Operators Becoming Investors

Most venture capital is written by people who have never built the thing they fund.

This was less of a problem when software companies followed predictable patterns. A SaaS playbook could be diligenced from outside. Distribution metrics, cohort retention, sales efficiency — these are visible from across a table. You did not need to have shipped a product to evaluate one.

AI breaks that pattern.

The hard parts of building AI in production are not visible from the outside. They live in routing decisions, recovery behavior, evaluation pipelines, vendor terms, latency budgets, and the slow accumulation of workflow data that makes the difference between a demo and a system that earns trust. None of that shows up in a deck. Most of it does not show up in customer references. Some of it does not even show up in the founder's pitch, because the founder is too close to the work to know which parts a generalist would miss.

This is the gap that has pulled a generation of operators into investing.

What operators see

When you have spent years building production AI in regulated industries, you develop a specific set of instincts that are hard to acquire any other way.

You learn what brittleness looks like in a pitch. The founder says the model handles edge cases. You know the question to ask: what happens on the fifth user turn when they change their mind halfway through. You learn what a thin moat looks like. The founder says they have proprietary data. You know how to ask whether the data compounds, who labels it, and what happens when the underlying model improves.

You learn what a real workflow integration is, versus a Zapier-level glue layer that will be ripped out in eighteen months. You learn what compliance posture looks like when it has been thought about from day one, versus retrofitted under pressure.

You learn which architectural choices buy optionality and which create lock-in. You learn what the actual cost of running a multi-model production system is, including the parts that do not show up in the AWS bill. You learn how long it takes to recover when a foundation model provider changes their terms.

None of these instincts come from reading about AI. They come from being responsible for the system when something breaks at two in the morning. And in 2026, that experience is one of the rarest and most valuable inputs into early-stage AI investing.

The shift in who can see what

For most of venture's history, the capital advantage was concentrated. A small number of firms with brand, network, and access could see deal flow that no one else could. The competitive edge was access.

That has been changing for a while, and AI is accelerating the shift. Deal flow is more democratic than it has ever been. What is becoming scarce is not access. It is judgment.

Specifically, judgment about which technical claims hold up, which workflows have real moats, which founders actually understand the work, and which companies are building something that will still be standing in three years when foundation models have absorbed most of the surface area their competitors are building on.

That judgment is hard to develop without operating experience. Reading about a thing is not the same as building it. Watching twenty pitches in a sector is not the same as having shipped one product in that sector to a regulated customer.

This is why the most interesting funds being built right now are being built by operators. Not because operators make better investors automatically, but because operators see things that generalists cannot see at the speed the AI cycle requires.

What I see now

I spent four years at Google working on large-scale machine learning. I spent the last few years building production AI in healthcare and financial services. I was the person responsible for routing decisions, vendor terms, multi-model architecture, and compliance posture in environments where mistakes had real consequences.

When I look at AI startups, I look at them through that lens.

I look for whether founders have a real point of view on architecture or are repeating talking points. I look at which workflows can actually be automated and which ones will need human-in-the-loop for the foreseeable future. I look at how companies are absorbing the structural risks of foundation-model dependency at the infrastructure layer, or whether those risks are about to transmit through to their customers' compliance posture.

I look for founders who understand that the model is the easiest part of the system, not the company.

These are not theoretical observations. They come from time inside the work. And they are the lens that shapes how 434 VC will evaluate the companies we back.

The other thing operators bring

Beyond judgment, operators bring access to a different network.

The relationships I built at Google, at Sprinter Health, at Goodfin, in the AI infrastructure community, and at fintech conferences are not the same relationships a career investor builds. They are deeper in specific places and thinner in others. They include the senior engineer who has shipped multi-agent systems in production, the compliance officer who knows how a regulator will respond, the wealth advisor who can tell you which workflow tools their firm has tried and rejected.

For founders building vertical AI in regulated industries, that kind of network can be more useful than a generalist VC's network of other VCs. The introductions point toward operators, buyers, and design partners — not just toward more capital.

This is the part of investing that institutional firms find hardest to replicate. They can hire operators. They can build platform teams. They can sponsor events. What they cannot easily build is the kind of trust that comes from having sat in the same seat as the founder, two or three years earlier, with the same constraints and the same stakes.

Why I think this generation matters

The AI cycle is moving faster than any technology cycle in software history. New model releases reset competitive landscapes every few months. Foundation model providers shift terms, pricing, and capability frontiers on quarterly clock cycles. Regulatory regimes are forming in real time.

In this environment, the founders who win will not be the ones with the best decks. They will be the ones who can react fast and well to a landscape that is changing under their feet. They need investors who can react with them.

That is hard to do from outside the work. It is hard to do without lived experience of having to make these decisions yourself. The funds that move at the speed of this cycle will be the ones run by people who have been builders recently enough that the speed feels normal.

This is the bet behind 434 VC. The founders I want to back are operators who have been close enough to the work to see what generalists miss. The judgment I want to bring is the kind that comes from having built production AI in a domain where trust matters.

The line between building and investing has been blurring for years. In this cycle, I think it has finally disappeared.

The best investors of the next decade will be people who can build. And the best builders will think like investors from the start.