In startup circles, "YC-backed" is still a gold badge, but YC's batches are inflating faster than tips in New York. YC now runs four batches a year, and 2026 is on track for roughly 770 companies, including 142 in Fall so far.
At this rate, just keeping up with YC companies is heavy work. So I decided to play a diligent investor and dig into all 142 companies in the Fall 2026 batch, using nothing but public information.
The goal
Let's set the goal: find the companies most likely to raise a priced Series A within 18 months.
First, think about this: what type of investor would set a goal like this?
In my observation, two kinds:
- Emerging managers, who need markups to raise their next fund. A priced Series A is the clearest early one.
- Platforms that give individual investors early access. Their investors want visible progress and, eventually, a way out, and a priced round gives them a milestone and a path to liquidity.
In the end, it comes down to whether the Series A investors say yes. They weigh a lot of things before the "yes", but two are hard to get around, one about now and one about what comes next:
- Now: the demand has to be real. It starts with a buyer who already pays for what the product replaces, so the decision to buy is easier and faster. And it has to be verifiable: paying customers and live deployments are hard to fake; adjectives aren't.
- Next: the company has to keep winning as things change. Both the product and the market will keep changing. The product will keep evolving as customers use it, and founders who know the problem firsthand, or bring a rare skill, are better placed to steer it. The market will shift too, as competitors and model labs move in, and only a lead that gets harder to copy over time survives that.
How I screened
These concerns became my four tests. The first two rest on facts and the last two on judgment, so I ran them in that order and counted only public evidence. A company was out at the first test it failed.
| Criterion | Test | Remaining |
|---|---|---|
| Start | All Fall 2026 companies in the YC directory | 142 |
| 1. Replaces existing spend | The product replaces a specific cost already in the buyer's budget, such as a role, an outside service or a tool. | 97 |
| 2. Verifiable demand | Someone outside the team has paid for the product or put it into live operations: named paying customers or a deployment with numbers. Logo walls, free trials, waitlists, GitHub stars and investor backing don't count. | 14 |
| 3. Founder–market fit | A founder has worked inside the problem or brings a scarce technical skill. Prestigious credentials with no link to the problem don't count. | 10 |
| 4. Moat | The company gets harder to replace as it grows, through proprietary data, regulated workflows or network effects. Speed or a better interface alone doesn't count. | 5 |
What's not on the list. Market size. A Series A investor will ask about it, but at batch stage a market estimate is hard to verify from outside, and under an 18-month clock, proof beats potential. A fund hunting for outliers would put it first.
A note on hardware. Hardware companies usually need more capital and a longer product and customer validation cycle to get from seed to Series A, so they sit less naturally with an 18-month path to a priced round. That shows in the screen: 29 companies, about a fifth of the batch, build hardware or physical systems, and none made it past the demand test. The one I would watch first is Rodeo Robotics, which builds robots for vehicle maintenance and says it earns service revenue while collecting robot training data, though it hasn't disclosed customers or volume.
What I found
In the end, only 5 of the 142 companies passed all four tests.
But the bigger cut came earlier. Of the 97 companies that cleared the first test, only 14 got through the second. Most could describe what they're building, but couldn't name a customer.
Those 14 are worth watching:
| Company | Industry | What it does |
|---|---|---|
| Ascenix | Healthcare | Clinical intelligence for skilled nursing facilities |
| CONUS | Healthcare | AI knowledge layer for medical device companies |
| Covera | Insurance | AI operating system for insurance brokers |
| Decent | Insurance | Health insurance brokerage for small businesses |
| Elimentary | Fintech | Agents for private capital back-office work |
| Forward | Fintech | Structures private credit deals and routes them to funds |
| Grep AI | Operations | Agents that automate operational work |
| Ignition Benefits | Insurance | AI-native benefits brokerage |
| Invertix | Energy | AI agents that run solar, wind and storage operations |
| Lyon | Fintech | Foundation models on enterprise transaction data |
| Quivly AI | Customer success | AI agents for customer onboarding and renewals |
| Shire Intelligence | Restaurants | Runs restaurants from CCTV in real time |
| Sintropix | Accounting | Autonomous accounting platform |
| Sutrix | Healthcare | Turns hospital EHR data into research datasets |
And the five that passed all four:
| Company | Who pays | Why it fits | Traction | Biggest risk |
|---|---|---|---|---|
| Ascenix | Long-term care operators | • Replaces clinical staff and consultant time • Founders built Fallyx for long-term care • Embedded in compliance workflows | • Named case study: 80% less time on fall analysis • Ten operators shown on its site | EHRs that nursing homes already run could build the same analytics |
| CONUS | Hospitals and device makers, via separate licenses | • Replaces regulatory and documentation work • CEO from healthcare consulting; product built with five chief surgeons • Holds each hospital's device knowledge | • Claims to be live in 20+ hospitals | Depends on device makers sharing their content; large makers could keep it in their own apps |
| Forward | Private credit funds and debt advisors | • Replaces analyst and advisor work • Founders from private credit • Network grows with every fund and advisor | • Claims $100M+ in deals originated | Large funds may not share deal flow, and matched parties can go direct |
| Invertix | Asset owners, power producers and O&M firms | • Replaces analyst labor • CTO trained in electrical engineering; strong founder-led sales • Learns from every asset it runs | • About 2 GW under management • Pilots with a 1.5 GW O&M provider and a 350 MW producer | Vendors holding operators' data could bundle similar agents; on-site engineering may keep it services-heavy |
| Sutrix | Hospital research teams | • Replaces manual data abstraction • CEO was a clinical researcher at HSS and Emory • Approved EHR access is slow to copy | • Names HSS, Emory and Mount Sinai as paying teams • Data prep cut from 3 weeks to 4 days | Early customers overlap with the CEO's former institutions; sales beyond them are unproven |
These five show promising signals, and a few question marks. The Fall 2026 batch is still ongoing, with Demo Day on December 2. As a more cautious "investor," I'll keep watching and take a closer look at a few of them along the way.
Now it's your turn: if you were the VC, what goal would you set, what criteria would you use, and which company would you back? Place your bets, and tell me why. We'll check back in 18 months.
Sources
- YC company directory, Fall 2026: the 142 companies screened
- FundingWatcher: Y Combinator statistics: 2026 batch sizes
- YC: 2026 Demo Day dates: Fall Demo Day on December 2
- Company details: each company's YC page and official website, as of October 2026
About this series
This series is how I look at early-stage companies: identify the question behind the business and keep testing it as new evidence comes in. All information comes from public sources as of October 2026; more Fall companies may still appear in the YC directory.
If you'd like the full spreadsheet, or want to trade notes on any of these companies, reach me on LinkedIn or at jz3938@columbia.edu.