Pre-Demo Day investing sounds like conviction. It is. But it's conviction built on a specific checklist — five things we verify before we commit, every single time. Miss one and we pass. Hit all five and we move fast.
Here's exactly what we're checking.
Check 1: Founder Pedigree That Is Directly Relevant
We don't care about prestige for its own sake. We care about credentials that are evidence of domain mastery — specifically mastery of the exact problem the company is solving.
Akira, co-founder of Arga Labs, is 17 years old and a former Goldman Sachs quant. Not impressive in the abstract — impressive because he built digital twin infrastructure that automates 90% of the twin creation process. The Goldman background isn't what matters. The pattern-matching skills from quantitative finance applied to physical-world simulation — that's what matters.
Ayman Saleh of Chronicle Labs spent 8 years at NASA JPL on the Mars rover landing system. His co-founder helped return the Boeing 737 Max to service. These founders understand failure modes in high-stakes systems that ordinary software engineers never encounter. Chronicle builds production simulation for enterprise AI agents — and only a team with aerospace credentials would think to build it this way.
The TesterArmy founding team came out of Vercel. They built developer tooling at the company that redefined developer tooling. When they said they were building the developer-native QA product, we believed them immediately.
The question we ask: does this founder's background make them the single most qualified person in the world to build this specific product right now? If the answer is yes, we go deeper.
Check 2: Enterprise Validation — Real Contracts, Not Pilots
There's a spectrum of enterprise interest: signed contracts, paid pilots, unpaid pilots, LOIs, 'warm conversations.' We only count the first one. Everything else is a hypothesis.
Prototyping.io came to us with Tesla, Zipline, Rivian, Generac, and Light Matter as paying clients — before they had raised a dollar of outside capital. These aren't pilot customers. These are the most demanding manufacturing companies in the world writing actual checks.
Astraea had enterprise contracts signed before their raise. Not LOIs. Not letters of intent. Executed agreements with healthcare systems for clinical trial data automation at 99.5% accuracy.
The enterprise validation check exists because it's the hardest thing to fake. An enterprise customer's procurement team doesn't sign contracts with companies they're not serious about. If a YC company has one or two real enterprise contracts in hand before Demo Day, that's an extraordinary signal.
Check 3: Existing Revenue — Money Already in the Bank
Revenue projections tell us what a founder hopes. Existing revenue tells us what the market has decided. We want the latter.
Prototyping.io was profitable and bootstrapped when we backed them. Not pre-revenue with a compelling model — actually profitable. For a pre-seed company in a YC batch, that's almost unheard of.
Arzana's anchor customer, Milltown Paper, was already paying. Manicule had a full pipeline before their raise formally began. They were turning away inbounds. That's not a revenue projection — that's a revenue constraint.
We don't need the revenue to be large. We need it to exist. A company with $10K MRR at the pre-seed stage has proven something that a company with $0 and a great pitch has not.
Check 4: A Technical Moat That Takes Time or Data to Replicate
The best early-stage technical moats aren't patents — they're datasets, proprietary architectures, or hard-won accuracy benchmarks that took years to build and can't be reproduced quickly.
Astraea's 99.5% accuracy on clinical trial data didn't come from a clever model. It came from years of building training datasets from healthcare systems that are notoriously difficult to access. A new entrant can't replicate that in 12 months.
Arga Labs' 90% automated digital twin creation is built on proprietary simulation pipelines that the team spent their entire careers developing. The moat isn't the idea — it's the 10,000 hours of domain knowledge baked into the training data.
TesterArmy's vision-based QA agent requires no scripts, no selectors, no setup — it perceives your UI the way a human does. Building that perception model is genuinely hard. The ex-Vercel team had spent years thinking about developer workflow before building it.
Check 5: A Timing Window Where Incumbents Are Sleeping
The best investments don't fight established players — they find the moment when the market is ready but the incumbents are still asleep. That window rarely stays open for more than 18-24 months.
Astraea's timing is textbook. SAS and its peers — the entrenched players in clinical data software — won't have a competitive AI-native offering until 2027-2028. The hospital systems know it. That's why enterprise contracts are already being signed with a YC startup. The window is open now.
TesterArmy is in the same position. Enterprise QA is dominated by legacy tooling that requires dedicated QA engineers, complex script maintenance, and weeks of setup. TesterArmy's developer-native approach is so different that legacy players can't pivot to it — it would require rebuilding from scratch.
The timing check asks: why is this the right moment? If the answer is 'because AI is getting better,' that's not specific enough. We want to know exactly which incumbent is asleep, exactly why they can't respond quickly, and exactly how long the window stays open.
Why All Five Matter
Four out of five isn't enough. A company with a brilliant technical moat and perfect timing but no existing revenue might be too early. A company with revenue and enterprise validation but no timing edge might be entering a market that's about to get crowded. Each check exists because we've seen companies fail along each dimension.
When all five align — and in the current batch they aligned for 11 companies — we move before Demo Day. The crowd will show up in a few weeks. We'd rather be there when the price is still right.
