58% of buyers now start their research in ChatGPT or Gemini, not Google. Most startups aren't showing up there yet.

The ones that are get cited by the AI tools their buyers, investors, and future hires already use. And they convert at 3×.

Download the free AEO Playbook for Startups from HubSpot and get the exact steps to start showing up. Five minutes to read.

Get the free playbook

Your ears rang. We took notes. You're welcome.

Friday · 6 min read · Lightly edited by a human who definitely didn't listen at 2.5x speed (lie)

Good morning, capo. This week the government noticed AI. Washington spent the week circling Anthropic's Fable model, the All-In table spent an hour on whether the whole industry is about to get nationalized, China got quietly locked out of America's biggest IPOs, and Perplexity's CEO went on 20VC to argue the export controls meant to kneecap China have instead been forging it into a stronger competitor. The story is no longer the technology. It's the state. Also, a founder explained why his company is a guitar and not a microwave, and a man built a newsletter for the price of hay. Let's go count someone else's money.

20VC — "Perplexity CEO: How Export Controls Helped Not Hurt China, and Power Is the Bottleneck to AI" (listen)

1h+ → 90 seconds

Perplexity's CEO came on 20VC and calmly detonated three pieces of conventional wisdom. Start with the one every investor needs: the bottleneck for AI is no longer chips or models. It is electricity. "The biggest problem today is actually power. You have to go secure land, or lease a property, and buy a bunch of turbines to generate power, or work with grid suppliers, and you also have to work on cooling." The most advanced industry on earth is now gated by the least glamorous thing on earth, and he draws the obvious trade: "It might not be inconceivable that Micron, the supplier of memory, is worth more than Meta in the next 6 to 12 months. Whatever is the bottleneck will command the price."

The second detonation is aimed at half the startups in your feed: "The model is no longer the product. If you're literally just a reseller of model tokens, you have no business, because the model will get commoditized." The thing everyone is racing to build is the thing about to be worth the least. The value moves to whoever owns the scarce layer, which this year is power and memory, not weights.

The third is the spiciest, because it says the policy worked backwards: "The only reason there's even a 12-month gap between open source and frontier is export controls. It's definitely helped. But there's a chance that because of that, they now get really good at the physical layer. You're converting them into a far more potent competitor." Sanctions as an unintentional training program.

Takeaway for operators: find your scarce layer and own a piece of it. In AI right now that is power, memory, and cooling, not another wrapper on someone else's model. The bottleneck always sends the invoice, and it always wins.

Odd Lots — "How a Vibecoded Newsletter Is Making the Hay Market More Transparent" (listen)

1h+ → one idea you can steal (and we mean steal)

This is the most Pod-Capo episode that has ever existed: a man built a one-person media business around the price of hay, and accidentally laid out the cleanest playbook for a niche newsletter we've heard. The whole strategy fits in one sentence: "Find a piece of public data that exists, that's consumable for free, that nobody is regularly looking at. Gather it, become the central repository for it, and then occasionally comment on it."

The genius is in what AI changed. "This is a very clear example of a new type of business that might not have been economical to run. You wouldn't have been doing it while you're in college." Vibecoding the data pipeline dropped the cost of "become the system of record for a boring market" to roughly zero, which means a thousand neglected markets are now one motivated person away from a Bloomberg terminal. His own motto, with no irony: "The Bloomberg for Hay."

And he understands exactly who he threatens, which is the test of whether a transparency business has teeth: "The middleman will always feel the outcomes of more transparency, because they're the ones making deals kind of blindly, kind of just guessing." Every opaque market is a middleman's salary. Publish the prices and you become the most important and most hated person in the room.

Takeaway for operators: pick a market everyone trades and nobody measures. Aggregate the free data, become the source of truth, monetize the attention. He said the quiet part with admirable honesty: "Hay is definitely not going to get me a girlfriend." Maybe not. It might get you a business.

Invest Like the Best — "Clay's Unusual Path to Building a Multi-Billion Dollar Company" (listen)

1h+ → a founder who thinks differently

Clay's founder built a multi-billion-dollar company while sounding less like a CEO and more like a monk who happens to run go-to-market software, and the result is the most quietly radical interview of the week. His theory of risk is the engine: "Capitalism rewards risk. You need courage to make a leap and commit to these ideas. To take real risk, you need to genuinely not know what's going to happen." Most founders manage risk down to zero and wonder why the upside went with it.

The part that will stick with you is about where he creates from. He spent a stretch, in his words, "going around pretty annoyingly telling people, I have everything I need, I'm good," and his point was not smugness. It was leverage: "You can create from a place of wholeness, and in fact you will be less destructive in the world." A founder who already feels full makes different decisions than one trying to fill a hole with a cap table.

Then the metaphor that reframes the entire scale-at-all-costs religion: "We're not building a microwave. Microwaves are useful, but they're commoditized. We're building a guitar. There's just six strings and twelve frets, but you could spend your whole life getting better at it." And the strangest, best idea in the hour, a genuine heresy in startup-land: "I wish there was a death doula for companies. If your company has achieved its mission, should we help it pass? Maybe it has children and people come out and do cool stuff, rather than scaling things that become zombies." A man running a growth company, arguing that some companies should be allowed to die well. Screenshot it.

All-In — Nationalizing AI and the Anthropic "Fable" Backlash (listen): The table renewed a take that was spicy eight months ago and is now close to consensus among them: "I said Anthropic was engaged in a very sophisticated regulatory-capture campaign based on fear-mongering. People thought that was a spicy take. Eight months later, a lot of people are saying it." The practical worry for builders is concrete: "Anthropic will increasingly take in your prompts, evaluate them, and decide what to do before generating output. If you're a company, censorship risk makes it almost a non-starter." One host claims he got auto-downgraded mid-demo for asking, as "a VC and podcaster," about fertilizer-bomb regulation, then deadpanned: "I just proved the point." The throughline of the segment: the labs asked to be regulated, and may be surprised what they get.

Prof G, China Decode — China Locked Out of America's Biggest IPOs (listen): The stat that stops you cold: "Elon is now worth more than a trillion dollars, about 3.2% of US GDP. Rockefeller at his peak was 1.5%. On a relative basis, Elon is more than twice as rich as the richest American ever." But the sharper point is about demand, not supply: "The biggest obstacle to AI in America isn't supply chains and it isn't power. It's popularity. Tens of billions of dollars of data centers were blocked in 2025 because of political pushback." The bottleneck Perplexity's CEO missed: the public.

My First Million — "Most People Are Chasing Type 3" (listen): On personal brand as a built thing, not a born one: "You think I just woke up like this? I created this Tony Robbins [character]. I decided that's who I needed to be, and then I created him." The actionable core is one line every operator-turned-creator should tattoo on: "The making of the product needs to be the content." And the gut-check on the whole grind: "At a certain point you're trading very valuable hours of life energy for useless dollars you have no ability to even spend to improve your life."

What turned AI political this week, measured in airtime:

For seven issues the AI story was money: capex, IPOs, fund math. This week it became politics, which is what happens to every industry the moment it gets too big to ignore. Washington wants a say, the public wants a protest, and China wants a workaround. The bottleneck used to be chips. Now it's permission. We remain, as ever, long the turbine and long the guy selling hay data.

If you only listen to one episode this week, make it Perplexity's CEO on 20VC. It is the rare interview that hands you three trades and a worldview in under an hour, and you will never again read an AI headline without quietly asking, "yes, but where's the power coming from."

If you only have 20 minutes, the hay newsletter on Odd Lots. It is proof that the best business advice this year is "find a boring number nobody tracks and become the person who tracks it," delivered by a man with zero illusions about what it's doing for his love life.

Still no David Senra sighting. The corpus has now gone two full issues without him. We are not worried. Cryptids surface on their own schedule, usually clutching an out-of-print biography of someone who manufactured something unglamorous with terrifying intensity. We wait.

Forward this to a friend who still thinks the AI fight is about better models. It's about power, permission, and who owns the bottleneck. We're a newsletter, not a guilt trip. (Okay, a little guilt trip.)

— The Capo 🎙️🤌

Made with coffee, too many podcast apps, and the firm belief that nobody has time to listen to 15 hours of content per week. That's our job now.