In partnership with


Today’s edition is written by:
Anna, Michael, & Thomas

🧋 Hello human and artificial friends,

China’s AI boyfriend business is officially next level.

Gen Z women are dating chatbots, customizing personalities, and in some cases even going on real-world dates with human cosplayers playing their AI partner.

Meet scalable romance.

STEAL THIS IDEA

Asia is where AI moves fastest. New products, new habits, new ways of using tech show up every week. Most disappear, but some quietly change how people shop, learn, travel, or create. We pick those ideas so you can study them and build your own version.

Today’s idea

💰 The $31M Bet on the "One-Man Empire"

Meet our new 10-person department

For the past year, we’ve heard about "Vibe Coding"—the idea that you can build apps just by describing them.

But a team in Shenzhen, DeepWisdom, just raised $31 million (backed by Ant Group and Baidu) to take this a step further.

They call it "Vibe Business."

What’s new about it

DeepWisdom builds an AI system called Atoms.
Instead of helping users write code, it helps them run a small business.

It deploys an entire virtual department (Product Manager, Engineer, Designer, and even an SEO Specialist) to build a finished, revenue-generating business in minutes.

Why it works

Beyond the "Toy" Phase: Atoms builds the "boring" stuff that usually kills solo projects: user authentication, databases, and Stripe payment integration. It’s a "Business-in-a-Box."

The "Race Mode" Strategy: To fix the problem of AI "hallucinating" bad code, Atoms runs several models (like Claude, GPT, and DeepSeek) against each other in the background.

  • It’s a "digital horse race" where only the best-functioning code survives and gets delivered to the user.

Built-in Growth The system includes an AI SEO Agent that automatically optimizes your site for Google from Day 1.

What you can build today

💡 The "Deep Niche" SaaS: Pick a highly specific, non-tech industry (e.g., HVAC technicians, boutique jewelry designers, or local boat rentals).

👉🏻 Use a tool like Atoms to build a "Full-Stack" tool that handles their specific booking, inventory, and payments.

💡 The Hot-Swap Business Use AI to spot "micro-trends" (like a sudden viral hobby on TikTok).

👉🏻 Use the "Vibe Business" model to launch a dedicated tool or community for that trend in 48 hours. If it fails, you lost $10 and a weekend. If it works, you own the category before the big players even hold a meeting about it.

🥡 Takeaway

The "One-Man Empire": It validates Sam Altman’s prediction of the first "one-person billion-dollar company."

By shifting the founder’s role from "builder" to "orchestrator," one person can now manage 10 different niche SaaS businesses simultaneously.

OUR PARTNER

Introducing the first AI-native CRM

Connect your email, and you’ll instantly get a CRM with enriched customer insights and a platform that grows with your business.

With AI at the core, Attio lets you:

  • Prospect and route leads with research agents

  • Get real-time insights during customer calls

  • Build powerful automations for your complex workflows

Join industry leaders like Granola, Taskrabbit, Flatfile and more.

RECENT AI

💰 Moonshot AI eyes a $4.8bn valuation: The Beijing-based startup behind the Kimi chatbot is close to wrapping up a new funding round that would lift its valuation by about $500m in just weeks. Investor appetite is rising as Chinese AI peers like MiniMax and Zhipu go public, while Moonshot’s latest Kimi K2 Thinking model is scoring near the top of global benchmarks, trailing only DeepSeek among Chinese models.

🇰🇷🇮🇹 Korea and Italy deepen tech ties: President Lee Jae Myung and PM Giorgia Meloni agreed to step up cooperation in AI, semiconductors, and aerospace, while also expanding collaboration on defense and critical-mineral supply chains. The two sides signed an MOU between Korea’s semiconductor industry association and Italy’s electronics association to boost business links, information sharing, and supply-chain coordination. They also plan a new 2026–30 action plan and to resume strategic dialogue soon.

🇨🇳 China trains AI without Nvidia: Zhipu unveiled GLM-Image, a multimodal image model it says is fully trained on domestic hardware using Huawei’s Ascend chips and MindSpore stack. It’s the first major Chinese model to publicly claim end-to-end training without foreign GPUs, underscoring Beijing’s push for AI self-reliance. The milestone comes as Washington inches toward allowing limited Nvidia H200 sales to China—and just days after Zhipu’s IPO, whose shares have surged on domestic-AI optimism.

TREND TO WATCH

🧭 Face-as-wallet in Korea: facial recognition checkout hits mainstream

Cashless Korea is moving from cards and phones to faces. With AI vision and 3D capture cutting errors and spotting fakes, “glance-to-pay” is showing up at scale.

Toss’s FacePay passed 1 million users months after launch, while Naver’s campus deployments normalize face payments for students. A country that’s already 90%+ cashless is pushing the next step. 

Face ≠ photo. Modern systems don’t just match an image, they verify “liveness” (blinks, micro-motions) and run fraud-detection in the background. Templates are encrypted and stored separately from public networks.

Merchants get speed, users get one-second, hands-free checkout.

What changed

  • Toss FacePay jumped from pilot to mass use. Launched in late 2025, it is expanding beyond early pilots. Korean coverage describes a path from tens of thousands of stores today toward hundreds of thousands to ~1M locations by 2026, driven by fintech distribution and lighter hardware demands. 

  • Naver brought “pay-by-face” to cafeterias. Naver Financial’s FaceSign went live at Kyung Hee University in 2024, using AI face recognition at campus checkouts—no phone or card needed. 

  • A perfect market for cashless. South Korea already tops global rankings for non-cash payments, creating a receptive base for face-as-wallet experiences. 

Why this could still stall

  • Privacy & oversight. Biometric data is sensitive under Korea’s PIPA; regulators have already flagged procedural gaps (e.g., Naver Financial’s early rollout needed prior adequacy review).

  • Deepfake cat-and-mouse. Liveness checks and 3D capture raise the bar, but attacks evolve. Providers will need constant model updates and stronger anomaly detection to keep spoofing at bay.

  • Merchant economics & interoperability. Scaling from campus and chains to long-tail merchants hinges on cheap hardware, easy onboarding, and acceptance across multiple processors—not just a single app.

BIG DATA

That’s how much of the world’s Physical-AI patent power now sits in China.

🤖 Patent gravity shifts east: Baidu, Huawei and Tencent take the top three spots, with China placing five companies in the global Top 10 for physical AI patents.

🏭 From code to concrete: Physical AI covers robots, autonomous machines and intelligent hardware. Beijing has elevated the field to national-strategy level, pushing applications from humanoid robots to factories, cars and logistics, even as US firms still edge China on average patent quality.

Watch: Physical AI locks in manufacturing, supply chains and long-term competitiveness. With patents piling up and policy backing in place, China is positioning itself where software meets steel.

TOOLS OF THE WEEK

🎬 Tongyi Wanxiang is Alibaba Cloud’s multimodal video system moving AI from clip generation to full narrative output.
Its edge: auto-generated story videos with multi-shot structure and professional voiceovers, combining high-fidelity visuals with a flexible voice engine including regional Chinese dialects.

🚀 DeepSeek-V4 is an upcoming large-scale code model designed as infrastructure for AI-driven development.
Its edge: handling up to 300,000 lines of code in a single pass, near–frontier performance in internal benchmarks, and extreme cost efficiency on domestic hardware.

🌐 SKT A.X K1 is a commercially open-source foundation model built by an SKT-led consortium.
Its edge: Apache 2.0 licensing, strong reasoning and coding performance, and a credible open alternative to closed frontier models with a clear multimodal roadmap.

STARTUP LAB

🤖 Singapore’s Botsync orchestrates robot fleets with Series A: Backed by SGInnovate, the AMR and vendor-agnostic “robot OS” startup is doubling down on SyncOS™ to coordinate multi-brand factory robots—after hitting 1M+ live production trips in 2025 and reporting ~230% revenue growth.

🇻🇳 Vietnam’s NamiTech scales VoiceDNA + AI CRM tools: The FPT spin-off secured $4M from Japan’s Toho Gas and existing backer TVS to expand its voice biometrics and conversation analytics used by clients in finance, insurance, retail, and telecom across Vietnam, Japan, and the US.

💼 Hupo AI trains sales reps with an “AI coach”: Meta's first Asia investment, the Singapore startup raised a $14M Series A led by DST Global to scale its platform that listens to sales conversations and coaches reps in banking/insurance, claiming faster onboarding and higher conversions.

BITS TO DO


See how fried rice can be prepared by a robot.
Find out what the best AI tools to make money in 2026 are.
Spend 48 hours in the world’s most advanced city.
Let AI plan your next Asia trip, tailored to your interests and budget.
Save unnecessary fees when shopping internationally and pay directly in multiple currencies.*

*Yes, it’s an affiliate link. No, you won’t pay more. But it helps support asiabits. Thanks :)

FUTURE COOKIE

On a cold, rainy day, Thomas would be glad to have someone else walk Einstein. 🐕

How would you rate today’s issue?

It helps us improve the content for you!

Login or Subscribe to participate

Keep reading

No posts found