I used to think deal flow was a network problem. Then I watched a $4M Series A I should have been in.

01 · The old journey

For ten years, I sourced deals the way every angel does.

I’m an engineer first. Fifteen years of shipping production code, half of those years on infrastructure or AI plumbing. I started writing small angel cheques somewhere around deal #5 of my life — €5K to €25K, mostly into devtools and AI infra, mostly through warm intros from other engineers I’d worked with. The thesis was simple: I know technical founders, I can recognise good code, I’ll see early-stage software companies before the general-purpose investors do.

For ten years that worked, kind of. I got into a few good rounds. I missed dozens. I told myself the misses were because I lived in Athens, not San Francisco; because I didn’t go to enough dinners; because my network wasn’t senior enough. The obvious explanations.

They were the wrong explanations. The real one took a small fintech startup to show me.

02 · The wall

The Series A I watched in real time, from the wrong side of it.

In late 2024 I was tracking a small fintech team. Three founders, one repo, beautifully boring product. No press. No AngelList buzz. Nobody in my network had mentioned them. I had drafted a three-line email I never sent, the kind that starts with “I saw your repo and I love what you’re doing with the settlement layer.” The email sat in drafts for two weeks.

On a Saturday I idly opened their GitHub org. Their commit velocity had tripled in the prior fortnight. Four new contributors had joined. They’d spun up three new infrastructure repos. None of this was visible anywhere except on github.com. Nobody in my warm-intro network was talking about them. Nobody on AngelList had posted about them. They were not, in any sense, on the radar yet.

Three weeks later they announced a $4M Series A led by a top-tier fund. The investors who got in had either been told by a warm intro — fine, but slow — or had been watching the same data I had and acting on it. The data was free. It was public. It updated in real time. I didn’t need a better network. I needed a better lens on data I already had access to.

That’s when the three false beliefs started to collapse.

03 · False belief #1 — collapsed

“The right answer is to grow my network.”

I had spent ten years trying to fix a network problem. I went to conferences. I cold-DMed founders. I introduced myself in Slack groups. I wrote tweets. None of it solved the actual problem, because the actual problem wasn’t my network — it was the sequence. By the time a deck reached me through any warm intro, no matter how senior, three other investors were already in the room. The network was getting me to the table. It was not getting me there first.

Once I named that, the question became: what gets me there first? The fintech team had told me the answer without saying anything. Their code was visible weeks before any human heard of them. Engineering acceleration was not noise. It was sequence. I just needed a system to read it.

04 · False belief #2 — collapsed

“Public data is too noisy to be useful.”

The first thing every investor said when I described what I was building was “isn’t GitHub data just noise?” A fair question. Raw commit counts are noisy. Bots inflate them. Hackathons spike them. A single developer pushing config files looks the same as a team shipping features. The whole point of the objection is that the data is visible to everyone, so it can’t be edge.

But that argument falls apart the moment you stop looking at absolute numbers. Quant hedge funds make billions on SEC filings — data anyone with a browser can read. The edge isn’t the access. The edge is the lens. When you stop reading commit counts and start reading acceleration relative to a company’s own baseline, the noise drops out and the regime changes show up. We ran the regression on 219 startups. The 21-to-47-day lead time held. SSRN-indexed it. The lens existed. Nobody had built the product around it.

05 · False belief #3 — collapsed

“If this worked, the big incumbents would already do it.”

The third objection was the one I had to argue with myself. If GitHub momentum is real edge, why hadn’t Crunchbase or Affinity or Harmonic built it already? They have engineering teams. They have data scientists. They’ve had a decade.

The honest answer surprised me. The big incumbents are optimised for the funnel they already monetise — the warm-intro funnel. Their customers are funds with hundreds of associates whose job is to work the deck queue. A 47-day lead time isn’t a feature for them; it’s a different funnel they’d have to retrain their entire customer base on. The opportunity for a new product is precisely the funnel the incumbents won’t cannibalise. That’s why category-creators always win the next platform — the people serving the old one are too profitable to disrupt themselves.

07 · Identity shift

I stopped trying to know more people. I started trying to read better.

The shift isn’t tactical. It’s identity. A warm-intro investor optimises for who they know. A code-side investor optimises for what they read. The first compounds with seniority and dinners. The second compounds with rhythm and tooling. They feel different from the inside. The second one scales without your physical presence in any room. That’s the move I want my readers to make.