One Kick, 10,000 Times: Why Depth Beats Breadth in the AI Era
I still remember the exact moment the quote hit me.
I was 19 or 20, staying up way too late trying to learn everything at once. No-code tools, basic Python, content marketing, UI design, growth hacks — the whole buffet.
I knew enough to sound smart in any group chat, but nothing felt solid. My Notion was a graveyard of half-finished projects.
Then this line from Bruce Lee stopped me cold:
"I fear not the man who has practiced 10,000 kicks once, but I fear the man who has practiced one kick 10,000 times."
I read it, closed my laptop, and genuinely didn't know what to do with myself.
Because I was exactly the guy practicing 10,000 kicks once.
Fast forward to February 2026 — and that quote isn't just good advice anymore. It's basically a survival guide.
AI didn't come for the experts first
Here's the thing nobody wants to say out loud: AI came for the people who knew a little about a lot.
Entry-level hiring in AI-exposed fields has dropped 73% since 2023. U.S. programmer employment fell 27.5% in two years — but software architect roles barely moved.
An analysis of 180 million job postings found that execution roles like copywriters, graphic artists, and photographers declined 28–33%, while creative directors held steady.
The pattern is the same everywhere: surface-level skill gets automated, deep judgment survives.
The people at the top said it out loud
Dario Amodei — CEO of Anthropic — warned in mid-2025 that AI could eliminate half of all entry-level white-collar jobs within five years.
Mark Zuckerberg casually mentioned that Meta would have AI performing at the level of a mid-level engineer by the end of 2025.
Salesforce's CEO said his company might not hire any new engineers this year because of AI productivity gains.
That world is already here.
The prompt-and-template era is over
AI writes emails, summarizes reports, spins up starter code, makes decent mockups — all in seconds.
The person operating at the prompt-and-template level, the one who knows a little about everything but nothing cold?
That person is the easiest to replace.
But depth? Depth is eating.
Here's where it gets interesting.
A landmark Harvard and BCG study gave 758 management consultants access to GPT-4 for real work tasks. The results: consultants using AI completed 12% more tasks, 25% faster, with 40% higher quality output.
But when those same consultants blindly trusted AI on tasks requiring deep expertise, their performance dropped 19 percentage points below those working without AI at all.
They stopped thinking. They outsourced their judgment to a machine that doesn't have any.
Ethan Mollick, the Wharton professor who co-authored the study, put it simply: AI consistently produces better-than-average results, but it doesn't match a genuine expert.
The GenAI Wall
A follow-up study from Harvard and Stanford named it: the "GenAI Wall."
Marketing specialists could use AI to write articles adjacent to their expertise. But developers trying to cross into marketing hit a wall — the output was mediocre because they lacked the foundational knowledge to guide the AI or even evaluate what came back.
You can't fake depth. Not even with the best tools in the world.
The market is already pricing this in
Workers with strong AI skills now command a 56% wage premium over their peers, per PwC's 2025 Global AI Jobs Barometer.
Wages in AI-exposed sectors are growing twice as fast as sectors AI hasn't touched.
A study in Science analyzing 30 million GitHub commits found that senior developers benefited most from AI coding tools — using them to expand into entirely new domains. Junior devs saw almost no meaningful productivity gain.
The people going deep are not just surviving. They're pulling away.
What Bruce Lee actually meant
Here's what most people get wrong about the 10,000 kicks quote: they think he was saying "only learn one thing."
He wasn't. Not even close.
Lee created Jeet Kune Do by studying Wing Chun, boxing, fencing, judo, wrestling, and a dozen other disciplines. His whole philosophy was about absorbing from everywhere.
His most famous motto: "Using no way as way; having no limitation as limitation."
Anti-shallow, not anti-breadth
Lee wasn't against breadth. He was against shallow.
He studied many arts, but he drilled each technique with obsessive, bone-deep intensity until it became instinct.
The quote isn't about limiting your repertoire. It's about the quality of mastery — proficiency so deep it fires without conscious thought.
In 2026, the winning formula isn't "pick one thing and ignore the rest."
It's: go so deep on your core skill that nobody — and no AI — can touch you there. Then use AI to handle the breadth.
AI becomes your horizontal bar
There's a concept called "T-shaped skills" — deep expertise in one area plus broad working knowledge across many domains.
It's been the gold standard since McKinsey coined it in the '80s.
Here's how AI rewrites the framework: AI becomes your horizontal bar.
You go deep. AI goes wide.
I lived this
A couple of years ago I needed to build the backend for an idea I couldn't stop thinking about. No CS degree, no years of experience.
Before AI, that would have meant months of tutorials or paying someone expensive.
Instead, I stayed deep on the one thing I actually owned: understanding the user problem and the product feel. Then I let a team of AIs guide me through Supabase, Drizzle, edge functions, and auth flows.
They explained trade-offs, caught security issues, suggested better patterns, and — critically — taught me the concepts so I understood why we chose one path over another.
I didn't become a senior engineer overnight. But I became dangerous in my own lane because I had tireless coaches who never got frustrated with my questions.
It's everywhere now
Backend developers are shipping full-stack apps using Cursor despite barely knowing JavaScript.
Product managers are going from idea to working prototype without waiting on engineering.
A Microsoft employee who'd never written code built a custom expense management app in two hours by describing what he wanted in plain English.
Cem Karaca, an electronics engineer and AI developer, put it perfectly. He compared Person A — a senior Python developer who knows nothing about materials science — with Person B, a PhD materials scientist who picked up Python six months ago using Claude and Cursor.
His verdict? Person B wins every time. The code is temporary, but the discovery is permanent.
His key insight: the rarest skill is no longer coding. It's knowing what to build.
The numbers are staggering
25% of Y Combinator startups in 2025 built the majority of their codebase with AI assistance.
41% of all code written last year was AI-generated.
The barrier to entry for execution has collapsed.
But the barrier to entry for judgment — for knowing what's worth building, what's secure, what scales, what users actually need — that barrier is higher than ever.
So what does this actually look like?
Pick one problem domain you want to own for years. Securing SaaS applications. Making developer workflows fast. Understanding how users actually make decisions.
Whatever lights you up enough to keep coming back when it gets boring — and it will get boring, because mastery always does.
Then use AI to accelerate the reps. Explore edge cases. Run simulations. Test five different approaches to the same problem and slowly build your own library of patterns that no model can replicate.
AI becomes your heavy bag. It lets you practice at a speed that wasn't possible before. But you still have to step into the gym every single day.
The five rules
Choose your kick. Pick the one problem you want to become terrifyingly good at. Not three problems. Not "full-stack everything." One domain where you want your name to mean something.
Use AI to compress learning, never to skip it. Ask it to explain specs, compare options, generate test scenarios — then verify everything yourself. The moment you stop checking is the moment you hit the GenAI Wall.
Run deliberate practice loops. Take one task. Solve it five different ways. Let AI suggest alternatives. Then articulate why one approach is better. If you can't explain the reasoning, you haven't learned it yet.
Build real artifacts from your depth. Turn what you learn into checklists, templates, internal tools, and frameworks. AI can help draft them, but the substance comes from your own pattern recognition.
Stay bored on purpose. AI will keep throwing shiny new things at you. The edge goes to the people who keep coming back to the same hard problem until it cracks open.
This is just the beginning
Startups have always rewarded speed. The AI era rewards speed in the same direction for a long time.
This blog and Enclave are launching into exactly that world. Every post here is shaped by my multi-AI team — I'm not hiding it. I'm showing the process.
The future belongs to humans who know how to direct the orchestra, not the ones trying to play every instrument themselves.
I'm 23. I grew up in the endless scroll, jumping between hobbies and side projects, always feeling a little behind.
For the first time, I feel ahead.
Not because I know everything — I very obviously don't — but because I finally understand what's worth going deep on, and I have the best training partners in history sitting right beside me.
MIT economist David Autor said it better than I can: AI extends the reach of expertise when built on a solid foundation. Without that foundation, it becomes a structural hazard.
The tools don't care how smart you are. They care how deep you are.
So if you're tired of trying to master 47 new things every month — welcome.
If you want to obsess over the few things that actually light you up while AI handles the rest — you're in the right place.
Human depth plus AI breadth equals an unfair advantage.
Let's practice our one kick 10,000 times.
What's your one kick?
Hit me up on Twitter/X — I want to hear it.
Crafted with my multi-AI team. Human direction, human voice, human depth — guaranteed.