In 1935, Robert Watson-Watt was asked to build a “death ray.”

He said no.

Instead, he built something far less impressive — a crude system that could detect aircraft using reflected radio waves. The signal was weak. The screen flickered. It had blind spots.

It looked like a toy.

But Watson-Watt understood something deeper than physics:

The third best, delivered on time, beats the second best delivered too late.
The best may never arrive at all.

He didn’t wait for perfection.
He shipped what worked.

And that mindset — more than the hardware itself — is what mattered.

Today, we are at a similar point with AI.

AI agents hallucinate.
Robots are clumsy.
Automation breaks.

Many organizations are waiting.

Waiting for:

Perfect models

Fully autonomous agents

Zero-error systems

But history suggests something uncomfortable:

Perfection is usually late.

Meanwhile, “third best AI” already exists:

Agents that summarize meetings

Bots that extract tasks

Systems that route information

Automations that reduce manual friction

They are imperfect.
But they are operational.

And operational systems teach you faster than theoretical ones.

Robotics and AI are scaling quickly. When automation compounds, the advantage shifts to those who integrated early — not those who optimized perfectly.

Watson-Watt ignored fantasy and built infrastructure.

That’s the lesson.

Don’t chase the “death ray.”
Build radar.

Deploy the imperfect AI.
Integrate it into real workflows.
Improve while running.

Because in fast transitions, the third best — on time — quietly wins.