WWDC 26 补上了 AI，但离真正的 AI 助手还差什么？| S10E15

WWDC 26 Adds AI, But What's Missing for a True AI Assistant? | S10E15 - What's Next｜Tech Early Know｜Xiaoyuzhou - Listen to Podcasts on Xiaoyuzhou

On June 9, WWDC 2026 was held in Cupertino, California, as scheduled. This was the last keynote hosted by Cook during his tenure. Apple officially renamed its voice assistant Siri to Siri AI, introduced a new logo, made it an independent app, and announced a deep collaboration with Google Gemini in the model aspect. After two years of anticipation, Apple Intelligence finally delivered a response. Why, despite owning 1 billion iPhones, the world's deepest software ecosystem, and the most complete hardware matrix, has Apple fallen far behind other tech giants in the AI field?

In this episode, we invited Dong Hongguang, founder of Guangfan Technology, to join us in discussing AI assistants and smart hardware. Beyond the app silos, privacy restrictions, and edge-cloud collaboration, is there a completely different path that bypasses the smartphone form factor and builds an AI-native software infrastructure from scratch? What underlying infrastructure is needed for an AI-native personal assistant? And in this race between hardware and software, do Chinese startups have a chance to outperform the big companies?

This Episode's Guest: Ding Jiao Diane, Co-founder of "Shengdong Pohuo", Host of "Tech Early Know", and Founder of Guangfan Technology.

Timeline: 00:14 Starting with WWDC, what did Apple submit this time?

• Mainly catching up, under expectations: Renaming Siri AI, standalone app, integration with Gemini,

• Compared to innovations in domestic apps, the gap is obvious, but Apple's ecosystem strategy naturally requires a cycle.

04:38 The Gold Mine is Hard to Mine: Apple's Four Weaknesses

• Long decision chain + weak model capabilities, relying on OpenAI and Gemini only to meet the minimum standard

• Data is locked within super apps, OS layer cannot access it, so the so-called gold mine is actually hard to mine

• The ceiling of the smartphone form factor: If something can be done with two taps, there's no need to go through AI speech

08:46 Lessons from AI Hardware: Why Smart Hardware Failed?

• What's missing is not hardware, but an AI-native software infrastructure

• Traditional OS architecture needs to be rebuilt from scratch

• Today we have large models, but no infrastructure built around large models

11:36 Google vs. Apple: Advantages of a Full-Chain Player

• From the Gemini model to TPU chips, cloud infrastructure, internet services, and Pixel hardware, there are almost no obvious weaknesses

• Apple's gap lies in the model layer, which determines its upper limit

• "Coolness" has quietly shifted

20:31 Personal AI Assistant: A Non-Consensus Not Everyone Supports

• There are two layers of difficulty in implementation: technically, whether it can truly understand users, and ecologically, who is willing to open permissions

• Autonomous driving achieved "everyone has a driver," and a personal AI assistant is "everyone has a butler"

23:20 What Guangfan is Doing: Taking the Phone Sensor and Reinstalling It into the Earbuds

• Earbuds + Watch: An accepted wearable form + all-day wear + the golden location near the ear and eye

• Software side: Native AI integration rather than going through the phone app interface

36:09 Real Difficulty of Hardware Innovation: How Much is in One Earbuds Case

• Camera: No small camera customized for earbuds, mirror surface is arc-shaped, glass is replaced with resin

• Earbuds case integrates eSIM + GPS + fingerprint, can connect independently, power consumption management is extremely complex

• Current stage is "AI Assistant 1.0": Still far from the final state, waiting for the ecosystem

42:26 Who Will Emerge First in AI Hardware Between China and the US

• Hardware: Supply chain is fully in China, China has an absolute advantage

• Software/Model: On par, China is already leading in some areas

• US big companies dominate, investors don't like AI hardware; China has higher startup density and can innovate more

50:01 Dumb vs. Playable: User Segmentation Strategy for General AI Hardware

• Short-term: Target geeks first, support connecting local LLM, use voice to let AI vibe coding write skills

• Long-term goal: Ready to use out of the box + AI actively learns user habits to form personalization

Behind the Scenes Production: Supervised by Yaxian, Post-production by Dika, Operations by George, Design by Fanta.

"Knock Knock World" New technologies from Chinese companies, how to block the world's number one killer of infectious diseases? sourl.co In "Knock Knock World," you can hear global news and become a "global observer," submit topics, and participate in topic meetings. The 2026 program is still being updated.

Business Cooperation The commercial team of Shengdong Pohuo, click the link to go directly to the Shengdong Business Lounge (sourl.cn), or send an email to business@shengfm.cn to contact us.

Join Shengdong Pohuo Shengdong Pohuo is currently hiring full-time business operations managers, early coffee content interns, and community interns. If you are also interested in content production in the podcast industry, please click the recruitment entrance.

About Shengdong Pohuo "Collide with the world through sound," Shengdong Pohuo is committed to providing people with endless thinking nutrients. We also have these podcasts: Shengdong Early Coffee, Shengdong Ji Xi, Eat, Drink, Play, and Be Remarkable, Anti-Flow Club, Effervescent VC, Business Why Sauce, Jump into the Rabbit Hole, and Not Just Money. Please interact with us on Ji Ke, Weibo, and other social media platforms. Search for "Shengdong Pohuo" to find us. We look forward to your emails, the email address is ting@sheng.fm. Please scan the code to add Sheng Xiaoyin and stay in touch with us outside the program.

Special Guest: Dong Hongguang.