy Problem
Coverage was always for sale. A faithful copy of judgment never was.
Everyone in venture knows the beloved page one of the oldest firms in the country keeps, cataloging the great companies it passed on. The most famous entry has a partner who, offered an introduction to the two Stanford students building Google in his friend’s garage, asked how he could get out of the house without going anywhere near the garage. Every firm keeps a private mental version of their own list, my firm NextView included. But every company on it at least reached a partner who knew it existed and deliberately said no.
The longer list, the startups the firm never saw at all, is the one no one keeps. Its biggest entries are no secret, since everyone knows the winners that got away. But a deal you never got near does not sting like one you looked at and declined, so no one writes it down. That which remains is still the longest list in the building.
The famous list was written by judgment. The tacit shadow anti-portfolio was written by the calendar.
The limiting factor in any venture firm is a senior partner’s time, and that time is expensive. A bad pick costs the check, and the check is capped. A miss costs the multiple that got away, and in a business where six percent of deals produce sixty percent of the returns, the multiple that got away is frequently the fund. Judgment only ever works on what attention puts in front of it, and attention runs a few partners’ calendars deep.
The industry knows all of this, which is why the largest firms attack coverage with leveraged structure. They build pyramids of associates screening upward, run accelerator programs that standardize intake (nineteen thousand pitches in a recent cohort, for fifty to seventy slots), and deputize scout networks to write small checks beyond the partnership’s reach.
Coverage has always been for sale. It just takes resources, and the management fees on ever-larger funds pay for them.
But every one of those structures is a copy of partner judgment. The accelerator copies it into a standardized screen, trading resolution for volume. The scout carries a copy out of the building entirely. The junior investor pyramid copies it by apprenticeship, the slowest copy of all, and still a lossy one.
I know that copy from the inside, because I was one. I came up in the venture business as a junior investor under two mentors who could not have been less alike. One was an intuitive, nonlinear thinker with a nose he could never quite put into words, the kind of taste that made him an early backer of one of the era’s breakout consumer companies. The other was relentlessly linear, sizing up market and product and whether a company was truly venture scale or merely a good business, and ruthless once a bet was working, piling in when warranted and moving aside any founder who couldn’t keep up. The judgment I use now is an amalgam of both their eyes plus what I have learned since, and it is a lossy copy of each. No apprentice carries across everything his teachers had.
Venture has never had a way to scale judgment except by copying it, and every copy was lossy. That is the copy problem.
The problem exists because judgment comes installed in people, and people come with calendars. For the entire history of the asset class those two facts were one fact. Buying attention at scale meant accepting worse judgment, and protecting judgment meant rationing attention. Every coverage structure above was rent paid to the bundle.
What AI changes is the copy. Codified judgment is digital, and it duplicates at constant fidelity. Founder number two thousand gets screened by the same judgment as founder number twenty. No pyramid, no program, no scout network has ever offered that, and no apprenticeship ever could.
The objection a skeptical peer will raise is that a perfect copy of mediocre judgment is mediocrity at scale. Copy the screen and you copy its blind spots, at ten thousand companies instead of two hundred. It’s the right objection, and most of it stands, because the codified screen is only as good as the judgment that went into it. What it misses is that a codified screen can be inspected, measured against outcomes, and improved in place. The copy in an associate’s head can be audited only the way the anti-portfolio audits it, in hindsight, one regret at a time, years later. Whether codification captures enough of the real thing is a fair fight, and coming posts will take it on.
Solve the copy problem, though, and the bottleneck moves back to where everyone always claimed it lived: judgment. Under the bundle, improving judgment was low-leverage work, because a sharper eye still only saw the sliver the calendar admitted. Unbundled, every improvement to the screen compounds across everything it touches. The pass that later raises at a markup stops being a war story told at dinner and becomes a labeled error the screen learns from. Judgment stops being rationed and starts being optimized.
The firms that win the next decade will not be the ones that see the most companies, because “seeing” has already stopped being scarce through automated data players like Harmonic, Specter and Gravity. The winning firms will be the ones that improve and codify their judgment the fastest.
That is also why this AI disruption to VC is bigger than a tooling upgrade. The modern venture firm is a machine built around the bundle. Fund size is partner count, times the deals a partner can carry, times the check. The Monday partner meeting exists to ration the firm’s scarcest shared resource: senior attention. The board seat prices that attention into every lead check. The warm intro, our most cherished ritual, is at bottom a rationing device, dressed up as wisdom about trust. These are load-bearing walls with nothing left to hold up. The firm worth designing now starts from a blank blueprint.
