1999 Called. It Wants to Know If You're Making the Same Mistake Twice.
I want to tell you about three windows.
Not windows on a computer. Windows in time — the brief, chaotic, overlooked moments when a new technology is so new that the early movers get advantages that become permanent.
Window One: 1999
The internet had just become accessible to regular people. Websites were ugly, dial-up was slow, and most people thought it was a fad.
A small number of people — not necessarily the smartest or the richest — built websites, started forums, launched email newsletters. The technology was primitive and the future was unclear.
Those people became the first internet millionaires. Not because they were geniuses. Because they showed up when the room was still empty.
The people who waited two years to 'see how it plays out' — they're still playing catch-up.
Window Two: 2007
Facebook had existed for three years. YouTube was a year old. Twitter just launched. Instagram was still three years away.
The people who started creating content in 2007 — the early bloggers, the YouTubers who filmed potato-quality videos in their bedrooms — built audiences that became empires.
By the time the mainstream caught up and social media felt 'normal,' the early movers had distribution that was almost impossible to replicate. Algorithms favored early engagement. Audiences were loyal. The channels were established.
Most people watched from the sidelines while this happened.
Window Three: Right Now
OpenClaw was released recently. Claude, GPT, Gemini — these tools emerged in the last two years. The concept of AI agents that can take actions on your behalf, run 24/7, and integrate with your tools — this is 2025 technology.
The mainstream hasn't caught up yet. Which means the room is still relatively empty.
The people building AI agent skills right now are where the 1999 website builders were. They're where the 2007 YouTubers were.
They're early. And early, in technology, tends to be very, very good.
But This Time It's Different (In One Important Way)
In 1999, you needed to learn to code to build a website. The barrier was technical.
In 2007, you needed a camera, an idea, and the guts to show up on screen. The barrier was psychological.
In 2025, building an AI agent requires neither code nor camera. The barrier is almost purely informational — you just need to know how the tools work. And the tools themselves help you learn.
The irony is that the lowest barrier ever has also produced the most hesitation. People are waiting for AI agents to be 'more mature,''more proven,''less complicated' before they start.
That's exactly what the 2001 internet late adopters said. And the 2010 social media late adopters. And the 2014 smartphone app late adopters.
The Salim Ismail Prediction
At a recent India Today Conclave, Canadian entrepreneur Salim Ismail made a striking claim: the next trillion-dollar company will have 40 to 50 employees. Probably five.
Not five thousand. Five.
His reasoning: AI will handle the execution work that previously required large teams. Founders who know how to build and manage AI systems will be able to scale to revenue levels that previously required hundreds of people.
Whether his exact numbers are right doesn't matter. The directional arrow is clear.
What 'Too Late' Actually Looks Like
Social media isn't dead, but it's harder to build an audience now than in 2007. The channels are crowded. The algorithms are competitive. The first-mover advantage is gone.
That's what 'too late' looks like. Not impossible. Just much harder, much slower, and requiring much more capital.
AI agents aren't at that point yet. The practitioners are still few. The clients are many. The tools are just mature enough to work reliably. And the people who can build and deploy these systems are still rare enough to command real rates.
You can still be early. Early enough that being good at this makes you extraordinary, not just competitive.
The window is open. The question is never whether the window exists. The question is whether you walk through it.