Wrapping Up 2025, Looking Ahead to 2026
That’s a wrap on 2025. Hello 2026. 🎉
If there’s one word that captures the past year in AI, it’s acceleration. Not just in model quality or benchmarks, but in competition, strategic divergence, and the sheer speed at which long-held assumptions collapsed.
Looking back, it’s hard not to feel energized.
A Few Defining Moments from 2025
Vibe coding went mainstream.
AI-driven software development crossed a psychological threshold this year. Andrej Karpathy gave it a name, but the shift was already underway: the mental model of “writing code line by line” is being replaced by something more fluid, more collaborative, and more AI-native.
There’s no going back.
Open models closed the gap.
DeepSeek sent shockwaves through the industry and made something undeniable: open models are no longer “good enough”—they’re genuinely competitive. For many use cases, they’re now the better choice.
For enterprises, this means more leverage, more flexibility, and more control than they’ve had in years.
We launched Oumi.
This one is personal.
We launched our open-source stack and saw early traction that many household-name tech companies—and even multi-billion-dollar startups—would envy at launch. I’m deeply grateful to the community that showed up early and believed in what we’re building.
Google came roaring back.
Between Nano Banana, Veo 3, and especially Gemini 3, Google flipped the narrative. Google dethroned OpenAI as the frontier AI leader and forced OpenAI to go into “code red.”
Why 2026 Feels Even Bigger
As exciting as 2025 was, 2026 feels more consequential—and, in a few cases, more concerning.
Building on recent Bloomberg coverage and The Puck interview, four trends are dominating my thinking as we head into the new year. This isn’t exhaustive, but these are the ones I keep coming back to.
1. Consumer AI - The App Layer Becomes the Real Battleground
As closed models—GPT, Claude, Gemini—converge in quality, raw model performance stops being the differentiator. The fight shifts decisively to the application layer: distribution, UX, brand, and habit formation.
Meta’s acquisition of Manus AI is an early signal. Expect more consolidation, more experimentation, and more brutal competition for consumer mindshare.
2. Enterprise AI Grows Up
Enterprises are maturing fast.
We’re seeing a clear move away from undifferentiated, slow, expensive, and often unreliable large off-the-shelf models toward smaller, specialized, open models that are faster, cheaper, and higher quality for specific workloads. As enterprises progress from early prototypes and MVPs to scale outs, they require higher performance and more control over their key underlying AI technology.
This transition is happening now faster and faster — it’s exactly why we built Oumi.
3. Market Rationalization (A Controversial Take)
It’s becoming increasingly clear that the “closed model + capital is the moat” strategy increasingly favors large established incumbents as opposed to startups.
With roughly $400B in annual revenue as opposed to a $10B loss, and ownership of the full vertical—chips, models, platforms, and applications—Google is structurally advantaged in a way that OpenAI and Anthropic simply aren’t.
In my view, the window for OpenAI and Anthropic to win purely on this strategy has largely closed. Investor outcomes now hinge on either strong, differentiated applications—or an IPO before the bubble pops.
4. US vs. China—and the Open-Source Gap
This one worries me the most.
The US is quietly losing the most important battleground in AI: open source. US enterprises are increasingly adopting models like Alibaba’s Qwen, while some of the most influential voices in US AI advocate for closed strategies that —conveniently— align with their financial incentives.
If open-source leadership continues to slip, the long-term consequences could be profound.
Mark Zuckerberg—we need Meta to make a real comeback here.
I’ll end with a question instead of a conclusion: what did I miss?
I’d love to hear what trends you’re watching—and where you think I’m wrong.
Onward to another wild year for AI and Oumi. 🚀