Private notes, made public. · 7th Aug, 2026

Hey, it’s Gabriel.

These are my private notes this week on building wealth, raising my family, and breaking generational patterns.

This week
⚡ I stopped writing prompts. My agents interview me instead
❤️ a fake ultrasound, and the lies we were all raised on
💰Cloudflare beat earnings, my LEAP calls crossed +150%, and I’m converting
🧬 and the detox verdict I owed you

Here’s what happened. I needed an email for my list, so I did the thing everyone does: typed a prompt telling my AI to draft it, hit enter, and walked away.

The draft came back. I deleted the whole thing. Generic step-by-step filler I would never send, the kind of content that sounds like everyone and belongs to no one.

And before you blame the AI: mine should know me. It’s plugged into my second brain. Two years of my calls, my clients, my stories, all of it. It still produced something generic.

Because the problem was never what the AI knows about me. The problem is what I never told it: what I actually wanted from this specific email. I say this to my students all the time, and this week I proved it on myself: your agent cannot read your mind.

See, for months now my real workflow has been the opposite of prompting. Before my agents plan anything, build anything, or write anything, the first step is always the same: they interview me. Ten questions, one at a time. When my answer is vague, they push back and ask for the story behind it. Only then does the plan get written, from my answers and nothing else.

This week I skipped that step. Not because I decided to. Because prompting is the reflex. We’ve all been trained since day one of ChatGPT to open a box and type instructions at it. That habit runs so deep that even after months of working the other way, my fingers still defaulted to it. One lazy prompt, one generic draft, straight into the bin.

Think about what a prompt actually is. A prompt is me guessing what the AI needs to know, then writing it from a blank page. And half the time I don’t fully know what I want until something forces me to say it out loud. But answering questions? Ten questions in, what I actually want is sitting there in my own words. The interview extracts the one thing no database of me contains: what I want, right now, for this specific thing.

You don’t walk into a doctor’s office and dictate your own prescription. The doctor asks questions first, because the information that matters is inside the patient. And notice: even a doctor holding your complete medical file still opens with the same question. Where does it hurt today? The file knows your past. Only the interview gets at what you need now.

Regular readers have watched this sneak up on me without me naming it. The AI twin in #023 started with a 31-question interview to catch what the data got wrong about me. Back in #014, my agents interviewed my team members about their workflows before redesigning them. This week I finally admitted the pattern: the interview is the method.

The more capable the models get, the less value there is in me telling them how to do things. The value left is what only I know and what only I would choose. My lived experience and my taste. The interview is how you extract both.

When was the last time your AI asked YOU a question before it started working?

Getting interviewed is step one. Inside Two Hour CEO, I’ll show you how to turn those answers into systems that run without you, step by step, even if you’ve never touched a line of code.

I dropped a note into my second brain this week after a video on X stopped me cold. Parents cured their kid’s junk food habit by showing him a fake ultrasound: look, there’s a monster growing in your stomach. The kid was terrified. The kid stopped.

It worked. That’s the problem.

Because that was the whole playbook of the typical Asian family my generation grew up in. If you don’t behave, the policeman will catch you. Keep crying and the old rag-and-bone man will take you away. Finish your rice or your future wife’s face will have pimples, one for every grain you leave behind. If you grew up in an Asian household, you have your own list.

Our parents weren’t cruel. Fear is just efficient. It’s the cheapest tool in the drawer and it works instantly, which is exactly why an exhausted parent at 9pm reaches for it. I understand that temptation now in a way I never could as a kid.

But here’s the long-run bill. NTU followed 379 Singaporean young adults and found that the more their parents lied to them as children, the more they lied to their parents as adults, and the harder they found it to adjust to adult life. The fake ultrasound teaches one lesson, and the lesson has nothing to do with junk food. It teaches that the people you trust most will invent reality to control you.

Fear buys obedience. Truth builds judgment.

The scared kid obeys while you’re watching. The kid who understands why makes the right call when you’re not in the room. Only one of those grows up able to think.

What’s the lie you’re tempted to tell this week because it’s faster than the truth?

The News: Cloudflare (NET) reported its second quarter last night: $696 million in revenue, about 5% above what Wall Street expected, and they raised their guidance for the full year. The stock jumped past $300 after hours, up around 16% as I write this, clean through its previous all-time high.

The Position: You’ve watched this one build across three letters. I entered LEAP calls in the $190 to $220 range, betting on the wave of agentic AI traffic that has to pass through Cloudflare’s pipes. Last issue I told you I sold half during the run-up and got back more than my entire original stake. The calls I kept are now up over 150%. The stock itself is up roughly 50% from my first entry; the calls multiplied that because LEAPs are leveraged, in both directions, which is the whole point of the instrument and the whole risk.

The Move: I’m converting the remaining calls into shares to hold long term. LEAP calls expire. My thesis doesn’t. When a leveraged trade finishes its job, you either take the money and leave, or you become an owner on purpose. I’m choosing owner, and after last issue’s harvest, I’m doing it with my original capital already back in my pocket.

The Thesis: Why hold this one for years? Because the internet is getting its second population. By Cloudflare’s own count, automated traffic passed human traffic on the web this year for the first time. Soon your agent will talk to my agent, negotiate, and pay. And an economy of agents needs the same things every economy needs: identity, rules, and a way to move money. Cloudflare is building all three. Web Bot Auth gives agents cryptographic passports so sites know who’s knocking. Pay Per Crawl charges AI companies for the content they consume, built on HTTP code 402, “Payment Required,” a code that sat dormant in the web’s spec for nearly thirty years waiting for exactly this moment. And they’re building wallets with spending guardrails so agents can buy on your behalf. Even beehiiv, the platform that delivers this letter to your inbox, is wired into their AI control layer.

In #013 I called Cloudflare the plumbing of the agentic internet. After this week I’m upgrading the word: they’re building the toll booths. When one agent pays another for data, access, or work, some layer has to verify, meter, and settle that transaction. I want to own the layer, so I’m converting from renting the move to owning the road.

The Confession: Two issues ago I told you my brain had recalibrated to AI speed, and I prescribed myself a dopamine detox with a promised before-and-after. Honest part first: I scoped it down. The full monk-mode day with nothing engineered in it hasn’t happened yet. What I actually ran was one full day with social media and news completely cut off.

The Setup: I didn’t trust my willpower, so I removed the choice. I used an app called Opal to block it all at the phone level, and in its strictest mode you cannot unblock it early. No “just five minutes.” The door stays locked until the timer runs out. Long-time readers know this about me already: I’m still the guy who needs software to keep him off his own phone. I’d rather be humbled and improving than proud and scrolling. That’s the Architect-First rule I keep coming back to: design the environment so the decision is already made before the willpower gets tested.

The Verdict: One day is all it took to feel the difference. I slept noticeably better that night. Woke up in a better mood. And I had more drive to get things done the next day than I’d had all week.

One day.

The Honest Part: this is one person, one day, zero control group. Your mileage may vary. And the deeper detox I promised is still owed. Consider this the first installment, and it worked well enough that the full version is now a matter of when, not if.

As promised, my private notes. No fluff.

See you next week,

Gabriel Judah

P.S. The Interview-First Prompt, the machine audit, the twin: they’re all “Loops.” Inside Two Hour CEO, over 200 members are building their own, step by step. You don’t need to be technical. If you can answer ten questions, you can build one. Grab a seat here.