What I noticed
Capabilities that seemed impressive only a short time earlier were already beginning to look primitive.
The “dial-up” analogy captures the danger of evaluating a fast-moving technology only through the limitations of its current version.
Why I’m watching it
AI strategy is unusually vulnerable to snapshot thinking.
A leadership team experiments with today’s model, identifies weaknesses and quietly assumes those weaknesses are durable. But capability, cost, speed, multimodality and integration can all change before the organization’s planning cycle finishes.
The important signal is not that every AI capability will improve indefinitely. It is that the rate of capability change itself belongs in the strategy.
What it might signal
Organizations need technology plans based on trajectories and scenarios rather than fixed capability assumptions.
That changes procurement, workforce planning, governance and experimentation. A task that is not practical to automate today may need to be reassessed much sooner than traditional technology cycles would suggest.
What to watch next
Watch cost per useful task, agent reliability, multimodal reasoning, inference at the edge, workflow integration and the speed with which experimental capability becomes a dependable product feature.
The question it raises
Which parts of our AI strategy assume today’s limitations will still exist two years from now?