After AGI, the only thing left is you
I remember that my first impression of AGI came about a year and a half ago from an essay called Situational Awareness. I found it fascinating, especially given that it was presented at a time most people were dismissing AGI as a distant concept. And when we take a moment to realise how fast things have moved since then, the timeline is actually reasonable. Nvidia's CEO tells Lex Fridman he thinks we've achieved AGI. Anthropic's CEO says powerful AI is one to three years out. So the preceding “will AI stall out” conversation is settled. What comes next is the interesting part. Not the capability itself. The opportunity.
And this is already playing out. The repetitive work that used to take days is being done in minutes. Faster, often better. But that isn't the most interesting part yet. The point is what it leaves to us. The work that can't be described in a prompt. Not because it's too complex, but because the thinking behind it doesn't follow a process that can be simplified to a prompt. And if you can't prompt it, no system can touch it.
That's not a limitation of AI today. It's a limitation of what AI is for. Some decisions can't be achieved by logic and processing alone. The answer comes from a human sitting with it, weighing what no dataset captures, and making a call. That's real intelligence. Not computation, but human perception leading to a decision. Understanding what's really at stake.
I'm not speaking theoretically. I see this clearly in my own work.
I'm developing a mental health app. Having to figure out what it should actually be for the people who use it, what direction to take for the next feature, where the line sits between something that helps and something that just looks like it does. Every one of those decisions has to come from understanding the person on the other side of the screen. What they actually need, what would make the product feel right to them, what would make them come back because it genuinely made a difference. That's the part that can't be engineered. And it's also what matters most.
AI can hugely support you in what you are building. But the hard calls, deciding what would make something truly matter to the person on the other side, those are yours. And they always will be.
What changes isn't the work. It's how much of your time goes to the part that counts. AI doesn't replace the thinking. It removes everything between you and the thinking that actually matters. And the question now isn't whether AI is smart enough. It already is. It's what you'll do with all the time it's about to give you. Because thinking well, creating with intent, and knowing when to change course has never mattered more than it's about to.
Panagiotis Tzavaras
panagiotis.tzavaras@tzavaras.ai