We’re on the brink of three major shifts in Al: huge context windows, agent-based learning, and text-to-action capabilities.
Context windows are expanding sos rapidly that we can hand an Al something as large as a library of books and have it make sense of the whole set, even though it “forgets” in the middle, much like human short- term memory.
Meanwhile, agents are emerging as LLMs that can learn by reading, testing, and reconfiguring their own understanding.
Finally, text-to-action is where you’ll be able to say, “Build me the next (XYZ),” and the Al will code, iterate, and deploy it, instantly at scale.
Put all three together, and we’re talking about an impact that dwarfs what social media did to the world.
Imagine everyone having a personal programmer: not just for silly projects, but for real innovation or disruptive tools.
Right now, these systems are limited by cost and complexity, but those barriers won’t last.
The moment they come down, you’ll see a wave of new creations roll out in mere hours or days, things no single developer or team could’ve spun up alone.
That’s the power of big context, self- learning agents, and instant action, all merging in the very near future.
