We’re Not Racing to AGI – and That’s OK

GPT-5 is finally here! And yes - it’s extraordinary. Anyone looking at today’s AI from the vantage point of 2022 would be stunned. Yet, for all the hype, we’re not witnessing the rapid take-off towards Artificial General Intelligence (AGI) that many predicted.

Instead of one model exponentially self-improving and leaving the rest behind, we’re seeing something far more interesting. The top AI labs (OpenAI, xAI, Google, Anthropic, DeepSeek, Mistral...) are neck-and-neck, playing weekly catch-up with marginal improvements in a crazy cycle of competition. Progress is real, models keeps getting better, they are more usable, cost less, but we are hitting a ceiling, the inflection point of the S-curve.

This fits a pattern we’ve seen in many technologies before AI. The S-curve happens when rapid gains give way to slower progress as current approaches hit their limits. Scaling up today’s large language models won’t magically get us to human-level intelligence - Yann LeCun, Gary Marcus and others have been saying this for years. The next research breakthroughs will come from new architectures, smarter designs. And models that don't just learn from patterns but from principles.

The true AI revolution will come from what we do with these models. And it is already unfolding: millions of people using these tools to think, create, and solve problems in ways we couldn’t imagine a few years ago. We will tackle bigger problems. Work will change. Companies will rewire entire processes and all industries will transform themselves upside down. That’s where the real transformation lies folks, no Terminator, no AGI in the real world.