… And for any new readers, welcome to Signal // Noise — the newsletter read by CEOs, founders, & scrappy builders every Thursday. Each week, what I’m listening to, one deep dive, notes from the field, three links worth your time. No buzz, no bullshit.

🎧 Want the whole vibe? Find & bookmark the running playlist right here.

Last week's issue on barrels vs. ammo really hit a nerve.

I got more texts and DMs about it than anything I've sent in a while, all some version of the same thing: "yep, that's my exact problem."

Here’s a quick refresher.

As a business owner, your output isn't capped by how much talent you have or solely raw headcount.

It's capped by your barrels, the rare people who take something from idea to done, who pull others along, who edit their own work without you hovering.

The concept comes from Keith Rabois's "How to Operate" talk, where he categorizes strong people into ammunition and barrels and says you can only ever shoot through the number of barrels you've got.

Everyone agreed with, or could relate to, the diagnosis above.

Which means everyone's now stuck on the obvious next question: okay, so how do I actually find a barrell?

A barrel is defined by a behavior, carrying something all the way to the finish.

But interviews mostly test a completely different skill: how well someone talks about work. So if you strictly hire on articulation, communication, past experience and/or pedigree, then you can’t be shocked when the killer résumé turns out to be something less than ammunition.

And it matters more this year than it did five years ago.

Agentic AI has collapsed the cost, time, and headcount it takes to produce work.

HBR just named it the "second great compression." When producing stuff gets cheap, "can produce" stops being the valuable trait. What's scarce now is "can own an outcome and finish it."

What’s scare now is judgement and discernible taste.

That's a barrel. Which makes spotting one the highest-value hiring skill you've got right now.

Hunt for end-to-end ownership.

Ask what they drove alone, idea to shipped. Ask what metrics they were solely responsible for. Then dig into the part where it got hard. Barrels push through the friction. Ammunition waits for direction. Barrels have resilience. Ammo often folds.

Watch them work before you commit.

The research is blunt here: unstructured interviews and résumé reviews are some of the weakest predictors of performance we have, while structured interviews and real work samples are among the strongest. A small paid trial project tells you more than another hour of talking.

But the project is where people sometimes drive me crazy. Don’t treat this lazily, as something to cross off or just quickly produce with ChatGPT.

For example, right now I'm helping a founder hire a VP of marketing, and I spent five hours last week building the project brief and the supporting data workbook. Five hours. Because the candidates’ analysis is only ever as good as the inputs you hand them.

So it kills me when people think a boring one-pager or "eh, I'll just have ChatGPT knock something out," is gonna surface a real operator.

⚡ The quality of what you learn is capped by the quality of what you gave them.

Reference-check and back-channel for autonomy, not competence.

Don't just call two references and follow boilerplate corporate jargon. Ask what the candidate owned end-to-end, and what happened when nobody was watching? Ask what the one thing was that the candidate was actively trying to improve?

And please, learn how to back-channel properly and incorporate it into your process — you cannot rely solely on candidate-supplied recommendations.

Every one of these tries to really focus on performance.

A barrel can’t fake a real work sample, can’t script a truly honest back-channel, can’t rehearse what they owned when nobody else was watching.

Keep those three items top of mind next time you’re hiring for a barrel.

Incase you missed Tuesday’s launch email, I'm officially opening up Hiring OS to readers: done-with-you help landing a senior marketing leader (Director, Head of, VP, CMO), an ops leader like a COO, or a comparable senior seat, using the same process I run with clients.

If you've got a senior role to fill this quarter and you're done guessing, the details are in the link above or you can reply to this email, too.

Microsoft CEO Satya Nadella dropped a long one on X (already north of 11M views and climbing), and the core idea is sticky: in the AI age you pay for intelligence twice, once with money and again with the proprietary knowledge you have to feed the model to make it useful. His fix is a "trust boundary" so your company's know-how compounds for you, not for whoever owns the model. You don't have to buy the whole framework to feel how real that tension is.

Uber CTO Praveen Neppalli shared some numbers: 99% of Uber's engineers now use AI tools, and more than 70% of pull requests come from local or cloud agents. This week's Signal argued that agentic AI is collapsing the cost of producing work, which is exactly why "can own an outcome" now beats "can produce." Uber is what that shift looks like at scale: when agents handle most of the output, the human edge moves to ownership and judgment, not raw production.

Every time someone tells me they're selling stocks to "let things cool off," I think about my favorite investing data point (”days in market”) or just honestly, how silly that is. Now, this piece by Nick Maggiulli, only helps my cause. He ran a full century of US stock returns and found there's never been a 20-year window with a negative real return. As he puts it: "The bull market didn't start in March 2009. It started in 1776."

When was the last time you made an incredible hire?

—> Tell me. I read every single reply.

—> Or pass it to someone who’d get value from it.

Hope everyone has a great weekend.

And until next time, thanks for reading.

Jordan