What I Learnt From 199 Pitches at the YC W26 Demo Day
I attended YC’s Winter 2026 Demo Day. 199 companies. Here’s everything I took away: the data, the patterns, and what it means if you’re a future founder.
Key Takeaways & Lessons for Founders
On the Market / Problem Statement
1. AI isn’t a category, it’s the substrate. 60% of the batch is AI-native. Another 26% is AI-enabled. Only 14% has no AI. The question isn’t “are you using AI?” It’s “what does your AI do that a foundation model can’t do out of the box?”
2. Replace, don’t assist. The defining theme is “AI employees,” not copilots, not assistants. The pitch is always “we replace [expensive human role] end-to-end,” priced as a fraction of that person’s salary. Copilots assist. Agents act. The industry moved on.
3. Find the “Claude Code” for your domain. Every profession has structured outputs AI can now generate: contracts, CAD files, financial models, surgical plans, specifications. Find a profession where practitioners cost $100-500+/hr, tools are 10-30 years old, and there’s a clear verification step. Wide-open domains: tax prep, civil engineering, management consulting, clinical trials, patent drafting, music production.
4. Consider the service model. ~20% of the batch is building AI-native service firms (law, recruiting, accounting, insurance) that charge for outcomes with software margins. They show the fastest revenue in the batch. The playbook: start as a service → get revenue and data → ship the automation → graduate to a platform.
5. B2B dominates. AI agents replacing B2B knowledge workers. 87% B2B. Only 14 consumer-facing companies (~7%). The current AI capability unlock maps perfectly to business workflows. This is a good trade, but the legendary companies from this batch will likely be the outliers: the uranium explorer, the moon hotel, the robot cowboys, the parasite drug company.
6. Build a data flywheel. Every customer interaction should make your product better. LegalOS trained on 12K visa petitions → 100% approval rate. Perfectly improves with every hire. Without a data flywheel, you’re a wrapper.
7. Don’t build a general-purpose AI wrapper. “AI for everything” loses to “AI that replaces one specific $80K/yr role.” Go deep into an unsexy industry. The best opportunities are in industries you’d never pitch at a cocktail party.
8. The absence of consumer is the opportunity signal. Zero education companies. Zero consumer social. Zero mental health/wellness. Zero govtech. The least-funded categories historically produce the largest outlier returns. The founder who cracks AI-native entertainment, social, or education has the category to themselves.
9. Hardware is back. 18% of the batch has hardware components (robotics, drones, wearables, space tech). That’s a notable jump from recent batches. The SpaceX/Tesla alumni building physical products are among the most differentiated companies in the batch.
On Distribution
10. Distribution is a prerequisite, not an afterthought. 60% of the top 15 companies by growth acquired customers through founder network or YC network. If your first 20 customers require “figuring out distribution,” you picked the wrong market.
11. Your previous employer is your first market. The dominant GTM motion (~35% of B2B): founders who spent years in an industry, left, and sold back into their network. Their Rolodex IS the distribution channel.
12. The PE roll-up channel is massively underexploited. Ressl AI and Robby independently discovered that PE-backed roll-ups are desperate for margin improvement tools. One PE deal = 50-200 locations.
13. Choose markets where you already have a distribution network. The companies struggling with GTM are almost always the ones that built a product first and then asked “how do we sell this?” The winners asked “who do I already have access to, and what do they desperately need?”
On Teams
14. Founder-market fit is the single strongest predictor of speed to revenue. The founders who literally did the job they’re now automating close deals in days. Everyone else takes months. Proximitty ($700K ARR in <3 weeks): CEO was a McKinsey bank risk advisor. Corvera ($33K MRR in 4 weeks): CEO ran a CPG brand.
15. Your cofounder relationship is your moat. 46% of the batch is 2-person teams. The strongest teams worked together for years: former colleagues, classmates, siblings, repeat cofounders. If you haven’t shipped something with your cofounder, you haven’t validated the most important part of your startup.
16. Domain expertise beats pedigree. The most compelling founders lived the problem: the dentist building surgical AI, the aircraft maintenance head building mechanic tools, the lobbyists building policy AI. “Ex-FAANG” is table stakes, not a differentiator.
On Pitching
17. The crazy closer matters. When 199 companies pitch in one day, you need to be the one they talk about at drinks. “The first AI Oscar will be made on Martini.” “You can book a hotel on the Moon for 2032.” Make your vision specific, falsifiable, and quotable.
On What to Avoid
18. Avoid undifferentiated agent infrastructure. 8-10 companies building agent monitoring/testing/compression. Foundation model providers will build these natively. If “[Existing DevOps tool] but for AI agents” describes you, that’s the danger zone.
19. Avoid AI-native services without a data moat. Fastest revenue but lowest defensibility. Core tech replicable in weeks. Traditional firms will adopt AI in 12-18 months. Without proprietary data or embedded distribution, the moat is thin.
20. Avoid commodity workflow wrappers. AI for a single well-defined task where GPT-5 might do the same thing natively in 6 months.
The Scene
199 pitches. Something about freshly baked startups coming out of the YC oven that just smells so good. Exciting, high-energy, never a dull moment.
Some moments that stuck:
A startup pitching the first hotel on the Moon, with White House invitations and $500M in LOIs
Robot cowboys herding cattle with autonomous drones
An AI presentation company generating its pitch deck live during the demo
One company casually zooming into Tehran, Iran while demo-ing satellite imagery (the room went quiet)
Martini’s founder closing with “The first Oscar for an AI-made movie will be won by Martini!”, the kind of line that makes investors either roll their eyes or reach for their checkbook
The hardware demo area was buzzing: robots, drones, microscopes with lifesciences proteins, radars mounted on cars. Real, physical things you could touch. This wasn’t just a batch of SaaS dashboards.
After 199 pitches, you stop hearing individual companies and start seeing patterns. Here’s what I found.
The Macro Numbers
Total companies: 199
Business Model:
B2B: 174 (87%)
B2C: 14 (7%)
B2B2C: 11 (6%)
Product Type:
Pure software: 163 (82%)
Hardware + software: 24 (12%)
Pure hardware: 12 (6%)
AI Classification:
AI-native (AI IS the product): 120 (60%)
AI-enabled (existing workflow + AI): 52 (26%)
Non-AI: 27 (14%)
Traction:
Estimated median ARR: ~$50-100K
Estimated median growth: ~30-50% MoM
Companies >$1M ARR: ~5%
Pre-revenue: ~50%
Top industries: B2B software (59%), Industrials (15%), Healthcare (10%), Fintech (8%), Consumer (4%).
Only 14 companies are consumer-facing, and YC officially classifies just 7 as “Consumer.” The rest are consumer products wearing enterprise labels, tucked under B2B, Healthcare, or Fintech.
The Ten Themes
1. AI Agents Replacing Entire Job Functions
The defining theme. Not copilots, full replacement.
Beacon Health replaces admin staff doing prior authorizations
Perfectly replaces recruiters end-to-end
Lance replaces hotel front desk across 50+ Marriott/Hilton/Hyatt properties
Mendral (Docker’s co-founders) replaces the DevOps engineer
Canary replaces QA
The “copilot” framing dropped from ~4% of pitches in early 2025 to 1% in W26.
2. “Claude Code for X”
Claude Code and Cursor proved agentic AI works for code. W26 founders are applying the same paradigm to every profession with structured outputs:
REV1 for mechanical engineers (3D→2D drawings)
Avoice for architects (specifications, documentation)
Synthetic Sciences for scientific research
Maywood for investment bankers
Alt-X for RE underwriting (works directly in Excel)
Cardboard for video editing
Mango Medical generates surgical plans in minutes instead of days
3. AI-Native Professional Services (”Service Business, Software Economics”)
Not building tools for incumbents, building AI firms that compete with them:
Four AI law firms (Arcline, General Legal, Vector Legal, LegalOS)
AI recruiting agency (Perfectly)
AI accounting (Balance)
AI insurance brokerage (Panta)
AI policy consulting (Fed10, founded by three ex-lobbyists)
Panta says it explicitly: “A service business with software economics.” Charge for outcomes, operate with software margins, because AI does 80% and humans handle 20%. Arcline has 50+ startup clients. LegalOS has 100% visa approval rate.
The bear case: humans-in-the-loop cap margins at 60-80%. Liability is real. The moat question: if the core tech is “LLM + domain prompts + human review,” what stops replication? The emerging answer: start as service → ship automation → graduate to platform. The service is the wedge; the software is the moat.
4. Infrastructure for the Agentic Era
Every tech stack layer is being rebuilt for agents:
Agentic Fabriq = “Okta for Agents”
Sponge (three ex-Stripe crypto leads) = financial infrastructure for agents
Moda/Sentrial = Datadog for agent reliability
Salus = runtime guardrails
21st (1.4M developers) = React components for AI-first UIs
Zatanna turns pre-LLM SaaS into databases agents can query
The risk: foundation model providers build these natively. The ~30% competitive overlap in this layer confirms it’s crowded.
5. Vertical AI for “Unsexy” Industries
The biggest ROI is in industries tech has ignored:
Zymbly automates aircraft maintenance paperwork (45 min docs per 5 min repair)
GrazeMate builds robot cowboys, autonomous drones herding cattle. When they pitched, you couldn’t help but grin. It sounds absurd until you learn the founder grew up on a 6,000-head cattle station.
OctaPulse does CV for fish farming
Squid tackles power grid planning ($760B annual inefficiency, still on spreadsheets)
These founders went deep. Scout Out’s founder is fourth-generation construction. LegalOS’s co-founders grew up in their family’s immigration law firm (10,000+ hours each since age 12). Zymbly’s co-founder was Head of Aircraft Maintenance at Virgin Atlantic. The best opportunities are in the industries you’d never pitch at a cocktail party.
6. Physical AI / Robotics Renaissance
18% of the batch has hardware components:
Remy AI and Servo7 build warehouse robots learning from human demonstrations (80% of warehouses have zero automation)
Origami Robotics builds robot hands
RoboDock went viral deploying their MVP in 60 days, landed a $100K Waymo contract
Fort (three ex-Tesla engineers) tracks strength training, something Whoop/Oura still can’t do
Pocket shipped 30K+ units, $27M annualized revenue
The hardware demo area was the most energizing part of the day.
7. Defense & National Security
Milliray (three Oxford/St Andrews PhDs) builds drone detection radar for NATO ($470K in batch sales)
Seeing Systems builds AI strike drones for UK Royal Marine Commandos
DAIVIN! builds tankless dive gear for U.S. Special Ops
Defense budgets are large, contracts long, credibility transfers to commercial.
8. Data as the Moat
When everyone has the same foundation models, proprietary data is the primary defensibility:
Shofo: world’s largest indexed video library
Human Archive: dropped out of Stanford/Berkeley, moved to Asia, collecting data from thousands of homes for humanoid robots
LegalOS: 12K successful visa petitions → 100% approval rate
The pattern: every customer interaction makes the product better. Without a data flywheel, you’re a wrapper.
9. Hard Tech & Space
The most audacious pitches. GRU Space is building the first hotel on the Moon by 2032. When they pitched, the room recalibrated: half thought they were crazy, half thought they might do it. $500M in LOIs, White House invitations, 1B+ views. Beyond Reach Labs builds football-field solar arrays for orbit (500x power demand increase by 2030). Terranox uses AI to discover uranium deposits (single discovery = $200-700M).
Ditto Biosciences might be the most creative thesis: parasites evolved proteins that control the human immune system over millions of years. Ditto uses AI to identify them and engineer autoimmune therapies. Evolution already solved the problem, they’re just reading its answers.
10. AI-Native Research & Science
Talking Computers deploys fleets of AI scientists ($1M+ ARR)
Aemon (twin brothers, published at ICLR/EMNLP before age 20) set a world record on an NP-hard math problem with <$10 compute, beating Google DeepMind
Ndea, co-founded by Zapier’s Mike Knoop and Keras creator François Chollet, is explicitly building AGI that can innovate
The Founders: Patterns From 429 People
Demographics:
~60% immigrant/international
86% male, 14% female
Top schools: Berkeley (~45), Stanford (~35), MIT (~20), Waterloo (~15)
55% studied CS; 45% didn’t
Backgrounds:
~30% ex-FAANG
~25% had a previous startup
~12% ex-finance/trading (Citadel, Jane Street, Jump)
~12 founders from SpaceX alone, overwhelmingly building hardware and aerospace
Teams:
46% are 2-person teams, 15% solo
Most common archetype: two technical cofounders with different specializations (~35%), not the classic “hacker + hustler”
19% of companies have at least one PhD founder
How they met: ~35% university classmates, ~25% former colleagues, ~15% repeat cofounders, ~10% family/siblings
Domain experts who became founders are the most compelling stories: Adrian Kilian (dentist → surgical AI at Mango Medical), Robbie Bourke (25 years in aviation → Zymbly), Pamir Ehsas (outside counsel to OpenAI → Arcline), Conor Jones (years inside National Grid → Squid).
Some observations:
Deep domain expertise + a technical co-founder who can build = the strongest companies in the batch
The most successful-looking teams either built and sold a company together before, or worked side-by-side at the same company on the same problem they’re now solving
31% of companies have at least one PhD or researcher founder, primarily concentrated in healthcare/biotech, hard tech, and AI infrastructure
How They Found Their Markets
B2B (88% of batch)
“I Lived This Pain” (~40%): The strongest pattern. End Close’s founder spent 6 years at Modern Treasury processing $1T+ in payments. Squid’s founders spent years inside National Grid. They didn’t need customer discovery, they were the customer.
“I Built the Platform This Replaces” (~20%): Docker’s co-founders building Mendral. TikTok’s ML scientists building Perfectly. They know the architecture intimately and see where AI creates a step change.
“The 50-Conversation Sprint” (~15%): Systematic discovery. Ritivel had 50+ pharma conversations before writing code. Ressl AI started as consulting, found trades had the most glue work.
“Infrastructure Prophecy” (~15%): Thesis-driven. “If agents exist, they need auth” → Agentic Fabriq. Risk: building for a future 2-3 years away.
“Research → Commercialization” (~10%): CellType (Yale professor + DeepMind). Valgo’s co-founder literally wrote the textbook on safety-critical systems.
B2C (7% of batch)
“I Am the User” (~50%): Fort founders are lifters frustrated by wearables. Doomersion’s founder doomscrolls and studies languages, combined them.
“Format Shift” (~25%): Existing behavior + new medium. Pax Historia: love of strategy games + AI alternate history.
“Hardware Wedge” (~25%): Physical product creates data loop software can’t replicate.
The meta-lesson: Not a single successful W26 company was born from a hackathon or a “what if we used AI for...” brainstorm. Every one traces to deep personal experience or obsessive customer discovery.
How They Found Distribution
The data is clear: founder network is the #1 mechanism for the fastest-growing B2B companies. 60% of the top 15 by growth rate acquired first customers through founder network or YC network.
B2B patterns:
“Sell to former employer’s peers” (~35%): Fed10’s three ex-lobbyists, their Rolodex IS distribution
“YC as launch pad” (~25%): Cardinal runs outbound for 40+ YC companies, Palus Finance signed 33 in weeks
“Open source” (~10%): 21st has 1.4M developers, only works for infrastructure
“PE roll-up channel” (~8%): One deal = 50-200 locations
“Systematic outbound” (~15%): Finite buyer lists with quantifiable pain
“Wedge product” (~7%): Land narrow, expand everywhere
B2C: The product IS the distribution. Doomersion got 15K downloads in 2 weeks, zero paid marketing. Pax Historia built tens of thousands of DAUs organically. Hardware founders bet that physical presence generates word-of-mouth.
The biggest takeaway: The companies struggling with GTM are almost always the ones that built a product first and then asked “how do we sell this?” The winners asked “who do I already have access to, and what do they desperately need?”, then built that.
Anatomy of a Great Pitch
Seven components that separate pitches that stuck from ones that blurred:
1. Hook. Three archetypes work:
The Shocking Stat: “It takes 500,000 days to bring a drug to market. We want to make it 5” (Rhizome AI)
The Reframe: “Every file you’ve ever uploaded uses a protocol from 1974” (Byteport)
“I Am the Problem”: “I spent 6 years building reconciliation at Modern Treasury processing a trillion dollars” (End Close)
2. Problem (specific, not generic). “Technicians spend half their time on paperwork” (Zymbly) beats “We automate back-office workflows.”
3. Team (one-line credibility bombs). “Andrea wrote Docker’s first lines of code” (Mendral). “Our team invented the MPIC standard securing every HTTPS connection on the internet” (Crosslayer Labs).
4. Market (inevitable, not just large). “Satellite power demand: 500x increase by 2030” (Beyond Reach Labs). The strongest market pitches explain why now and why this is inevitable, not just how big the TAM is.
5. Traction (velocity > absolute number). “$0 to $33K MRR in 4 weeks” (Corvera) beats “$100K ARR” with no timeframe.
6. Unique insight. “Parasites evolved proteins that control the human immune system. We read their answers” (Ditto Bio). “Insurers can’t price autonomous systems because historical claims data doesn’t exist” (Valgo).
7. Crazy closer. “The first AI Oscar will be made on Martini.” “Book a hotel on the Moon for 2032” (GRU Space). The pitches that blurred: generic “AI for [industry],” team credentials without connection to the problem, and (critically) no crazy closer.
Competitive Overlap: YC’s Multiple Bets
~30% of companies have a direct competitor in-batch. Only ~5% face truly high overlap.
High overlap: LLM context compression (Token Company vs. Compresr), med-legal docs (Wayco vs. Docura Health), robotics data (Human Archive vs. Asimov)
Medium: Startup legal (Arcline vs. General Legal vs. Vector Legal), AI SRE (IncidentFox vs. Sonarly), agent monitoring (Sentrial vs. Moda), prior auth (Ruma Care vs. ClaimGlide vs. Beacon Health)
What it tells you: YC bets on markets, not companies. Three startup law firms = the market is real and large enough for multiple winners. Two identical-looking companies on Demo Day will look completely different by Series A. The most differentiated companies have zero overlap: Terranox, Zymbly, GrazeMate, Ditto Bio. In each case, the founder’s domain expertise IS the moat.
What’s Conspicuously Absent
Zero education companies
Zero govtech
Zero consumer social
Zero mental health/wellness
Near-zero marketplaces
Near-zero pure crypto (blockchain used as plumbing, never as product thesis)
Consumer at a historic low (14 companies total, just 7 officially classified)
Industrials went from 3.6% of W24 to 14.1% of W26, a 4x jump. The “atoms vs. bits” shift is real within YC.
The contrarian read: W26’s composition is a snapshot of what’s fundable right now, not what will be valuable in 10 years. The legendary companies missing from this batch are the consumer and social founders who will arrive in 2-3 batches, once AI capabilities catch up to their ambition.
What’s Likely to Fail
Undifferentiated agent infrastructure. 8-10 companies in agent monitoring/testing/compression. Foundation model providers will build these natively. Enterprise buyers default to existing vendors.
AI-native services without a data moat. Fastest revenue, lowest defensibility. Core tech replicable in weeks. Traditional firms adopt AI in 12-18 months.
Solo technical founder in relationship-sold markets. Construction, insurance, freight: if nobody can walk into a job site and speak the language, it stalls.
“AI for [Industry]” without domain depth. The tell: description leads with “We use advanced LLM agents to...” rather than the customer’s specific pain.
Pre-revenue deep tech with long horizons. Not wrong conceptually, but the failure mode is running out of runway.
Commodity workflow wrappers. Single-task AI where GPT-5 might do the same thing natively in 6 months.
The Fastest Companies Share Five Traits
Sold the outcome, not the tool
Founder had customer relationships before the product existed
Charged from Day 1: no free tier, no pilot purgatory
Customer was desperate, not curious (Proximitty: banks with $2B+ delinquent loans; Ruma Care: clinics denied $150K reimbursements)
MVP was embarrassingly simple: they describe outcomes, not architectures
The gap between “launched and learning” and “building and hoping” is where most of the mortality in this batch will occur.
Exciting times ahead! Never been a better time to build.
Written on March 25, 2026, days after YC W26 Demo Day.
The US is miles ahead in vision, risk and experiments then any other country.
Curious on how you analysed the data of 199 pitches? [What was the prompt and raw data :p]