I built the wrong kind of smart

I'm Jeremy Hurst, VP of Autonomous GTM at Swan. I left a $450K VP Sales role to take a job with a made-up title. I did it because I got a glimpse of the 100x seller - it stopped being a thought experiment and became the actual playbook. This is a field journal for GTM and growth leaders who can feel the model changing under their feet and prefer changing with it to being left behind.

I built an AI agent for spotting upsell opportunities. It has 12 weighted signals & a precise scoring formula. Then Amos Bar-Joseph sh*t all over it 😤

"Don't make such a rigid scoring model. Leave more space for the model to exercise its reasoning & judgment."

It's the 3rd time he's said something like that to me - it's finally sinking in.

Because when I think about the biggest calls I've made, none of them ran through a deterministic model.

Take joining Swan AI.

I took a 20% pay cut - went from managing a team of 30 to managing a fleet of agents.

There's no spreadsheet that makes that math work.

While I was weighing it, I was walking along the LIC waterfront and happened upon these 2 swans. That's what tipped it for me.

Same thing happened in 2009. I saw my future wife across the bar. No scoring system. No pro/con list. I just walked over.

Most decisions that matter work like that. Not measured. Felt.

When you hear all the talk about "vibe coding" or "vibe GTM-ing" (shout out Ariel Cohen), that's what "vibes" actually is.

Not fluff. Compressed judgment. Years of pattern recognition your brain doesn't bother explaining to you.

I built an agent that tried to replace that with a formula.

Amos wasn't saying add more signals. He was saying stop engineering out the one thing that makes the decision good.

I don't fully know what that looks like in practice yet. But the rigid version wasn't it.

Vibes are not the opposite of rigor.

They ARE rigor. Just compressed (if done right).

getswan.com

THE SIGNAL

I used to think "vibe GTM-ing" meant the industry had stopped being rigorous. Now I think it's what rigor looks like after enough reps that you don't need the spreadsheet anymore. That's what I'm trying to build into these agents, not engineer out of them.

THE PLAY

The Judgment Audit

• Pull the last AI agent or scoring model you built for a GTM decision.
• Count how many inputs are there because they're measurable, not because they're what actually moves the call.
• Ask whoever made that call best before the model existed what they were really weighing.
• Rebuild one rule as a judgment prompt instead of a weighted score.
• Run both versions for 30 days and see which one your best rep would trust. (~2 hours to set up, 30 days to know)

SIT WITH THIS

What's the one call your best rep makes that you couldn't turn into a rule?

I don't have the rubric for this one yet. That might be the point.

-Jeremy