Feb 18, 2026
Alex Armstrong
The past year has been a wild ride for anyone following predictions about artificial general intelligence. In early 2025, tech leaders were confidently declaring that AGI was just around the corner. By the end of the year, those same timelines had quietly stretched out by years. What happened? And more specifically, how does this apply to something uniquely human, such as art?
The Great Hype Cycle of 2025
When OpenAI released its reasoning models in late 2024 and early 2025, it generated a great deal of excitement. Sam Altman declared that his company now knew how to build AGI. Dario Amodei from Anthropic talked about creating "a country of geniuses in a data center" within just a few years. The narrative was clear: we were on the verge of something transformative, and it was happening fast.
But here's the thing about predictions from tech company leaders. They have skin in the game. Every bold proclamation about imminent breakthroughs may very well serve a purpose beyond simply sharing technical assessments. These announcements generate buzz, attract investment, and position companies as leaders in an intensely competitive race. When billions of dollars in funding and market positioning are at stake, it's worth taking even the most confident predictions with a healthy dose of skepticism.
As 2025 progressed, reality began to temper expectations. The Metaculus prediction market, which tracks forecasts from a broad community of analysts, tells a revealing story. At the start of 2025, forecasters predicted strong AGI would arrive in July 2031. By year's end, that forecast had shifted to November 2033. That's a two and a half year extension over the course of just twelve months.
Why the Timelines Stretched
Several technical realities became clearer as the year went on. First, while the new reasoning models were impressive at math and coding, they didn't magically generalize to other domains the way many had hoped. A model that excels at solving logic puzzles still can't reliably book you a flight or organize an event. The leap from narrow capabilities to broad, human-like intelligence turned out to be steeper than the initial excitement suggested.
Second, the methods used to create these improvements came with serious limitations. The technique of giving AI models more time to "think" about problems showed impressive results, but it's prohibitively expensive to scale up. We went from giving models essentially zero thinking time to one minute, which made a big difference. But going from one minute to ten minutes, or ten to one hundred, would require computer resources that simply don't exist at affordable prices yet.
The third challenge involves something called reinforcement learning, which requires enormous amounts of computation to produce relatively modest improvements. Think of it as an AI trying to learn by making millions of failed attempts before stumbling on the right answer. It works, but it's incredibly inefficient compared to how humans learn from just a few examples.
AGI In the Context of Creativity
Here's where things get particularly interesting for those in creative fields. While technical domains like mathematics and software engineering have clear right and wrong answers that can guide AI training, art occupies fundamentally different territory.
Art isn't about finding correct solutions. It's about making choices that resonate, communicate something human, and connect with other people's experiences and emotions. When you're writing a story, painting a portrait, composing music, or designing a space, there's no algorithm that can tell you whether you've succeeded. The validation comes from how your work makes people feel, what conversations it starts, how it shifts perspectives or touches hearts.
AI models can mimic artistic styles and generate technically competent work. They can create images that look professional, write text that follows grammatical rules, even compose melodies that sound pleasant. But they do this by pattern matching from their training data, not from lived experience, emotional truth, or genuine creative vision.
The same challenges that are making full automation of AI research and development take longer than predicted apply even more strongly to creative work. Just as improving AI systems requires human judgment about what constitutes progress, creating meaningful art requires human judgment about what makes something worth making. That judgment comes from being alive, from having experiences, from caring about things, from wanting to say something that hasn't been said before.
Renewed Perspective
Rather than seeing these extended timelines as disappointing, we should view them as clarifying. The breathless predictions of early 2025 created unnecessary anxiety for creative professionals. The reality is more nuanced and, in many ways, more encouraging. AI is becoming a more useful tool. It can help with research, generate initial ideas, handle tedious tasks, and accelerate certain parts of creative workflows. This is genuinely valuable. But there's a world of difference between being a useful tool and replacing human creativity altogether.
The evidence from 2025 actually supports a more optimistic view. Despite all the predictions and hype, AI hasn't fundamentally disrupted most creative workplaces. The gap between what looks impressive in a demo and what actually transforms real-world work has grown more apparent, not less. Models keep getting technically better, but their practical impact on creative industries remains modest.
This doesn't mean we can be complacent. Technology does advance, and creative professionals should absolutely explore how AI tools can enhance their work. But the idea that we're about to be replaced wholesale by machines? That narrative doesn't hold up when you look closely at how AI actually works and what its limitations are.
Looking Into the Crystal Ball
Even the most optimistic forecasters now acknowledge that meaningful AGI is likely decades away, not months. And when it comes to the kind of intelligence required for genuine creative work, the challenges are even more substantial. The inability of AI to learn continuously from experience, to develop taste through living, to make creative choices based on personal vision rather than pattern matching, these aren't minor technical hurdles. They're fundamental differences in how humans and machines process the world.
The story of 2025's timeline roller coaster is ultimately a reminder to look beyond the hype. Tech leaders have their reasons for bold predictions. But the reality is more complex, more human, and ultimately more interesting than any simplistic narrative about imminent replacement can capture.
The unique human perspective, lived experience, and ability to make choices that matter to other humans, remain irreplaceable. Not just for now, but for the foreseeable future. And quite possibly, forever.
