Year-End AI Sprint: How Silicon Valley and China Just Rewrote the Rules
Stop calling 2025 the “year of AI applications.” The truth is simpler: idealists were pinned to the floor by reality while the “six-dimensional warriors” harvested the battlefield. While we cheered for yet another pretty wrapper startup, the giants of Silicon Valley and China quietly completed the encirclement.
This year’s People’s Park Talks AI year-end awards skip the feel-good pork-splitting. We’re honoring the products and people who actually tore the industry’s fig leaf off—or slapped everyone awake.
Product of the Year: Nano Banana, the missing piece of the AGI puzzle
First, a moment of silence for Alex Xiu’s 2024 flame, Cursor. His verdict: Cursor choked on its own IDE half-measure, never reaching the Agent summit. This year he nominated Claude Code because Cursor still needs a human conductor, while Claude Code is the “perfect workhorse” that doesn’t wait for your cue. It doesn’t just autocomplete code; it seizes your laptop, runs tests, fixes bugs, even reads the docs. Alex’s words: “With Cursor I feel like a kung-fu master with junior disciples; with Claude Code I feel like scrap material about to be optimized away.” It’s so good it’s scary—and it’s not just a tool win, it’s the ticking clock on every developer’s career.
Rambo Liu shouted out Gemini 3, but our 2025 medal goes to the underdog that will outrun it: Nano Banana.
Why? First, it kept half of this year’s AI startups alive—every shiny new image or slideshow tool is just Nano Banana in a trench coat. Second, it’s real “paradigm shift.” It doesn’t process text; it models physics. Feed it a chalkboard sketch and it infers the missing equation; give it a GPS pin and it renders the exact sunset you’d see standing there. Behind the curtain is DeepMind chief Demis Hassabis flexing. While OpenAI is still gaming the next-token lottery, Hassabis used Nano Banana to argue that a “world model” is the only on-ramp to AGI. The banana doesn’t retrieve answers; it simulates gravity, light, causality. After years lost in the desert, Google didn’t just catch up—it lapped OpenAI at the origin. Nano Banana isn’t a model; it’s DeepMind’s entire R&D, compute and engineering stack forged into one beast. Bottom line: Claude Code keeps you employed; Nano Banana makes you question what employment will look like.
Team of the Year: Altman plays social butterfly, OpenAI builds skyscrapers on a cliff
Not a trophy for Sam personally. Sure, Google’s DeepMind integrated flawlessly; Alibaba’s Qwen team shipped so hard they even plugged into Gaode Maps—textbook commercialization. But next to OpenAI they look merely “excellent.” In one year OpenAI ballooned from hundreds to 4,000 heads, kept startup cadence, and dropped Sora, Atlas browser, ChatGPT-5, and Codex—the thing cornering programmers. That’s not just tech; that’s management witchcraft. While the boardroom coups and exec exits played like a soap opera, the engineers didn’t fall off the cliff—they threw up a skyscraper, faster and faster. Google and Alibaba are mining gold on solid ground; OpenAI is pouring concrete mid-air. If it doesn’t collapse, that organizational muscle is the real moat.
Boss of the Year: While Hassabis terrorizes Silicon Valley, a small-town founder nails the full stack
Hassabis takes the crown—no suspense. But we also nominate the “social butterfly” Altman and Yan Junjie, founder of MiniMax. Yan is one of the few Chinese founders who has sprinted the entire lap: research → engineering → consumer product. While investors were still debating “is multimodal a hoax?” he had video, audio, and code models all in global first-tier, on a budget one-tenth of US rivals. He’s the county-kid math-olymp type who solves a 12-variable business equation with a stubby pencil. In an age of TED-talk theater, that “think clear, ship fast” ruthlessness is pure pheromone.
Dark-Horse Cash-Out: Anthropic, the “get rich quietly” masterclass
While rivals scream for C-end retention, Anthropic slipped Claude Code into hundreds of thousands of enterprises. Revenue up ~10×, run-rate heading toward $9 billion. Claude Code alone may out-earn Anthropic’s entire 2024 top line. It proves the boring truth: enterprises don’t care if your model can write haiku; they care if it ships working code while the CEO sleeps. OpenAI won the volume; Anthropic won the certainty. In B2B you don’t need fireworks—just a digital ox that plows overnight.
2026 Checklist
Vibe awards done. The world flipped in six months. First half we gasped at DeepSeek and Liang Wenfeng; second half we saw Claude Code graduate from copilot to agent, and Nano Banana turn image models into world-model windows. Will 2026 fuse the two? A 3-D world model plus an agent with kernel privileges—AI flips from manual to automatic transmission. Meanwhile in China, Alibaba’s Qwen could close the loop with Gaode, Taobao, Fliggy, Xianyu, Alipay—model + ecosystem locked tight. And it might happen even faster.
One AI day, one human year.
Subscribe to JustSayAI Daily Brief
🌈 New here? Grab the AI Daily Brief · two emails a day · one-click listen
⭐ Old friend? Join the membership for brief + deep dives + text columns
Find us: