New companies kill themselves trying to be everything to everyone from day one. “If we only target busy parents, what if busy parents don’t actually want this?” So they expand to busy professionals, students and anyone else who might be interested.

The problem is that a solution that’s amazing for busy parents will naturally attract busy professionals and anyone else juggling priorities. But a solution designed for “everyone who’s busy” will be mediocre for all of them.

Going broad feels safer because it seems like you’re hedging your bets. But you’re actually making a much riskier bet. You’re betting you can out-execute every focused competitor who’s solving the specific problems you’re trying to solve generally. The restaurant that tries to serve “food for everyone” ends up with a massive menu that’s mediocre at everything.

Fear-driven broadness is the opposite of product-market fit.

A big mistake companies make when building products is confusing market with target. Thinking going broad means capturing more market share. But it usually means building something mediocre for everyone instead of something essential for someone specific.

Market is the opportunity. Target is who you’re solving for. When you nail it for a specific target customer, you often discover they’re a wedge into a much larger market. Solving it really well for the target, opens the door to every other closely related segment.

This confusion shows up in two predictable ways. New companies go broad because they’re afraid their specific target won’t be enough to sustain a business. Existing companies go broad because they think new features need to work for their entire customer base. Both approaches backfire for the same reason. Specificity in who you’re serving leads to specificity in what you’re building.

Start every new project with Claude Code by asking it to create an idea.md file. Tell it to listen and structure what you write. Nothing more.

Then just talk.

Dump your thoughts into the chat. Explain the tech stack. Share what you expect the system to do. Write freely. Show screenshots of similar apps. Reference existing patterns you like.

Claude Code will organize everything into idea.md as you go.

Review the file periodically. You’ll spot gaps in your thinking. Add those missing pieces. New requirements will emerge from seeing your ideas structured. Feed those back in.

This creates a feedback loop: Talk. Read. Think. Repeat.

When idea.md feels complete and you’re ready to build, ask Claude Code to break the implementation into testable phases. Each phase should work independently:

• Phase 1: API endpoints you can test with curl
• Phase 2: Mock data and basic responses
• Phase 3: Minimal UI that connects to the mocked backend
• Phase 4: Real data sources
• Phase 5: Polish and edge cases

Every phase produces something you can verify works before moving forward.

The magic is in the constraint. By forcing all your ideas through a single markdown file first, you build clarity before code. By demanding testable phases, you avoid the trap of trying to build everything at once.

99% of the time I use a QR code it’s to browse.

The ads say: fill out the form over here, read more about it over here.

The mental model is to open the browser and… well yes… I need to open the camera app… then scan… then it automatically goes back to the browser.

That little detour through the camera breaks the flow every single time.

You can’t judge a new technology by how well it copies the old one.

It’s like looking at the first cars and asking “Can this plow my field better than my horse? Can it carry as much hay?”

The question isn’t whether it can do what came before. The question is what can it do that nothing else could do.

However, if the new tech is proclaiming to plow the field and the only thing it does is autocomplete, then judge away.

Real innovators are students first. They study, understand, then move.

Innovation is movement. You need to map the territory before you can navigate it.

When using AI most people jump straight to “make me a website about X” without doing the hard thinking first, but AI doesn’t just make us faster when used right, it makes us more thorough. The opportunity isn’t speed, it’s combining research with execution to create work that’s better than what came before. Think like a researcher and ship like startup.

I’m in my first ever improv class at age forty-six and my teacher said something that got me thinking: “I’ve forgotten what it’s like to be new.” He was explaining why teaching was harder than performing. “I have this wealth of experience that I want to share, and I just want to funnel it all into you. But that’s not how learning works.”

He introduced us to his stop word, a specific phrase we should use when we didn’t understand something. Something that would snap him out of expert mode and back into beginner’s mind.

The idea of stop words has been stuck in my head all week. Thinking about conversations where I’m the experienced one. When I feel myself jumping ahead, assuming knowledge that isn’t there, skipping steps that have become invisible to me.

In my thirties I decided to take up dancing and signed up for the laidback hip-hop class called “Mondays Groove”, thinking my dance background would carry me. I’d spent seven years learning Latin and ballroom as a kid, from eight to fifteen. Countless cha cha’s and waltzes. I knew how to move to music.

Except I didn’t. Standing in that circle while the instructor demonstrated a simple hip hop sequence, my feet felt like lead weights attached to someone else’s body. The rhythm that should have been automatic wasn’t there. The groove everyone else seemed to find effortlessly had abandoned me completely.

“Mondays Groove” might be my stop word. That moment of standing still while everyone else moved, of realizing that being new isn’t something you remember. It’s something you have to experience again.

I haven’t given anyone that phrase to use when I’m racing ahead in explanations. But I’m thinking about it. About finding my way back to the beginner’s mind I’ve forgotten.

Maybe that’s what stop words are really for. Not just to pause and clarify, but to pause and remember. To snap us out of expert mode and back into the humbling, valuable discomfort of not knowing.

It is the strangest feeling. You write a word that you know exists. And it is spelled in this exact way. But the auto correcting or spelling checker is marking it as faulty.

You cannot ignore it. It just feels right. But computer says ”no”.

You selected it, open google, paste and press enter.

Eureka, it exists. It is spelled like that.

You’ve just been gaslighted by auto correct.

One interesting way to game the meeting AI assistant is to wait until the very end, when everyone has left, and then repeat the most important decisions according to you.