YC W26 2C Company Complete Breakdown: Only 14 Out of 199 Are Consumer-Focused, So Why Is the Traffic Champion Here?

Written by: Lang Hanwei Will

This is the first article in our YC W26 series analysis. We spent a week tagging all 199 companies in YC W26—primary industry classification, secondary niche focus, whether AI-first, whether Agent-type, B2B or B2C, and website traffic data from December 2025 to February 2026. Next, we will analyze each track one by one, diving deep into each focus area. Starting with the most scarce and counterintuitive track: Consumer.

An Unexpected Data Point

YC W26 has a total of 199 companies. After tagging each, we found: 91% are AI-first, 27% are working on Agent, and 93% are B2B.

Consumer companies number only 14, accounting for 7%.

This figure isn’t high in any YC batch, but in W26—a batch almost entirely AI and B2B—the scarcity of Consumer companies stands out. Lobster Capital’s analysis also points out that Harshita Arora is the only YC Partner in this batch to heavily focus on Consumer.

Even more interestingly, among these 14 Consumer companies, the top two in traffic are here: Pax Historia (3.6 million monthly visits) and Pocket (460,000 monthly visits). One develops AI strategy games, the other AI note-taking hardware. Their combined monthly visits surpass the total of most of the remaining 197 companies in the batch.

Except for these two, the other 12 Consumer companies have monthly visits below 15,000.

This polarization itself is a story worth telling.

Summary: 5 Key Takeaways

1. The truly successful Consumer companies are not those simply adding AI to existing products, but those creating entirely new categories with AI. Pax Historia created an “AI alternate history strategy game,” and Pocket developed “portable AI note hardware.” Category innovation has a higher ceiling than feature innovation.

2. Hardware is an underestimated direction within Consumer AI. Out of 14, three are hardware companies (Pocket, Fort, Button Computer), founded by teams from open-source hardware communities, Tesla, and Apple. Pocket has already achieved $27 million annualized revenue, proving AI hardware is more than just a PPT concept.

3. Gaming is the largest subcategory in YC Consumer (3 companies), but their paths differ greatly: Pax Historia uses AI-native gameplay, CodeWisp offers AI game creation tools, and CatchBack Cards develops traditional digital card games. Only the first truly leverages AI’s unique capabilities.

4. Consumer AI in this batch is a “high-reward, low-probability” track. B2B companies sell with clear ICP and willingness to pay, while Consumer companies face the triple challenge of distribution, retention, and monetization. YC Partner choices reflect this—only Harshita Arora is systematically betting on Consumer.

5. For Chinese teams, some product logic directions in these Consumer companies are directly applicable, but distribution channels need to be completely rebuilt. Chinese teams’ product capabilities are on par with US teams, but their outbound distribution is a weakness. Pax Historia’s 3.6 million monthly visits were driven by word-of-mouth in English-speaking communities and creator economies—these flywheels require different strategies in China and for international markets.

Pax Historia: The Batch’s Traffic Champion, Created by Two University Students

Official website:

An AI-driven sandbox strategy game set in alternate history. Choose a country and a time period, then rewrite history with AI.

Key Data: 3.6 million monthly visits (batch top), 35,000 daily active users, launched token monetization within 8 weeks from zero

Business Model: Token-based—free players get 0.2 tokens daily, which are spent interacting with AI; 1 token ≈ $1.00. Paid options include token packages (Basic ~$6) and Patron subscriptions (first month free, 300 free AI requests per day, supports custom API keys).

Team Highlights: Two CS students from Virginia Tech, started in April 2024. Ryan’s previous “Qing Dynasty simulation” gained 13,000 YouTube subscribers.

Competitors/Risks: Paradox Interactive (HOI 4 / CK 3) has dominated grand strategy games for decades, but their user base doesn’t overlap—Paradox appeals to hardcore players willing to spend 200 hours learning systems, while Pax targets history enthusiasts who prefer simpler gameplay.

Other Highlights: YC CEO Garry Tan called it “one of the highest retention LLM games ever”; Columbia students built an autonomous agent to play Pax on HN; strategic partnership with Senhwa Biosciences.

Traditional strategy games rely on detailed stats and tech trees, but Pax Historia replaces these with AI-generated real-time narrative text. No fixed victory conditions, no pre-set story branches. Uses Infraon AI to manage over 30 model providers for intelligent routing, reducing AI costs by 60%.

Creator economy is the core flywheel: players create presets (custom maps, characters, backgrounds), and after over 150 rounds, earn 10% of token commissions. The clever part: AI inference costs are directly passed to users (more tokens for better models), content is crowdsourced (no official level design needed), similar to Roblox UGC but with lower creation barriers—just write a historical background, no coding required.

Monetization differs from Paradox—Paradox sells DLC (over 20 DLCs totaling over $300 for HOI 4), Pax sells tokens, with theoretically unlimited AI-generated content, no need for DLC development. A lighter model, but more dependent on AI inference cost reductions.

Traffic trend: December 2.77 million → January 2.67 million (slight dip) → February 3.61 million (+35%). The rebound likely relates to YC Launch exposure; sustained growth depends on retention and monetization.

Pocket: 30,000 units sold in 5 months, $27M annualized revenue

Official website:

A small AI device designed specifically to record face-to-face conversations—meetings, doctor visits, chats—then transcribe, summarize, and create to-do lists.

Key Data: 30,000 units shipped in 5 months, $27M ARR, 50% MoM growth (confirmed by YC). 472 reviews with an average rating of 4.89.

Business Model: One-time hardware purchase ranging from $79 to $327, no subscription. $79 is an entry-level model (magnetically attaches to phone back), $327 is a standalone device.

Team Highlights: Akshay Narisetti previously developed open-source AI note devices, turned down Google three times. Gabriel Dymowski founded DoxyChain.

Competitors/Risks: Plaud ($169+) offers more features—30-hour battery life, 60 days standby, 98% transcription accuracy, 112 languages, many free AI features. Pocket’s battery lasts only 4 days, AI features behind a $20/month paywall. Otter is purely software for online meetings, not the same market. Limitless (“Memory Necklace”) was acquired by Meta in December 2025 and taken off the market.

Other Highlights: Pocket features a contact microphone—magnetically attached to the phone’s back to record calls—considered more natural than competitors’ physical attachment methods.

Pocket’s founder’s product philosophy: “Amazon Kindle succeeded by doing less—no notifications, no Instagram, only books. Devices that do one thing beat devices that do many.”

Pocket’s advantage over Plaud isn’t in features but in its low entry price of $79, greatly reducing trial costs. Shipping 30,000 units proves supply chain viability, something PPT companies can’t match.

Traffic trend: December 2.3K → January 3.9K → February 4.6K, steadily growing.

Fort: Three Tesla engineers develop a wearable strength training device

Official website:

A screenless wearable that automatically recognizes over 50 strength training movements, tracking reps, sets, speed, and fatigue.

Business Model: Pre-order (retail $349), shipping in Q3 2026. One-time hardware purchase.

Team Highlights: All three founders are ex-Tesla engineers—Zac worked on Cybertruck batteries (prototype to production), Miranda on Cybercab chassis and powertrain (previously at SpaceX on Starlink), Paul on Semi and Robotaxi systems architecture. Zac and Miranda got engaged after meeting at the gym.

Competitors/Risks: Whoop, Oura, Garmin focus on cardio, so strength training devices may be less relevant. Past failures include Beast Sensor (Italy, poor accuracy, discontinued), Push Band (Canada, $1,000/year subscription, high cost), Trackbar (closed ecosystem). Common issues: unreliable accuracy, high user effort (manual movement selection, inputting weights, start/stop).

Other Highlights: Covered by media outlets like New Atlas, NotebookCheck, Fitt Insider. Funded by YC, Afore, Weekend Fund, Theory Forge, and angel investors from OpenAI/Tesla.

Fort’s differentiation: AI automatic recognition (no manual input), all-day wearability (not just during workouts), and Tesla’s mass production experience. But if accuracy isn’t sufficient, the entire product logic fails. User reviews after Q3 shipping will be critical.

Traffic: December 0.26K → January 0.24K → February 1.2K (+390% MoM), mainly driven by media coverage.

Button Computer: “Conversational Button” developed by former Apple Vision Pro engineer

Official website:

A small button on your shirt that, when pressed, enables AI conversation.

Team Highlights: Ryan Burgoyne, ex-Apple senior software engineer involved in Vision Pro, with background in software and mechanical engineering. Chris Nolet, 6 years at Apple, involved in Vision Pro, previously founded Skyglass (mobile virtual production).

Competitors/Risks: Similar in appearance to Humane AI Pin but different positioning—Humane aims to replace phones, Button only provides a “low-friction voice AI interface.” A button with no screen or camera, with very narrow functionality.

Other Highlights: Strong funding background—seed funds from Sequoia, Index Ventures, plus angel investments from OpenAI, Hugging Face, ElevenLabs, Unity.

Product is still early-stage, no public user or revenue data.

Other two gaming companies: CodeWisp and CatchBack Cards

Besides Pax Historia, two other companies in gaming are taking very different paths.

CodeWisp is an AI game creation platform—describe the game you want in text, and the platform automatically generates code and assets, supporting 2D/3D/multiplayer browser games. Founder Elvin Fu has been making games since age 10, with 22 million views on YouTube teaching game development, and a self-built audience of millions. Over 2,000 weekly active creators. But the challenge is fundamental: Unity and Roblox are integrating AI into their engines, so independent “AI game creation” tools face increasing pressure as engine giants embed AI capabilities.

Official website:

CatchBack Cards makes digital collectible card games, more like digital collectibles/NFT variants. Traffic in February was 9,000.

Official website:

Looking at these three together, the conclusion is clear: AI’s value isn’t just in “helping create content” (CodeWisp), but in “making the experience itself different” (Pax Historia). When AI becomes the core engine of the product experience rather than an auxiliary tool, it opens the door to entirely new categories.

The Remaining 5: Quick Scan

While much of the earlier analysis focused on leading companies, five smaller companies among the 14 are worth a brief mention—they are smaller but some focus areas are quite interesting.

Martini/C 47 (15,000 monthly visits, MoM +157%): AI-assisted professional video collaboration tool. Real-time multi-user editing and generative media. Interesting direction, but limited info.

Librar Labs (15,000 monthly visits, MoM +944%): AI library management system. Coach Jared Friedman. Rapid growth but small base. A highly vertical niche.

Vela (6,000 monthly visits, MoM +175%): AI scheduling assistant. The space is crowded (Reclaim, Motion, Clockwise), so differentiation is needed.

Resonate (9,000 monthly visits, new site): AI-native messaging platform. Too little info to judge.

Doomersion, Remix, Return Signals, Zero/ZeroSettle: No public traction data yet. Doomersion’s concept of “learning languages via short videos” is interesting—an AI-native alternative to Duolingo. Remix automates social media content creation. Zero/ZeroSettle offers SDKs to bypass app store commissions; not strictly AI companies.

Looking at all 14 together: The survival rules for Consumer AI

Data reveals:

(Buttons and some others lack traffic data)

Key observations:

First, traffic polarization is severe. Pax Historia’s 3.6 million exceeds the combined traffic of the other 13 companies by over 20 times. Pocket ranks second with 460,000, and the third has only 15,000—there’s a 30x gap. The distribution in Consumer is not normal but follows a power law—either explosive or nearly nonexistent.

Second, successful companies share a common trait: category innovation. Pax Historia isn’t just “strategy game with AI,” it’s a completely new AI-native game category. Pocket isn’t just “AI voice recorder,” it’s a “designed-for-face-to-face conversation AI device.” This distinction is crucial—adding AI to existing products works in B2B (where ROI matters), but in Consumer, the product must be something that “didn’t exist before.”

Third, hardware companies outperform expectations. The three hardware firms (Pocket, Fort, Button Computer) have the strongest founding teams—backgrounds from open hardware, Tesla, Apple. Pocket has established supply chain and revenue models. In a batch that’s 92% software, hardware companies are the most traction-rich in Consumer.

Beyond YC: The “Cemetery” and Survivors of Consumer AI Hardware

YC W26’s Consumer companies aren’t isolated. Over the past two years, several major events have shaped the landscape—understanding these helps assess these 14 companies’ prospects.

Humane AI Pin is the most famous failure. Priced at $699, it aimed to replace phones with a laser projection device on the shirt—“AI device replacing smartphones.” Launched in April 2024, reviews were almost universally negative—slow, hot, limited features. The company raised over $200 million but abandoned consumer hardware by 2025. Key lesson: trying to make users abandon phones is unrealistic; AI hardware should complement, not compete with, smartphones.

Rabbit R 1 is another cautionary tale. An $199 orange box promising to help operate apps via “large action models” (LAM). Experience fell far short of expectations; user interest waned after the first week. The issue isn’t hardware design (Jesse Lyu’s industrial design is excellent), but core functionality—why buy a device for tasks your phone already does?

Limitless Pendant took a different approach—a “memory necklace” worn around the neck, recording continuously, automatically organizing searchable memories. The concept was good, but after being acquired by Meta in December 2025, the product was taken off the market. This highlights another risk in Consumer AI hardware: the market is too early. Even if the product is right, the company may not reach scale before being acquired.

Looking back at YC W26’s three hardware companies (Pocket, Fort, Button Computer), they’ve learned from these failures:

Pocket learned “don’t replace phones”—it’s an extension, not a substitute. Priced at $327, much lower than Humane’s $699, reducing user trial costs.

Fort learned “focus on one thing”—not tracking your life or pushing notifications, but dedicated solely to strength training.

Button Computer learned “minimum friction”—a button that, when pressed, activates AI conversation, no learning curve.

Can these avoid the fate of predecessors? Data speaks: Pocket has already proven product-market fit in shipment volume and revenue. Fort and Button Computer still need validation after shipping.

Implications for Chinese Teams

These 14 Consumer companies offer several lessons for Chinese outbound teams.

First, product capability isn’t the bottleneck—distribution is. Chinese teams can produce Pax Historia or Pocket-level products—China’s AI application innovation isn’t slower than the US, and Shenzhen’s hardware supply chain is world-leading. But Pax Historia’s 3.6 million visits relied on word-of-mouth in English communities, Hacker News discussions, YouTube and Reddit communities, and YC Launch exposure. Recreating this distribution flywheel in China or overseas markets requires entirely different skills: English community management, Product Hunt/Hacker News cold start, and overseas KOL/creator networks. These are orthogonal to product development but may be more critical for Consumer success.

Second, AI gaming is a promising outbound direction. Pax Historia proves AI-native games can achieve 35K DAU and 3.6 million monthly visits in browsers—no Steam or App Store needed, short distribution chain, low global entry barrier. China has many experienced game teams; if they find similar “AI-native gameplay”—not just adding AI assistants to traditional games but making AI the core experience—browser + global + creator economy is feasible. Pax Historia’s token model, passing AI inference costs directly to users and crowdsourcing content via creator commissions, is worth studying and replicating.

Third, AI hardware’s overseas success hinges on supply chain advantages. Pocket’s 30,000 units shipped demonstrate real demand. China’s hardware ecosystem is unmatched globally—Shenzhen can help produce similar products at lower costs and faster. Fort’s early media coverage shows high market interest. But beware: Plaud (a Chinese company) already dominates the AI voice recorder space; new entrants need differentiated hardware categories rather than direct competition.

Fourth, “focus on one thing” product philosophy applies overseas. Pocket’s founder’s Kindle analogy is instructive—“a device that does one thing beats one that does many.” Chinese outbound products often suffer from feature bloat and trying to satisfy everyone. In Consumer AI, extreme focus is an advantage. Failures like Humane AI Pin and Rabbit R show that more features and fuzzy positioning lead to quick demise.

Fifth, pay attention to the market gap left after Meta’s acquisition of Limitless. The “all-day AI memory device” concept was validated but product was delisted. No strong competitors exist in the short term. If Chinese teams can differentiate on privacy and edge processing (a potential strength—edge inference + data stays on device), this is a promising direction.

Key Takeaways

1. Consumer AI in YC W26 is a “high-reward, low-probability” track. Only 2 out of 14 companies (Pax Historia and Pocket) gained significant traction, but their traction exceeds most B2B companies. If you pursue Consumer AI, clarify whether you’re creating a new category or just adding AI to existing ones.

2. AI hardware is not a false proposition. Pocket’s $27M ARR and Fort’s media buzz confirm real demand. But supply chain risks are real—slower, heavier, and more expensive than software.

3. Gaming may be the earliest monetization path in Consumer AI. Pax Historia’s token and creator economy are already operational. AI’s ability to generate “infinite” game content fundamentally changes monetization—no more DLC, just sell tokens.

4. For Chinese teams aiming at Consumer AI overseas, distribution is the first hurdle. Product capabilities are comparable, but “how to make the first users know you” is 10x harder abroad. Pax Historia and Pocket relied on English community word-of-mouth, requiring language and community skills.

5. Tomorrow is YC W26 Demo Day (March 24). The funding performance of these 14 Consumer companies will be a key signal—whether investors are willing to back Consumer companies in a batch that’s 92% B2B. We will continue to monitor.