During his 2025 keynote at CES in Las Vegas, Jensen Huang, in his usual approachable style, delivered some important updates regarding what I call the Paradigm of Jolting Technologies as it applies to AI architectures.
Historically, a given technology approach would deliver improvements only up to a point, after which further efforts would generate progressively diminishing returns. This is why we are accustomed to seeing S-curves when analyzing the history of any given technology: progressive development at the beginning, then a rapid adoption, followed by plateauing at the end.
However, in his Law of Accelerating Returns, Ray Kurzweil demonstrated how technological innovation follows a series of exponentials, whose sum results in a super-exponential curve. Moore’s Law, observed for over 50 years as a self-fulfilling prophecy due to the efforts of engineers competing and collaborating globally, is a special case of this phenomenon, where the rate of doubling remains constant. Today, however, we are witnessing a series of innovations that progressively shorten the doubling rate of whatever metric we seek to maximize. This pattern is becoming increasingly evident in the world around us.
The big surprise that fueled the current wave of AI enthusiasm 12 years ago was an unreasonable result: it is exceedingly rare to achieve better outcomes simply by scaling inputs by orders of magnitude. For instance, if building a bridge requires a thousand people working for a thousand days, hiring a million people won’t reduce the build time to one day. Yet, AI defied this logic. By scaling data, hardware power, and the sophistication of training algorithms, we achieved greater intelligence.
This is what Jensen highlights in his slide about the three scaling laws, ensuring not only that AI applications avoid hitting a plateau but that their power will increase at an ever-shortening doubling rate.
This approach now extends beyond pre-training to post-training and test-time reasoning. The result? Applications that, if viewed through yesterday’s lens, would seem like science fiction. Your task will be to experiment with these innovations and update your world model, recognizing that they will rapidly become part of your daily life.