Every few months, someone in Silicon Valley announces we are close to AGI. Or that we are already there. The goalposts move. The press release lands. The valuation holds.
I am getting a little tired of it.
Let me be clear upfront: I think AI is genuinely extraordinary. I use it every day. I build with it. I am not a sceptic in the dismissive sense. But I am a sceptic in the scientific sense - show me the evidence, and let's be honest about what it actually proves.
So let's talk about what is actually being claimed, what history tells us, and what I think AGI actually means.
What OpenAI Is Actually Saying
In early 2025, Sam Altman said OpenAI was "now confident we know how to build AGI as we have traditionally understood it." In August 2026, OpenAI's Chief Research Officer Mark Chen told TIME the company is "80% of the way" to AGI. They expect an internal AGI system before the end of 2026.
That is a bold claim. OpenAI's own definition of AGI is "highly autonomous systems that outperform humans at most economically valuable work." Even that definition is contested - but let's take it at face value and ask: are we 80% there?
I do not think so. And I am not alone.
We Have Been Here Before
Jensen Huang, Nvidia's CEO, said in early 2026 that he believed we had achieved AGI - then immediately hedged. Elon Musk and Dario Amodei at Anthropic both predicted AI would outsmart humans by 2026. Google DeepMind's Demis Hassabis has shifted his prediction from "a decade away" to 2030, now to 2029-2030. The window keeps shrinking, but we never quite arrive.
Here is the one that really gets me: in 2016, a study of AI experts put the median AGI timeline at 2055. Nearly forty years away. Now, nine years later, we are apparently 80% there?
The goalposts have not just moved. They have been picked up, carried to a different suburb, and quietly set down without anyone admitting it happened.
There is also this: the OpenAI/Microsoft partnership agreement defined AGI partly in financial terms - generating $100 billion in profit. That is not a technical milestone. That is a business one. And the agreement requires an independent expert panel to verify any AGI claim. Even they do not trust self-declaration. Worth sitting with that.
Yann LeCun, Meta's Chief Scientist and one of the most credible voices in the field, says AGI is not imminent and may require decades. He is not a pessimist. He is just being honest.
My Definition of AGI
Here is where I land on this. AGI is the point where AI can do everything a human can do, as well as a human can do it. Not just tasks. The whole job.
Let me give you a simple test. I call it the pizza test.
If I asked an AI system right now to order me a pizza - not just find a phone number, but actually figure out which place to order from based on what I feel like, navigate a website or app it has never seen before, handle the payment, deal with a problem if the order goes wrong, and follow through until the pizza arrives - could it do that reliably?
No. Not yet.
That is a task most teenagers can do without thinking. It involves navigation, novel problem-solving, adapting to unexpected situations, and following through on a multi-step process in the real world. AI is not reliably doing that in 2026.
AI is excellent at tasks. Better than humans at specific tasks in certain domains. But it takes thousands of tasks to make a job role. And doing a task is not the same as doing a job. And doing a job is not the same as being a human employee who can self-direct, adapt to novel situations, build relationships, and handle the thing nobody anticipated.
The Agent Reality Check
Agents are genuinely exciting. I am not dismissing them. The pace of improvement is real and it matters.
But multi-stage agentic tasks are still unreliable. They break in unexpected ways. They require careful setup and monitoring. Edge cases kill them. A human employee does not fail when the task has more than four steps.
Sam Altman predicted that AI agents would "join the workforce" in 2025 and "materially change the output of companies." Has that happened at scale? Not really. There are pockets of impressive deployment. But the economy has not transformed. Whole job functions have not been automated away. The mass agentic takeover has not arrived.
That is not a failure. AI is progressing fast. But it is a data point. We are not at AGI.
The Valuation Question
I am going to say the thing that some people think but few write.
OpenAI is burning extraordinary amounts of cash. Its valuation is enormous. Every "we are 80% there" statement is fuel for the next funding round. Every AGI announcement shores up a narrative that supports billions in investment.
That does not mean the technology is not real. It is real and it is impressive. But the framing of these announcements is financial as much as it is technical. When a company needs to justify its valuation to investors, the story it tells matters enormously. "We are making a useful chatbot" does not carry the same weight as "we are building the most transformative technology in human history."
Even Altman himself has acknowledged that "AGI has become a very sloppy term." They know this. And they keep using it anyway.
Will AGI Come?
Honestly? Nobody knows. Not even the people building the models.
AI is not like a programming language where you can trace the logic end to end. It is built more like wiring a brain - and no one fully understands how a brain works. The trajectory is upward. The progress is real. But the gap between "impressive at tasks" and "does everything a human does as well as a human" is enormous. Possibly unbounded.
The honest answer is: maybe. Eventually. But not now, and probably not soon in any meaningful sense.
What This Means for Your Business
If you are running a technology team or making workforce decisions based on AI headlines, here is the practical takeaway.
Plan for AI augmentation of humans. Not replacement of humans. That is the real near-term story.
The "AI will replace all jobs" narrative is as overblown as the AGI claim itself. Specific tasks? Yes, absolutely. Whole roles? Not yet. The roles that survive - and thrive - will be the ones that combine human judgment, adaptability, and relationship-building with AI as a powerful tool underneath.
For hiring managers and recruiters: the market for strong technical talent is not going away. If anything, the people who understand how to work alongside AI effectively are becoming more valuable, not less.
Stay curious. Stay sceptical. And do not let the marketing noise drive your workforce strategy.
I'd love to know where you sit on this. Do you think AGI is as close as OpenAI claims? Or does the pizza test give you pause? Reach out or drop a comment - genuinely curious whether people in tech are buying the narrative or quietly raising an eyebrow alongside me.
If you want to talk about what this means for your hiring strategy, get in touch with the Talent Aligned team.