198 companies. 10 weeks. 17 investment decisions. That's the W26 batch in numbers for Eight Capital.
Backing 8.5% of a YC batch is not a small commitment. It requires a process — a real one, that survives contact with time pressure, competing demands, and the fog of a 198-company batch where the best ones aren't always obvious at first glance.
Here's how we actually did it.
Weeks 1-2: The Full Batch Scan
We review every company in the batch. All 198. This sounds exhausting — it is — but it's non-negotiable. The best companies in a YC batch are not always the ones with the loudest signal on day one. Some of them are quiet, heads-down, and slightly under the radar.
The first-pass filter is deliberately blunt: does this problem matter at scale? Is there any early signal of traction? Does the founder's background suggest they're uniquely qualified to solve it? If the answer to all three is yes, the company moves to our watch list. If any answer is clearly no, we move on.
At the end of two weeks, we typically have 55-65 companies on the watch list. That's still a lot. The hard work is ahead.
Weeks 3-5: Deep Dives on the Watch List
For every company on the watch list, we run a structured deep dive. This means reference calls with customers (not just 'do they exist' — what did they actually pay and what problem did it solve?), technical diligence on the product architecture, and market research on the incumbent landscape.
The question we're trying to answer at this stage is not 'could this work?' — almost anything could work. We're asking: 'is this the right moment, right team, right market, and right traction level to commit capital now?'
We use what we call the "already" test: what has already happened? Not what they plan to do, not what they expect — what has already been built, sold, paid for, and deployed? The best companies at this stage have a long "already" list. Prototyping.io already had Tesla. Astraea already had signed enterprise contracts. Arzana already had a customer that had referred two new customers in one week.
By the end of week 5, we've narrowed from 60 to roughly 20-25 serious candidates.
Weeks 6-8: Term Sheets
This is where speed matters most. The pre-Demo Day window closes fast. A company that's at $3M pre-money in week 6 will be at $8-10M post-Demo Day if the pitch goes well. We've learned — sometimes painfully — that hesitation at this stage is expensive.
We move on term sheets in week 6 for companies we have conviction on. Not 'we'd like more information' — conviction. The information we have at this point is as good as it's going to get before Demo Day. More time doesn't produce more clarity; it just produces more competition.
The term sheet conversations at this stage are also a signal. How a founder responds to a term sheet before Demo Day — their urgency, their transparency about other conversations, their clarity on use of proceeds — tells us a lot about how they'll operate as a portfolio company.
Weeks 9-10: Close and Watch Demo Day
By Demo Day, we've typically closed 12-15 of our 17 investments. The remaining 2-3 close in the week after Demo Day — not because we changed our view, but because the paperwork takes time.
We watch Demo Day carefully. Not to find new investments — at this point, the best entry prices are already gone — but to calibrate. How does the narrative that companies present compare to what we found in diligence? The answer tells us something about founder judgment and self-awareness that we carry into the portfolio company relationship.
What We Got Right (and What Surprised Us) in W26
Three things about W26 surprised us relative to our models going in.
First: the industrial companies were more technically sophisticated than we expected. In previous batches, industrial startups often had the domain knowledge but lacked the software depth. In W26, Prototyping.io and Arzana both had both — and their enterprise customers reflected it.
Second: the repeat founder quality was unusually high. We typically expect 30-35% repeat founders in a batch. W26 was 38%, and the quality of the repeat-founder companies was above average even within that group.
Third: the B2B AI agent wave moved faster than we modeled. Companies that we thought were 12-18 months from enterprise readiness had already signed contracts. The market's willingness to adopt AI-native tools in regulated industries — healthcare, manufacturing, financial services — accelerated sharply in the six months before W26.
That acceleration is what made W26 one of the strongest batches we've evaluated. The companies weren't just good on paper — they had market confirmation that the window was open. We moved accordingly.
