The Brown M&Ms Test

The smartest clause in the history of contracts was about candy.

Buried deep in the technical rider, between electrical load specs and structural weight limits, one line: "M&Ms (WARNING: ABSOLUTELY NO BROWN ONES)."

That was Van Halen. 1982 world tour. For decades, every journalist on the planet told you the same story: pampered rock stars throwing tantrums over the color of their candy. Peak excess. Celebrity brain rot.

They were all wrong.

Nine Eighteen-Wheelers

Van Halen's 1982 production was a monster. Nine eighteen-wheeler trucks of gear. Massive staging, pyrotechnics, lighting rigs that could crush someone if a structural number was off. They weren't rolling this into Madison Square Garden. They were playing third-tier markets. Arenas built for hockey games and county fairs.

The contract rider was so technically dense that David Lee Roth said it "read like a version of the Chinese Yellow Pages." Every page was load-bearing. Electrical requirements. Stage weight limits. Flooring specs. The kind of details that, if you get wrong, someone gets hurt.

So Roth buried a tripwire.

No brown M&Ms.

Deep in the rider. Easy to miss if you were skimming. Impossible to miss if you actually read the thing.

If Roth walked backstage and the M&M bowl was clean, good. The promoter read the contract. Probably got the rigging specs right too. But if he spotted a single brown M&M in that bowl, he knew instantly: this promoter did not read the rider. If they skimmed past the candy, they skimmed past the structural requirements. Guaranteed.

Now Let Me Tell You About the Word "Client"

At Sagan, we never call the people we work with "clients." Ever. They are customers or members. That language is deliberate. (this is a whole different discussion)

AI defaults to "clients." Every. Single. Time. ChatGPT, Claude, Gemini. It doesn't matter which model. If you don't override it, you get "clients."

So when a document crosses my desk from someone on my team and I see the word "client," I know two things:

1. It was mostly written by AI.

2. They didn't read the output before sending it to me.

Point two is the actual sin.

Using AI is fine. I want my team using AI. But using AI sloppily? Shipping output you didn't bother to review? That tells me something about your attention to detail across the board.

If you didn't catch the vocabulary, you didn't catch the logic gaps. You didn't catch the wrong number. You didn't catch the recommendation that contradicts what we discussed last Tuesday.

The word "client" is my brown M&M. It tells me the rider didn't get read.

The M&M was never about the M&M.

Find Your Brown M&M

Every business has one. Or should.

The word your AI reflexively uses that your company never does. The phrasing that sounds "almost right" but isn't yours. The little signature move that reveals whether a human actually reviewed the draft or just hit paste and send.

Maybe it's "leverage synergies" when your team says "stack." Maybe it's "I hope this email finds you well" when your culture opens with "Quick one." Maybe it's "clients" when you say "customers."

Whatever it is. Find it.

Build it into your team's standards. Make it explicit. Tell your people: "This is the word we never use. If I see it, I know the rest didn't get reviewed."

Here's what's coming. AI writing is about to flood every inbox, every doc, every Slack thread, every proposal. The volume will be overwhelming. Most of it will sound the same. Competent. Polished. Generic. Soulless.

The people who stand out won't be the ones who avoided AI. They'll be the ones whose output still sounds like them. Because they actually read the damn thing before shipping it. They caught the brown M&Ms. They verified the rider. They made sure the staging wouldn't collapse through the floor.

Roth wasn't crazy. He was the most detail-oriented person in the building.

Be Roth.

Yallah Habibi

Jon

P.S.

I made a rap video!