Member-only story
Everyone Said AI Hit a Wall. Google Just Smashed Through It with Gemini 3
The benchmarks are insane, “Vibe Coding” is terrifyingly good, and the 45% ARC-AGI score proves the scaling laws aren’t broken, they just got expensive.
Well.. well.. well.
We have a new hotgirl in town, and her name is Gemini 3.
You all have heard the rumors. You’ve seen the tweets. “AI has hit a wall,” they said. “The scaling laws are broken,” they said. “We are in a bubble.”
And honestly? For a minute there, looking at the incremental updates from other labs, I almost believed them.
But then Google dropped Gemini 3, and… holy sh*t.
Google claims this is the current best and “beast” AI model in the world. And after digging through their technical reports, playing with the model, and reading the absolute chaos on Reddit, I have to say: The “wall” might just be a mirage.
But in this article, I won’t just bore you with a list of benchscores (okay, maybe a little, because they are insane). I want to talk about the fascinating stuff, the “vibe coding,” the agentic capabilities, and the reality check on whether this is actually AGI or not.
