3 things I've learned, 3 years into VC

December 30, 2025

2025 was an unimaginable year for me.

When I joined my current firm 3 years ago, I was not new to VC, but I was new to investing (I'd spent 2.5 years at another VC's ops team, and another 3 years raising money for startups).

And upon joining I was told:

3 years later, the very first investment I made (seed round, zero revenue, invested via both our firm and with a lot of my personal capital) turned unicorn — Gamma, hitting $100 million ARR and $2.1 billion valuation with its Series B raise).

And it's not just Gamma. Several other investments have done incredibly well, e.g. with revenues shooting past the valuations I invested at.

What have I learned throughout all this?

In every org there are things that nobody seems to want to do, and it's easy to look at that pile and say: "If no one wants to do it, obviously it's not worth my time either."

Don't do that.

When I joined my current firm, I didn't have much going on. Others had lots of investments to manage, board meetings to attend, demo days to judge, deal flow, etc.

Me? I didn't have all that, so I had to hustle where I could: organizing LP events, writing newsletters, setting up fund admin software, figuring out how to automate things, and looking through piles of decks that others hadn't dug into.

And what a goldmine that turned out to be.

For one, ChatGPT exploded in late 2022, and suddenly I was the one using it + other no code tools like Zapier to automate tedious workflows within our firm, because well… I was the person handling these routine stuff, and I needed help. Things like newsletter signups, news monitoring, and writing blog posts became fully or semi-automated.

Very quickly I was the one that knew the most about the SOTA capabilities of AI and automation, because I'd spent weeks painfully testing everything, running into bugs and missing integrations.

Whenever LPs or much bigger partner firms came to us to ask how we were adopting AI, I was the one they talked to.

And then there was that massive pile of companies that others on our team hadn't gotten to.

These are mostly early stage US startups, courtesy of our relationship with the incredible Hustle Fund (we're an LP), and it's pretty tiring for non-native English speakers go sort through them. So I did it.

I went through ALL of those companies, one by one. Thousands of them, for weeks. And from those, I picked out ~8 companies I'd focus on.

Gamma was one of them.

And so was my second best investment in terms of both revenue growth and profitability.

Lesson: don't dismiss tasks just because others haven't done them or don't seem to want to.

Building automated systems at this historical moment is actually pretty frustrating, simply because the tech is changing too fast.

You can spend days solving a specific problem in the workflow, and the next week one of the foundational LLMs releases a new update and makes it all go away. You feel like you've wasted all that time.

But I've come to a different view.

Because I've had to build stuff from scratch, piecing together multiple tools (and trying many more to find the right tools), that entire ordeal did two things for me.

The first is agency: when I succeeded, and the 20-step workflow finally worked like magic in outputting a high-quality investment memo, I gained immense belief in my own abilities, and in the thesis that persistence and hard work pays off. The fact that Perplexity Labs came out months later and could do all that in a single prompt doesn't matter. My agency had already been strengthened.

The second is cognitive wiring: because I had to find, learn, and try all of those tools and how to fuse them together, the process in effect wired my brain. Yes, you could probably put all that into a masterful prompt and get an LLM to generate a master plan far better than what I did, but there's something fundamentally different in putting all that together yourself.

Here's an example:

Over months, I built and evolved a system to source startups in Japan. Whereas it took our interns a month to process 300 companies, now I can sort through >1,000 in 3 days. The process looks like this, and I'll mark out clearly where things are automated vs. manual

I've used this system to consistently identify companies that we otherwise would not have even looked at (hint: it's where a business and its growth signals don't line up with our prior beliefs).

Venture was always a "human-scale" business—you can only meet so many founders and read so many decks, right? Well, AI is changing all that.

I'm able to go through a thousand companies in a very short amount of time because the agents I've set up act as extensions of me: scanning thousands of sites, rewriting information in ways that I can process quickly, and re-ordering lists so I hit the high priority ones first.

And because I've built all this (and many other systems) from scratch, I know that if you dropped me into a new org and gave me no support, I could still build all of this back up.

Only faster and better, actually, because the state of the art has advanced.

Lately, there's been an explosion of posts on Twitter about the importance of agency. So I'm not gonna belabor this point.

I'll just say this: I was told I was on the "wrong side of 40" to be starting in venture. In this industry, that's usually code for "you lack the stamina", and "you can't compete against 20 year olds".

They were right about biology, but wrong about agency.

Agency isn't age-related.

I don’t have the luxury of a 30-year career arc anymore. I have a young child and a 10-20 year window to prove myself and secure a better life for my family. That hasn't made me slower; it's made me more focused.

When you have a young child and a ticking clock, and your mornings and evenings are entirely absorbed by childcare and house chores (our elders aren't around to help), you've gotta be hyper-efficient. You stop networking for the sake of networking, and you stop attending fancy events and social dinners to feel good about yourself.

And you stop waiting: You stop waiting for the firm to hand you deal flow. You find your own. You stop waiting for IT to give you a system. You build your own.