… And for any new readers, welcome to Signal // Noise — the newsletter read by founders, CEOs, and 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? The 2026 playlist is right here.

There's an a16z essay that took over my feed last week.

It's by Joe Schmidt, a partner there, called "Avoiding Death on the Yellow Brick Road."

Over a million views, reposted everywhere.

Dharmesh Shah, who co-founded HubSpot, told people that if they only had time for one long read that day, make it this one.

So I read it once - and had Claude summarize it for me once, too.

The setup is about AI. The labs (OpenAI, Anthropic) are swallowing the obvious, horizontal work. Code, writing, images, etc.

Schmidt's point is that the companies that survive get built in the messy, vertical corners the labs can't be bothered to reach.

He also talked about survivability in the age of AI, moats, and building something for the long-run.

He introduced a very specific (but simple) test: are you a system, or a tool?

Here's how he puts it:

System, or tool. A great way to think about durability in the age of AI.

It’s easy to think about system vs. tool when it comes to SaaS, workflows.

But let me introduce a way you can think about the system vs. tool concept in the not-as-obvious worlds of community, advice, media, services.

Years ago at The Motley Fool, I was in charge of our $100M+ renewal book.

For a long time I told myself the renewals came down really to one thing: the stock picks.

If subscribers made money off good picks, they renewed. If they didn’t, they churned.

And sure, the picks mattered.

People were making real money on the stocks, so they stuck around.

It’s hard to churn out of a portfolio that's going up.

A great stock pick is a tool. So is sell-side guidance.

"Buy this, sell that" feels like the product, but in the end, it could be copyable.

At some point a model is going to claim it can find alpha on both ends, the buys and the sells.

When that day comes, "we pick good stocks" saves nobody.

So why I don’t think a company like that will be replaced in the future?

Because investing isn't a math problem. It's a behavior problem.

The hardest, most valuable thing we ever did had nothing to do with picking.

It was keeping people from blowing themselves up.

When the market tanked, when an election broke the wrong way, when inflation ran hot and the banks wobbled, people wanted to dump everything - every single time.

And we were always on the other end.

The emails. The phone calls. The late-night "should I just get out?"

Part of our job was to talk them off the ledge. Hold. Breathe. Don't sell at the bottom.

Again, and again, and again.

That's brutally hard. And it was the whole ballgame.

A model can tell you what to buy. It can tell you when to sell.

But it can't be the steady, sympathetic voice that stops you from torching your portfolio at 2am when you're scared out of your mind.

That part - I think - is mostly human.

And that's the system.

Most of us think the moat is the product, the feature, the thing that gets demoed on a call. So that’s the thing that gets polished, bolted on to.

But the feature is the easiest thing in the world to copy.

What's hard to copy is being the one the customer trusts when it counts. The hand on the shoulder when sh*t really hits the fan.

And in some businesses, that could be the entire moat.

Think wealth management. Healthcare. Coaching. Law. Real Estate.

A model can hand you the portfolio, the diagnosis, the playbook, bu it can't be the person you trust enough to follow when every instinct says run.

In those fields the output was never the hard part. The trust was.

So before you spend another quarter making the product 10% better, sit with the harder question.

If a model did your job tomorrow, what would people still need you for?

If the answer is "nothing they couldn't get cheaper," you've got real work to do.

And it’s not just another feature.

It's the one who "seems good enough right now." You hire the conversation instead of the job, the friend of a friend b/c it was easy, the kid out of college because they were cheap… and then six months later the team's quietly reorganized around their weaknesses.

I put five of the frameworks I use to catch it before it happens into a free 8-page guide.

A founder I came across lost a closed-won deal because his ex-employee pulled public Fathom links, learned exactly how he was positioning, and undercut him on price. Same week, his Slack agent broadcast company revenue into a general channel by mistake. If you're stacking AI tools and trading privacy for speed, this is your reminder to actually check permissions. Be careful!

Anthropic just added $10 billion in revenue in a single month, roughly twice Databricks' entire annual run rate. ServiceNow took 20 years to cross $10b; Anthropic did it in under four. Tomasz Tunguz runs the napkin math on when that pace would make it the most valuable company on earth, and the answer (as little as three years if growth holds) is either thrilling or terrifying depending on where your money sits.

The biggest companies in the market are now also the fastest-growing, most profitable, and most dominant, which isn't supposed to be possible all at once. Josh Brown's read on the AI-fueled melt-up is the rare one that holds both truths at the same time: this is abnormal, but abnormal can run for years longer than you'd think. Ain’t nobody selling, here. It's a reminder not to build your whole financial identity on the bet that this pace compounds forever.

Take a hard look at your business and try to figure out what’s system, and what’s tool.

Hope everyone has a great weekend.

And until next time, thanks for reading.

Jordan