Yesterday we confirmed GR Angel’s first exit.
In 2025 we invested in a YC company building in AI. It was making about $100K a year. Twelve months later it was at $7M ARR, and this week one of the major AI labs bought it, before the company ever raised a Series A.
It felt like the right moment to write down what the last two years have looked like.
Guillermo and I started investing together in 2025, and since then:
▫️ We’ve backed 53+ companies, 42 of them from Y Combinator
▫️ Those companies have raised $1.2B+
▫️ One of them, HappyRobot, is now worth $1.2B
▫️ And this week we had our first exit
Here’s how it started, what we’ve learned and what we’re doing next.
The first exit
We’re keeping the names private, both the company’s and the lab’s. Here’s what we can share: it’s a YC company building in AI, revenue was around $100K a year when we invested, a year later it was $7M, and one of the major AI labs acquired it.
The speed is what stands out. The usual path is seed, then a Series A, then maybe an exit years later. This team grew 70x in twelve months and sold before the Series A conversation started.
We think this will happen more often. AI labs want products that customers already love, and buying a company that has one is faster than building it.
To the founders: congratulations, and thank you for letting us be part of it.
How GR Angel started
GR Angel started as a side effect.
Guillermo and I both worked in venture before we invested our own money. Guillermo was an M&A lawyer at PwC, started a B2B SaaS company, then worked as a VC at GoHub Ventures. I was an analyst at Mundi Ventures and later ran investments and media at Itnig Capital.
From those seats we noticed something. The best founders could raise money. What they struggled with was getting noticed by customers, by hires and by the next investors.
Meanwhile we had both been writing for years. I write The VC Corner, The AI Corner and The Founders Corner, which reach 370K+ subscribers. Guillermo writes Product Market Fit and AI Market Fit, with 120K+ readers. Between us we have 600K followers on LinkedIn, 10M+ impressions a week and 500M+ a year.
So when we started writing cheques, we did the obvious thing: we wrote about the companies we backed.
The model worked almost immediately. One newsletter feature led to a $1M investment offer for a founder. One of my posts doubled another startup’s ARR.
That changed how founders saw us. In a hot seed round, the allocation is usually full before the round opens. Founders started making room for us anyway, because coverage in our newsletters was worth more to them than our cheque.
In 2025 we made 25+ investments, all with our own money. Some of them:
▫️ HappyRobot: AI agents for logistics, now a unicorn
▫️ Quiver AI: $8.3M seed led by a16z
▫️ Agaton: $10M seed
▫️ Azimov: $12.5M from a16z Speedrun and NFX
▫️ CodeWords (Agemo): €7.6M, with the CEOs of Miro, ElevenLabs, Personio, Zalando and Supercell as fellow angels
▫️ Mito AI: backed by Lightspeed
▫️ Throxy: $6.2M with Base10
▫️ Momentic: $15M Series A with General Catalyst
The post
By the summer of 2025 we could see a pattern. Our best investments and our best relationships with founders all came from Y Combinator.
On September 9, 2025, I posted this on X:
Hundreds of founders replied, and GR Angel went from an idea to a name.
The other thing that happened was that investors started writing to us. They had been reading about the companies we backed and wanted to invest alongside us. In 2026 we said yes to a small group of them. It was our first year with outside capital.
How it works
GR Angel is two people. Guillermo and I make every decision, so we can go from a first call to a wired cheque in under 24 hours.
We invest in about five companies from each YC batch, through direct investments and private SPVs. Since 2026 a small group of private investors invests alongside us.
After we invest, we do what we’ve always done. We write about the company, put the founders on our podcast and introduce them to customers and to the funds that read us.
The numbers
Four out of five companies in the portfolio come from YC. The other 11 are teams outside YC we liked too much to pass on.
Our co-investors across the portfolio include a16z, Sequoia, Founders Fund, Coatue, General Catalyst, Peak XV, Lightspeed, Khosla Ventures, CRV and Kfund. Paul Graham personally invested in 5+ of the same companies.
Why we focus on YC
People ask why we only back YC companies when we could invest across the whole seed market. The short answer is that YC does the first round of filtering for us.
Here’s how it works:
▫️ YC runs four batches a year, and around 1% of applicants get in
▫️ Every company gets the same deal: $500K, made of $125K for 7% plus $375K on an uncapped SAFE that converts at the next round’s price
▫️ After three months, founders pitch thousands of investors at Demo Day, and the best rounds fill in a few days
The returns back this up. YC tracked investors who backed at least three companies at every Demo Day from 2018 to 2020, and compared with Carta’s data on 2,906 US venture funds from the same years, the gap is huge:
▫️ The median YC Demo Day investor reached 5x TVPI. The median US venture fund from those vintages sits between 1.1x and 1.6x
▫️ The top 10% of YC Demo Day investors reached 16x, against 2.6x to 4.2x for the top 10% of funds
▫️ Even the bottom 25% of YC investors reached 3.3x, more than the median fund
▫️ 6.5% of YC companies become unicorns, against 2.5% for the average seed-stage startup on AngelList, and a quarter of those unicorns go on to pass $10B
AngelList’s own Demo Day Funds tell the same story: 7 of the 17 YC batches from 2015 to 2023 rank top decile, and Winter 2019 returned 11.5x net.
So the hard part of investing in YC is getting into the rounds. The good ones are oversubscribed, and founders pick who gets in. That’s where our model helps: YC finds the companies, the best funds lead, and our audience is why founders save us a spot.
Some of the companies
-
HappyRobot (YC S23). Pablo Palafox, Luis Paarup and Javier Palafox build AI agents for logistics, the calls, emails and messages that keep freight moving. We joined the a16z-led Series A, revenue grew 10x after it, and in August 2026 they raised a $150M Series C at a $1.2B valuation. DHL, Kuehne+Nagel and Uber are customers.
-
Replit (YC W18). Amjad Masad and Haya Odeh built Replit so anyone can make software by describing it to an AI agent. In March 2026 they raised a $400M Series D at a $9B valuation, and Sacra estimates they reached $525M in annualised revenue the following month.
-
Peec AI (Berlin). Marius Meiners, Tobias Siwonia and Daniel Drabo help brands see how they show up in ChatGPT and other AI search tools. They went from $0 to $10M ARR in 16 months, have 3,000+ customers including Zalando, Hugo Boss and Chanel, and raised a $21M Series A in November 2025.
-
QFEX (YC P25). Annanay Kapila and Joshua Wharton are building a trading platform for markets that trade around the clock. They raised a $9.5M seed at a $95M valuation before making any revenue, with General Catalyst and Paul Graham as investors.
-
Nozomio (YC S25). Arlan Rakhmetzhanov is one of the youngest founders we’ve backed. He’s building infrastructure that gives AI coding agents the context they need, and raised a $6.2M seed led by CRV, with Paul Graham and Gokul Rajaram.
-
Apollo Atomics (YC P26). Assil Halimi and Drew Walker spun out of MIT to build nuclear reactors in a factory. They raised a $31M seed and report more than 20 GW of signed LOIs.
-
GRU Space (YC W26). Skyler Chan is building the first hotel on the Moon. He reports $500M in letters of intent. It’s the most ambitious bet in our portfolio, and we’d make it again.
Others we’re excited about: Karumi, Didit, Lightsprint, RonanRx, Exosat, Omnia and InfiniteWatch. The full list is at grangel.co/companies.
Compound
This year we also started a podcast, Compound. Our episode with Mark Cuban got 500 hours of watch time in its first 24 hours. We’ve also had Chris Best, who co-founded Substack, and Staffan Helgesson, founding partner at Creandum.
Every guest is someone our founders can meet later.
I’m also an official scout for a16z Speedrun, a16z’s accelerator.
If you missed them, here are the Mark Cuban, Staffan Helgesson and Chris Best conversations.
What we’ve learned
-
Distribution is a better pitch than money. Every investor offers money. Founders gave us allocation in crowded rounds because of what came after the cheque.
-
Speed counts. With two people deciding, we can answer in a day. Founders remember who moved first.
-
Focus beat breadth. Our best investments came from one place, so we put most of our effort there.
-
Make the ambitious bets. A hotel on the Moon sounds absurd until $500M in letters of intent show up.
-
Exits can come early now. Our first one came before a Series A.
What’s next
Our investors asked to put more money in next year, so we’re growing GR Angel. Three things are changing.
▫️ More founders. We’ll keep backing the best AI companies in every YC batch. If you’re in a batch, or applying to one, email me at ruben@thevccorner.com and tell me what you’re building.
▫️ A few more investors. We’re talking to a small number of accredited investors who want to invest alongside us. If that’s you, email me at ruben@thevccorner.com.
▫️ More people. We’re hiring across sourcing, content and operations. Same email.
You can see everything we do at grangel.co.
Two years ago this was two people writing about startups and writing small cheques. Thanks to every founder who let us in.
Rubén and Guillermo
Portfolio figures combine our personal angel investments and investments through GR Angel. GR Angel is a private investment company. This newsletter is for information only and is not investment advice or an offer to sell securities. Past performance is not indicative of future results.
Sources: Fortune on HappyRobot’s $1.2B valuation · TechCrunch on Replit’s $9B round · Sacra on Replit passing $500M · AngelList Demo Day Funds · Orange Collective on YC returns · Carta VC Fund Performance Q4 2025 · AngelList unicorn rate study · Pulse 2.0 on Apollo Atomics
