2025 AI Live: What We Witnessed and Imagined This Year

2025 AI Live: What We Witnessed and Imagined This Year

October 21, 2025

This week on Crossroads, we're joined by Minghao Zhuang (host of The Art of Dragon Slaying) to revisit the defining AI and tech moments of 2025, along with our own memories and impressions from living through them.

This year, we were witnesses — watching technology iterate at breakneck speed, seeing products disrupt daily life in unprecedented ways. At the same time, we were swept up in a collective fever dream about the future, riding waves of innovation and possibility that left us exhilarated one moment and disoriented the next.

We start with Minghao's keyword for the year: "inflection point." From DeepSeek R1 in January to the recent Sora 2, we unpack the model wars, the "Year of the Agent" ignited by Manus, the open-source ecosystem and talent flows, and finally turn to the capital markets — how should we understand this collective fever dream about the future?

Tune in as we distill our observations and experiences from 2025 into a shared memory of the AI era.

————

This episode is specially presented by HarmonyOS: "Code Your Way Forward, See the Power of the Ecosystem"

——

🎥 This podcast was recorded at AI Hacker House. The video version will be published on Koji Yang Yuancheng's Xiaohongshu, Bilibili, Channels, and YouTube.

🟢 02:09 2025: An Inflection Point, Up or Down?

"We've unwittingly walked into the limits of so many things — technology, products, money." Why does "inflection point" capture 2025 so precisely? The upward inflection: a data center construction frenzy foreshadowing a 2026 explosion. The downward inflection: when does the bubble burst? Have we already touched the limits of technology and growth without realizing it?

🟢 03:24 The Model Wars: Divergence and Convergence in China-US Tech Paths

How did DeepSeek R1, built for millions, upend the hundred-billion-dollar infrastructure narrative? Sam Altman redefines the "Turing moment": why might AGI arrive not with a bang but a quiet step? Survival strategies for leading model labs: Anthropic goes deep on enterprise, xAI takes the wild path, why was Microsoft forced to build its own models? Is the pure chatbot battle already over? Behind ChatGPT's 800 million weekly active users — moat or growth ceiling? The Chinese consensus and weapon: why did "open source" become the only means to counter American AI dominance? What does DeepSeek V3.2's release signify? And why might we not see V4 or R2 this year?

🟢 23:42 Another Path to AGI: From Sora 2 to World Models

"If visual models are also at the main table, they might achieve AGI through an entirely different route." Why is the multimodal battlefield more competitive than language models? Meitu, marketing video agents... the use cases and commercial paths are startlingly clear. OpenAI's product philosophy: why did Sora 2 reach millions of households when technically comparable rivals didn't? Does the world really not need an "AI Douyin"? Perhaps OpenAI never intended to build one. Google is back! Could its world model Genie be another path to AGI, even "the womb of worlds"?

🟢 39:04 The Year of the Agent — What Now?

Why will agents linger at the L3 stage for so long? Because for the first time, AI capability extends from "language" to "action." Manus's historical significance: it showed users for the first time what an agent should look like — defining the mental model is worth its weight in gold. How do agent startups survive? As general-purpose agents become gospel for the tech giants, vertical domains — law, finance, marketing — are paradoxically thriving. The attention economy's squeeze: when mainstream lanes get overcrowded, why can even niche tracks like "AI dynamic manga" — with limited upside — still raise money?

🟢 49:11 On-Device Agents for Smartphone Makers

Why couldn't Siri become the true "phone assistant," while today's on-device agents might? The HarmonyOS HMAF framework's insight: when the OS stops doing everything itself and instead delegates intent to apps' native agents, what new opportunities emerge for developers?

🟢 53:00 Open Source, HarmonyOS: Ecosystems with Chinese Characteristics

From "frontier models must be closed" to "open-source models will top the charts by 2026" — why did Sam Altman reverse his position? How does open source commercialize? DeepSeek's API still sells, at a fraction of OpenAI's cost. How does open source become a "weapon"? In competing for Europe, Southeast Asia, the Middle East — the intermediate zones — open source carries inherent trust advantages. How intense is local deployment demand? One laptop, oversized in RAM and VRAM for local model hosting, sold out instantly. What new "ecological niches" can developers seize? HarmonyOS developers are pulling in 70,000 yuan monthly with lightweight apps.

🟢 01:08:36 Secondary Market Euphoria, What About Primary Markets?

"Back then we thought AI was a technology, an industry. Today AI is the market itself." Sam Altman's endgame thinking: what happens when one company tries to package the next five years of growth expectations? On the flip side, Chinese VC's familiar mobile internet growth playbook — "can't be replicated today." Where do investors find conviction? When pure AI software stories stop working, everyone pivots to hardware founders with DJI, Roborock, and Dreame pedigrees. The ultimate question against the dot-com bubble: fiber optic cable could pave the future, but GPUs obsolete in three years? A curious signal: beyond NVIDIA, this year's best performers in the S&P 500 are Seagate and Western Digital — hard drive companies. Bubble alarm: when AI giants turn to debt financing, is the specter of "subprime crisis" drawing nearer?


Subscribe to the Crossroads podcast 🚦 We track the industry shifts and entrepreneurial opportunities brought by the new wave of AI.

🚦 Crossroads is Steve Jobs's metaphor for Apple — standing at the intersection of technology and liberal arts, where great products are born. AI is transforming every industry. We seek out, interview, and bring together a new generation of AI founders and proactive builders of the AI era, exploring and embracing new changes and possibilities alongside them.

👦🏻 Host Koji: Co-founded Jiepang / The Fair / Tangdao, launched AI Hacker House — a community space for the next generation of AI entrepreneurs. I believe technology, especially AI, represents the greatest value creation opportunity of our generation. Feel free to reach out to chat, exchange ideas, and connect on what's next. Koji on Jike | Koji's website

👧🏻 Host Ronghui: Formerly at a dollar-denominated VC, five years as a Silicon Valley correspondent, tracking technology development and business stories. Reach out anytime to chat and exchange perspectives. Ronghui on Jike

🎄 This podcast is supported by The Fair's Sound Forest Podcast Initiative.