X is buzzing about the Gemini 3.5 "Snow Bunny" leak. At first glance, some of the examples make you think, "If this is real, isn't it game over?" But taking a calm step back, I see it a little differently.
Honestly, the lab versions at all the companies building leading models are probably this wild by now. When companies are burning through hundreds of trillions of won a year to compete, it's hard to believe that just one has suddenly made a leap beyond everyone else. There is also a strong possibility that the leak itself is a deliberate signal. A way of telling both the market and competitors, "This is how far we've come." Competition among leading models is also a fierce marketing battleground, much like crypto, where you can't do anything without attention.
Everyone knows where Google has an advantage. Scale, data, infrastructure, TPUs, cash flow. Its disadvantages come from exactly the same place. The responsibility that comes with that scale, and the standards for AI safety and polish that this responsibility demands. The company that can run the fastest also has to run the most carefully.
OpenAI has benefited enormously from its first-mover branding. The equation "GPT = AI" still lives in people's minds, but the problem is that its shelf life is shrinking very quickly.
Anthropic has opened up a new paradigm in coding. Still, even as someone who actively uses Claude Code in its current form, I find it hard to answer the question, "How long can this lead really last?" A technical advantage and a business moat are different things. Its failure to channel the current developer energy into an open ecosystem, as seen in the Clawdbot ouster, is also a very big risk.
With xAI's Grok, you can feel Elon's will coming through. It makes its presence felt by charging straight into areas its competitors carefully avoid, like the bikini prompts flooding X. Its positioning is closer to a culture war than a pure technology race. The scariest part is the culture of religious gaslighting that makes developers willingly stay at work instead of going home. I've heard secondhand that the tents that filled the office ahead of the Grok4 launch are still there. A younger friend who recently interviewed at Grok in Silicon Valley said he ran away after executives smiled and pointed out a tent spot, telling him he could sleep and work there too.. (OpenAI, by contrast, apparently felt like a big company with plenty of money.)
Gemini still lacks a clear brand narrative. It is neither the smartest AI, nor the freest, nor the most developer-friendly. Being able to do everything well can also mean you haven't yet decided what you should do.
In the end, this race may be less about model performance than about which worldview you can get people to believe in. Technology that looks incredible now will gradually converge and become something close to a public good, and the real difference will come down to narrative. Open-source LLMs like Kimi will keep turning up with benchmarks that suddenly put them on par with the major models. Who will be smiling a year from now? What narrative will become the strongest, most lasting moat and lock in a community?
