Is AGI Here? A Sober Look at Today's AI Models
Everyone's debating if we've reached Artificial General Intelligence. We haven't. But today's models can do things that feel like magic—and fail in ways that are just as surprising.
June 22, 2026 · 1 min read · SuperThinking team
Let's cut to it: No, AGI is not here. We do not have a machine that can perform any intellectual task a human can. Not even close.
But for the first time, asking the question doesn't feel completely ridiculous. Models like GPT-4 and Claude 3 Opus display behaviors that are genuinely startling. They can synthesize, reason by analogy, and generate creative works that feel... thoughtful.
This isn't another article defining AGI. That's a boring philosophical debate. Instead, let's look at what these systems can actually do and where they fall flat on their faces. The gap between those two is where the truth is.
The 'Sparks' Are Real
The reason everyone got so excited, starting with a Microsoft Research paper on GPT-4, is that the model showed emergent abilities. These are skills it wasn't explicitly trained for, but that appeared as it scaled up. The classic example is theory of mind—the ability to infer another's mental state.
You can see this in practical tasks. Give it a messy codebase with a vague bug report, and it can often deduce the original developer's intent and find the logical flaw. It’s not just matching patterns; it’s building a mental model of the system.
Consider this prompt:
Explain the plot of Hamlet as if you were a VC hearing a pitch for a startup that's going to fail because of founder drama.The output isn't just a summary. It maps characters to startup roles (CEO, COO, board members), identifies the tragic flaw as a