Alchemy vs Chemistry

Before 1878, the only way to know if two chemicals would explode was to mix them and stand back.

Then Josiah Willard Gibbs did the math. He didn't make reactions safer — he made them calculable. Measure a couple of properties, and the result was known before the beaker left the bench. Chemistry stopped guessing.

But we still run business like alchemists. Especially with AI. We can sorta feel which experiments might be dangerous, but we can't size them. So we do the two alchemist things: mix and pray, or wrap projects in so much bubble wrap nothing moves.

Chemistry is the third option. The blast radius is calculable, from two numbers we can estimate before we hit go:

A) If this goes sideways, how long until we know? 3 minutes or 3 months?
B) Once we know, what’s the cost to undo? A dollar or a deposition?

Max out one number and the result is measured in megatons.
A change we'd catch in a minute or undo with a click is a whimper.

I know which one I’d risk my money on.

On August 1, 2012, a botched deploy at Knight Capital woke a piece of dead code, and its system fired 4M orders into live markets in 45 minutes — real money at machine-speed with no readout. By the time anyone knew, $440M was gone and none of it came back. They could have calculated that risk before they hit Go.

Compare to Morgan Stanley: Before putting a GPT-4 assistant in front of their 15,000 financial advisors, they built it so nothing reaches a client until a human signs off. Detection instant, reversal free. Blast radius: near zero.

Gartner expects over 40% of agentic AI projects to be scrapped by 2027, partly for inadequate risk controls. The survivors won't be the boldest or the most careful. They'll be the ones who ran the numbers first.

An alchemist only learns the blast radius when something blows up.
A chemist does the math.


—
Source: “Gartner Predicts Over 40% of Agentic AI Projects Will Be Canceled by End of 2027," Anushree Verma, June 2025