I used to judge my days in isolation.

Good day. Bad day. Productive day. Wasted day.

One bad meditation and the whole morning felt ruined. One missed experiment and I was “off track.”

This is the single-point trap. You evaluate yourself based on one data point. And one data point is noise.

Stock traders figured this out decades ago. They don’t buy or sell based on a single candle. They use moving averages: the 50-day, the 200-day. The longer the window, the less the noise matters. The trend is what counts.

I started applying this to my own systems.

Instead of asking “how many experiments did I run today?” I ask: what does my 55-day average look like?

Instead of beating myself up over one rough Monday, I zoom out to 89 days. Am I trending up or trending down?

The answer is different. It’s calmer. It’s more accurate.

📈

What would your two-month trend tell you that today can’t?

One guy paid me nine dollars for Multi before it was ready.

The page had elements rendering twice. The sign-in flow didn’t exist.

He paid anyway.

I saw the notification and sat with it. Felt good. Made me bullish on Multi.

If he came back, he couldn’t sign in. Couldn’t create another video. Couldn’t track his credits.

The first dollar changes the physics. Before it, you have a project. After it, you have traction. Traction creates expectations. Now you want to deliver.

I spent that night closing all three gaps. They were on the roadmap already. The dollar moved them to tonight.

⚡

What could you put in front of one person this week?

I built a tool that turns blog posts into videos. It’s called Multi. It runs from the command line: post in, video out. The output is rough. Captions drift. Stock footage misses.

The core pipeline file is called skateboard.mjs.

The name wasn’t random. It’s the philosophy.

“A complex system that works is invariably found to have evolved from a simple system that worked.”

John Gall wrote this in 1975. Most people ignore it. They try to build the spaceship first:

• The “perfect” system
• The “complete” feature set
• The “final” architecture

Then they launch. And it doesn’t work.

The skateboard works because it has one job: move you forward.

It isn’t polished. But it works. And once something works, you can iterate.

Henrik Kniberg drew the original five: skateboard, scooter, bicycle, motorcycle and car. I kept building past the car and needed four more.

Level	Vehicle	What It Is	Who It Serves
0	🛹 Skateboard	End-to-end. Manual. Proof of concept.	You
1	🛴 Scooter	One step scripted. Rest is manual.	You + early testers
2	🚲 Bicycle	Pipeline. Simple input, complete output.	You + first users
3	🏍️ Motorcycle	Self-serve. Landing page. Billing.	Early external users
4	🚗 Car	Outbound. Prospecting. Cold outreach.	Market
5	🚁 Helicopter	Referrals. Content. SEO. Inbound.	Growing user base
6	✈️ Jet	A/B testing. Funnels. Optimize everything.	Scaling user base
7	🚀 Spaceship	API. White-label. Others build with it.	Ecosystem
8	🛰️ Space Station	Marketplace. Community. Others build on top.	Everyone

The skateboard is a tool. The space station is a platform.

The mistake is trying to build the space station when you don’t have a skateboard.

---

Multi is at Level 1. It’s a scooter. skateboard.mjs taught me what to automate first: captions and scene timing. The rest is still ugly and manual. And it works end to end.

Gall’s Law is undefeated. The skateboard is what teaches you what to build next.

🛹

What’s the skateboard version of your biggest idea?

One day I heard a podcast on ratchets. I didn’t need it yet.

A ratchet turns in one direction and resists turning back. The first turn is the hardest. Each turn after that gets easier. In relationships, it looks like this:

• You overlook one moment. That’s the hard turn.
• The next one is easier to overlook. And the next one easier still.
• Each turn deepens your investment.
• By the time something is bad enough to act on, you’re too embedded to leave. And reversing is nearly impossible.

The frog is already boiling.

Months later I was staring down a friendship that had crossed a line. My instinct was to let it go. Talk it out. Give the benefit of the doubt.

Then I thought about ratchets.

I could see the shape of the next few years. Overlooking this moment would make the next one easier to overlook. Each turn would deepen my investment. I’d watched this pattern consume a past relationship. Years of overriding my own judgment.

The lesson sat dormant until I needed it. The ratchet concept wasn’t useful when I heard it. It became useful when I had to make a decision it applied to.

This is what learning does. It gives you lenses you can’t predict the use date for. They sit quietly in the background until a situation triggers them.

“How does reading books help me make money?”

That question looks at learning on the wrong timeline. It doesn’t produce immediate returns. It may look wasted until the moment you need it.

And in that moment, it can save you years.

🔧

What’s one thing you learned that sat dormant until the right moment?

I tried voice-to-text as an experiment. That made Cursor on mobile possible. Cursor on mobile made a daily questions system possible. The daily questions system made writing posts on walks possible.

Four experiments. Each one laid a foundation the next one required. I couldn’t skip steps. I couldn’t see step 4 from step 1.

That’s an experiment chain. One experiment creates the conditions for the next. You don’t plan the chain. You discover it by running the first link.

Some chains are sequential like that one. Each experiment is a prerequisite for the next.

Other chains branch. Someone learns to cook as an experiment. That one skill splits into meal prepping to save money, hosting dinners to build a social life and starting a food account to attract an audience. Same first link, three different chains running at once.

Sequential chains go deeper. Branching chains go wider. Either way, the first experiment is doing more work than you think.

The pattern is the same in both: the first experiment was driven by curiosity. The last one produced real value. The gap between those two is the chain you can’t see from the start.

You can plan an experiment. You can’t predict where it leads.

🔗

What got easier this year because of something unrelated you did last year?

I get one move per turn in chess.

That constraint rewired how I prioritize everything.

Each night I write down my top tasks. Typically around eight. Then I force rank them. A strict order from one to eight.

I started this around 2018 when I was getting back into chess. In chess, multiple moves compete for the top spot. But you only get one.

Chess taught me to reason through my options and force rank them.

If you don’t rank options, entropy does. Whatever is loudest or easiest wins.

I sort moves by going head to head. Ship the landing page vs. write the report. Landing page wins. Landing page vs. fix the onboarding bug. Bug wins. I keep going until I’ve been through the whole list. Whatever survives is number one. Then I run it again for number two.

Works with eight tasks or eighty.

If number one is blocked, it needs to be broken down further.

If it’s blocked by something you don’t control, it shouldn’t be on your list.

🏁

What’s one move you can make that would change your week?

One day I overheard a manager trying to place someone on a team. The person didn’t want it. I did.

So I cut in: “Take me instead.”

The manager said he couldn’t. But he asked the team lead behind the scenes.

The team lead said yes. That team became the best working environment I’ve had.

The ask is the experiment. The answer is just data.

A few months ago, mid flight, half the plane was coughing. I asked the attendant for a mask.

I had no idea if they carried masks. I controlled whether I asked.

They had a mask. But the ask was the experiment either way.

That’s the line between an experiment and a guarantee. You control whether you try. You don’t control whether it works. If you control the outcome, it’s not an experiment.

Some people wait until they feel certain before they act. But for most of what matters, that certainty never arrives.

The world runs on distributions, not certainties.

You can make the right call and still lose, like a good poker hand that doesn’t hit. Over enough attempts, the right moves trend up. But each individual one? Uncertain by design.

🎯

What have you been waiting to try until you’re “sure” it’ll work?

7 days a week I was checking emails, answering support tickets and coordinating with the team. Doing those tasks once a week seemed impossible.

I was in too deep to think about anything else.

Then I got to three days a week. Monday, Wednesday, Friday. Still couldn’t imagine once a week from there.

Then I got to two. And from two, once a week seemed possible for the first time.

I started putting systems in place.

• Tier 1 support handled most issues before they reached me.
• Recurring tasks got templated or delegated.

Each system I built at two days a week made one day a week seem more real.

Now I do admin once a week. And from here, I can see something I couldn’t see from two days a week: maybe once every two weeks. Maybe once a month.

That seemed impossible before I arrived here.

Each level reveals the next one.

🏔️

What’s your goal and what’s your next hill?

• Trying Cursor was an experiment.
• Trying Cursor on mobile was an experiment.
• Trying voice-to-text was an experiment.
• Daily questions to daily posts was an experiment.

And now, with all of those in place, I can do this: write posts while I’m out on a walk.

But I couldn’t have done this without the others.

Experiments are composable.

All experiments give you information. Some experiments give you infrastructure.

Infrastructure experiments unlock other experiments that weren’t possible before.

• Without voice-to-text, I can’t capture thoughts while walking.
• Without Cursor on mobile, I can’t write on the go.
• Without the daily questions system, voice notes never become posts.

Each one lays the foundation for the next.

🔗

What experiment could you run that unlocks others?

I did not start HeadsUp with code.

I opened Notion. Made a document by hand. Sent it to eight founders. “Here is what your competitors are doing.”

Ugly. Manual. Hours of work per document.

Three people wanted to pay for it.

That Notion document was the dirty vertical. The entire system, end to end, held together with “tape.” It did one thing: delivered value to a person. No infrastructure. No scale. No elegance.

It worked.

Thirteen years earlier, Shommi was the opposite. Months of coding. Zero users. Lesson learned the hard way.

John Gall wrote it in 1975.

“A complex system that works is invariably found to have evolved from a simple system that worked.”

This is the same principle as skateboard to spaceship. I’ll write about that one separately. But “dirty vertical” hits different because it emphasizes the mess. The skateboard metaphor is clean. It implies a polished, minimal product.

A dirty vertical is not polished. It is a Google Doc emailed after a Stripe payment. It is a spreadsheet you update by hand every Monday morning.

But it is vertical. It goes all the way from input to output. A user can touch it, use it, pay for it. That is what matters.

Some builders do the opposite. A beautiful database schema. A “perfect” API. Gorgeous components. The stack never reaches the user. Pretty. Useless.

The Notion doc became HeadsUp. Shommi became a lesson.

🏗️

What’s the simplest version of your best idea that someone could use today?