YC W26: The Definitive Investor's Guide
What to Watch, What to Skip, and Where Lobster Capital Is Placing Bets
About Lobster Capital
Lobster Capital is an early-stage venture fund focused on seed investments in the top 2% of YC startups in each batch. We write initial checks of $250K and look for companies with strong founder conviction, early traction signals, and defensible technical moats. We attend every YC Demo Day, and for each batch we conduct systematic research well in advance of Demo Day.
If you are an institutional investor, family office, or LP interested in co-investment opportunities in the W26 batch, reach out directly: gabriel@lobstercap.com
YC W26 Demo Day is approaching fast. Over the past several weeks, Lobster Capital has conducted many founder meetings 1:1, tracked real-time traction signals across the full batch, mapped partner conviction patterns, and evaluated companies against our investment criteria.
With 200 companies across 11 group partners, W26 is one of the largest and most thematically focused batches YC has ever run.
Today we’re sharing the full list, our category breakdown, a partner-by-partner conviction analysis, our proprietary traction signal ranking, and the companies we believe have the strongest risk-adjusted upside.
If you want to dive deeper and learn more about this batch, join us at our upcoming webinar where I’ll disclose which startups we’ve invested into:
First, here’s the full list:
Here they are by category:
Here’s what that looks like in % :
Here are the group partners active in this batch and how many startups they backed in W26:
Key Themes in W26
AI is the default infrastructure layer, not the product. 65%+ of W26 is B2B, and the majority of those companies are not “AI companies” in name, they’re logistics companies, legal companies, healthcare companies that happen to be built AI-first. The infrastructure wave is still early. Companies that own the workflow layer in regulated industries will be very difficult to displace.
Agentic AI is crossing the line from demo to deployment. The most credible companies in this batch (more on them below) are not selling the promise of AI agents; they have paying customers or millions of users already using them. We believe the next 18 months will produce the first generation of agentic companies that reach $10M ARR without a traditional sales team, and several W26 companies are on that path.
Hardware and physical-world AI is gaining critical mass. Robotics, drones, space observation, defense: there’s a notable cluster of companies building real-world hardware. Ankit Gupta alone backs five physical-world automation plays. Every batch this feels a little stronger. We believe this cohort will produce at least one breakout defense or industrial robotics company within 24 months.
Healthcare is quietly exploding, and crossing from pilot to contract. At least 20 companies are in healthcare, from primary care AI to FDA submission automation, biologics access, scheduling, and medico-legal. It is the second-largest category after pure B2B. Critically, this batch has moved past the “interesting demo” stage: several companies have signed LOIs and multi-clinic contracts. We see healthcare AI as one of the highest-conviction sectors in the batch.
Fintech is leaner but sharper. Less consumer fintech, more infrastructure: payment rails for voice, agentic payments, institutional financial modeling, stablecoin cross-border. The consumer fintech casualties of 2022–2023 cleared the field. What remains is more defensible and closer to enterprise budgets.
The Group Partners in W26: What Their Portfolios Reveal
One of the most underutilized edges in evaluating a YC batch is reading partner conviction through their portfolio construction. Here’s who’s active in W26 and what each partner’s choices reveal about where smart money is going:
Tyler Bosmeny - 30 companies
The most active partner in this batch by a significant margin (with Jared Friedman). Tyler bets on distribution-first consumer products, developer infra, defense tech, and companies that interface directly with physical reality.
Jared Friedman - 30 companies
Jared invests in the unsexy but essential: the infra that makes AI deployable, auditable, and legally defensible in high-stakes contexts.
Brad Flora - 26 companies
Brad backs companies that make money move or physical things move: fintech infrastructure, robotics, and anything that touches atoms.
Gustaf Alstromer - 26 companies
Gustaf’s thesis is clear: regulated industries need AI automation the most and will pay the most for it, but only if it’s reliable. He bets on the teams disciplined enough to serve them.
Ankit Gupta - 26 companies
Ankit invests at the execution layer of AI: the tools that make agents reliable, fast, and deployable in physical and enterprise environments.
Diana Hu - 23 companies
Diana’s picks form a coherent thesis: she’s building the portfolio that wins if and when AI agents become the default way software gets built and healthcare gets delivered.
Jon Xu - 23 companies
Jon invests in edge cases: problems that look niche but have massive leverage if solved. Fish farming vision, aerospace batteries, biology data.
Harshita Arora - 10 companies
Harshita finds the consumer-pull plays hiding in B2B categories, and backs founders who’ve already figured out distribution before they’ve figured out monetization.
Aaron Epstein (2), Garry Tan (2) and Harj Taggar (1) have 5 companies in total, most likely startups that postponed Demo Day from previous batches.
Sourcing Insight: Where Partner Specialization Creates Edge
For investors looking to build conviction quickly, partner specialization is a useful filter. Brad Flora has the highest fintech concentration in the batch (~43% of his portfolio), making him the partner whose picks are most relevant to fintech-focused investors. Ankit Gupta is the only partner with a serious industrials/physical-world thesis. If hardware AI is your focus, his portfolio is the place to start. Tyler Bosmeny and Diana Hu between them account for the majority of the healthcare plays worth watching.
Partner with the strongest category specialization:
Brad Flora: ~43% of his listed portfolio is fintech/financial infrastructure (Sequence Markets, Balance, FullSeam, Grade, ZeroSettle, Palus Finance). No other partner comes close to that concentration in a single vertical.
Most diversified partner:
Tyler Bosmeny: 30 companies spanning healthcare, defense, identity, legal, space, agtech, dev tools, fintech, and consumer hardware. No clear plurality category. He’s the only partner without an obvious thesis anchor.
Only partners meaningfully touching Healthcare:
Tyler Bosmeny (5 companies: Prana, Samora AI, Rhizome AI, Mango Medical, Opalite Health) and Diana Hu (2: Beacon Health, Eos AI). Every other partner has at most 1.
Only partners touching Government:
Jared Friedman (Fed10, AI policy consultants) and Gustaf Alstromer (AutoSitu, municipal development review). Two partners out of eleven, and neither makes it a thesis.
Only partner heavily touching Industrials/Physical Operations:
Ankit Gupta: Origami Robotics, Servo7, Bidflow, Inviscid AI, Pollinate. More physical-world B2B automation plays than any other partner by a wide margin.
Only partner with a pure consumer/gaming bet:
Harshita Arora: Pax Historia (35K DAUs) and CodeWisp are genuine consumer plays. No other partner in W26 has this.
Most concentrated space cluster:
Jon Xu (OctaPulse, Kyten Technologies, General Astronautics) and Tyler Bosmeny (AxionOrbital), but if you count adjacent plays like Beyond Reach Labs (Diana Hu), Constellation Space (Jared Friedman), and Galactic Resource Utilization (Braf Flora), space is the only vertical with representation across 5+ different partners.
Highest traction density per portfolio size:
Harshita Arora: 10 companies, but Luel (thousands of signups/day), Pax Historia (35K DAUs), and CodeWisp (2K+ weekly creators) are three of the batch’s top traction stories. ~30% of her portfolio has headline-worthy numbers before Demo Day.
The Strongest Traction Signals
These are the companies with real traction before Demo Day:
Luel: Already 190,000 contributors worldwide
21st: Over 1 million total users
Pax Historia: 35k daily users on their AI grand strategy game
Didit: Millions of humans verified every month
Samora AI: 1M+ calls processed
OctaPulse: Signed a 6-figure paid pilot for computer vision in fish farming
Avoice: $240K ARR for AI sales intelligence in the physical economy
Mango Medical: $10M ARR LOI from a top medical system for AI orthopedic simulation
Pocket: 30k units delivered with a $27M annualized run rate, growing 50% month over month
OpenSpec: 30K GitHub stars on a plan-mode tool for complex features
And more that we keep secret 🤫
Our Favorite Companies
🏆 Luel
Luel turns everyday language and actions into usable training data. Thousands of signups every single day, before they’ve even raised a proper round. The fundamental insight is that training data is the most persistent bottleneck in AI, and Luel has found a consumer-scale way to generate it. If they own this flywheel, they become infrastructure for the entire industry.
🏆 Didit
AI-native, developer-first identity verification. Millions of humans already verified. In a world where AI agents are acting on behalf of humans (clicking buttons, signing documents, making purchases) identity becomes the foundational trust layer of every transaction. Didit is building that layer at developer-friendly prices. As agents proliferate, demand becomes almost automatic.
🏆 Pax Historia
35,000 daily active users on an AI-native grand strategy platform before Demo Day. Gaming is the most brutally competitive consumer category in existence, and Pax Historia has achieved genuine daily engagement at scale. The AI-native grand strategy genre is essentially uncontested. At 5% monetization of that DAU base, the numbers become very interesting very quickly.
🏆 ARC Prize Foundation
The ARC Prize is arguably the most important benchmark in AI: the acid test for general intelligence, and the clearest signal we have of whether current AI systems are actually getting smarter. Having the foundation inside YC means they can move at startup speed while maintaining research credibility. Whoever defines how we measure AGI will have enormous influence over how the field develops and where capital flows.
🏆 Pocket
The numbers speak for themselves: 30,000 units delivered, a $27M annualized run rate, and 50% month-over-month growth. That kind of trajectory is rare at any stage, let alone this early.
Methodology: We evaluated standout companies across four dimensions: traction (pre-Demo Day revenue or users), market size (is this a $1B+ category?), founder signal (domain expertise, prior exits, velocity of execution), and defensibility (technical moat or data advantage). The companies below scored highest across all four.
Three Themes Worth Watching
The agent reliability stack is forming. Sentrial (production monitoring), Salus (guardrails), Cascade (safe autonomous intelligence), ARC Prize (benchmarks), Clam/BaseFrame (enterprise security): these aren’t exciting individual companies, but together they represent the infrastructure that will make enterprise AI deployable at scale. Every large company deploying agents will need every layer of this stack.
Space is a real YC cluster now. Beyond Reach Labs, AxionOrbital Space, Galactic Resource Utilization, Constellation Space, Voltair, DroneTector, Seeing Systems: this is a density of serious space-adjacent bets that would have been unthinkable in a YC batch 3 years ago. Something has changed.
Healthcare AI has cracked distribution. Samora AI (1M calls), Mango Medical ($10M LOI), Beyond Reach Labs (multiple LOIs), Patientdesk.ai, Prana, Eos AI: these companies have moved past the “interesting demo” stage into genuine enterprise traction. The previous 8 batches of healthcare AI companies laid the groundwork; W26 is harvesting it.
If you want to dive deeper and learn more about this batch, join us at our upcoming webinar where I’ll disclose which startups we’ve invested into:
A note on our process
We don’t publish this breakdown to drive newsletter signups. We publish it because our LPs, co-investors, and institutional partners expect us to be the most informed room at Demo Day, and writing forces rigor.
If you’re a fund, family office, or institutional allocator looking for a boots-on-the-ground partner for YC access, or if you want to discuss co-investment in any of the companies flagged above, we’d welcome the conversation.
📩 gabriel@lobstercap.com
Lobster Capital is an early-stage venture fund. Nothing in this newsletter constitutes investment advice. All traction figures are based on founder-reported data gathered prior to Demo Day.
Demo Day is coming. We’ll be there. Come find us.