the trap
Good code is cheap
Devs obsess over technical complexity cuz that's what gets respect. Job postings demand "rockstar ninjas" who know fifteen frameworks. Stack Overflow used to reward the cleverest solution.
We're trained to outsmart each other, not outsmart the market. The business of development has become a performance art where the audience is other developers.
U know what stays capped if u remain in developer role? ur earnings
Developers r in the business of looking smart, not earning.
the paradox
Market pillars in the age of LLM
On one hand: "LLMs can whip up a website for everyone." On the other: "how do u stand out in an agentic world when humans don't visit websites?"
Both things can't be true. Companies brag LLMs can build anything while simultaneously panicking about visibility in a world where agents do the browsing.
Here's what they're missing: agents don't give a shit about ur design system. They want markdown. Clean structured data. Ur $50K hero section with 47 gradients? Dead weight. MCP is becoming the standard because agents need utility, not beauty.
Bill Gurley called it: when ChatGPT rolls out ads, it wants a slice of what Booking.com pays Google. Marketplaces lose their moat when agents handle discovery and booking.
We're building cathedrals for congregants who'll never walk through the door. Design isn't dead—it's been demoted from necessity to luxury. Like calligraphy after the printing press.
hiring rituals
Silent insanity
Companies ask candidates not to use LLMs in interviews. Same companies braggin bout shipping code written with those LLMs.
Remember algorithmic thinking tests? LeetCode grinding? Years of testing devs on reversing linked lists. Then they got hired and... never used any of it. Day-to-day was API integrations, CSS layouts, debugging race conditions.
Now we're doin it again. Companies drowning in AI-generated code that "looks perfect but breaks in prod". Meanwhile comprehension debt piles up from teams who can't understand what fits and what not.
The irony: AI was supposed to make coding easier, but it's makin the thinking parts more valuable.
on value capture
Who signs off, gets $$$
Here's the extraction mechanism nobody talks bout: whoever signs off on ur work captures the big buck.
You ship the feature. Your manager approves it. Their director greenlights the roadmap. The VP presents it to the board. The CEO tells investors about "AI-driven innovation." Each layer adds nothing to the code but takes a cut of the credit—and the comp.
CEO-to-worker pay ratio: 30:1 in 1978. Now? 351:1. Not because they got 12x smarter. Because the sign-off economy rewarded leverage, not labor.
US finally admitted it: outsourcing critical manufacturing was a strategic error. CHIPS Act. Reshoring. Took ~50 years to recognize.
Same era, another pivot: the flip from workers to shareholders. Labor and capital decoupled. Productivity kept climbing. Wages flatlined. CEO pay launched into orbit.
Consider Pichai. Sitting on Transformer architecture for years. An engineer from google demos chat-with-ai to the public and gets shut down from above. OpenAI runs with it via ChatGPT and suddenly Google's playin catch-up with its own invention. The guy who couldn't recognize the product in his own research lab makes ~$200M annually. Stupid? Maybe. But he controls distribution, capital, and the signature line.
The medieval guilds understood this. Artisans owned their mark. You couldn't ship without your sign-off. Industrialization broke that pattern—factory owners captured what craftsmen produced. Now we're surprised when the pattern repeats in software?
Here's the uncomfortable truth: outsourcing wasn't a mistake. Neither was shareholder primacy. They worked exactly as designed. Shareholders won. Workers lost. The system delivered.
the influence game
Why speakers make more than builders
There's a whole economy built on extracting value from this dynamic.
Conference speakers don't need to be better engineers than you. They just need to be better at packaging insights. Course creators don't need deeper technical knowledge—they just need to position themselves as authorities.
The game isn't skill arbitrage anymore. It's attention arbitrage.
While ur optimizing for technical respectability, someone else is optimizing for reach.
my proof
What happens when coordination tax hits zero
Here's what happens when u stop playing the recognition game: u start building shit and a lot of it.
Ever notice how much energy goes into process? Waiting for review, waiting for clarification, waiting for the meeting that was supposed to be an email. Coordination is a tax. When ur solo, that tax is zero.
AI didn't make building easier. It made building alone viable. Maybe thats what Pichai was afraid of.
The "anyone can build anything with LLMs" hype is BS sold by ppl selling something. LLMs amplify existing skill and judgment—they don't replace it. But here's what is true: the minimum viable audience for a tool just shrunk from "thousands of users" to "one."
I built DevBrain. My own search engine. Not for the internet—just for my internet. A few thousand articles from indie devs I value. No SEO-gamed content farms. No Medium clickbait. Just signal. After 1 year of accumulating knowledge and fixes, it replaced Google for my programming searches. Ur domain expertise is the moat.
My proof points or my “wants to look clever” head talking:
DevBrain—a micro-search I use daily, because FU ads, FU dependency on massive search engine that directs me to other www with more ads and content locks
4 iOS apps (peak was 2.5k/mo)
Personal agent to build full-featured experiences for iOS
an in-house CDN, a "new web" with custom solutions to replace tls + http stack, F aws/gcp/azure
A pile of weekend-level projects that would've taken months pre-LLM
One person + AI agents now outputs what used to require a team. Not cuz AI is magic—because the coordination overhead of borrowing capabilities (copy, design, research, basic code) just disappeared.
This unlocks a Cambrian explosion of micro-tools. Robin Sloan called it "home-cooked apps"—software u build like a meal, just for urself. No product-market fit meetings. No edge cases for strangers.
35% of startups incorporated in 2024 had solo founders. One person, one product, one focused problem. Every.to runs five products—each maintained by one person with AI agents.
The barrier between "I wish this existed" and "okay I built it" has collapsed. A tool that makes $5k/month and serves 100 ppl perfectly? That's a life-changing business for one person. Ditch Big Tech and SaaS monstrosities.
There's a parallel with organic farmers. The ones who grow food just for themselves. They make this incredible food—real food with actual flavor—and maybe 10 ppl ever eat it. Same thing with developers building personal sites. Hand-crafted HTML, 4kb pages that load in 12ms, documented APIs with exactly three endpoints that 1 or 2 ppl ever visit.
The Arts and Crafts movement was this exact pattern. William Morris looked at industrial textiles and asked: what if we made beautiful things slowly?
Call urself a gardener, not a farmer. Farmers sell at scale. Gardeners grow for the love of growing. Both are valid. Just don't pretend ur beatin industrial agriculture at their own game. Ur playin a different one.
U r not a software developer/engineer anymore. U r solopreneur.
Begin by building shit that breaks ur dependency on others.
reality check
going solo is hard by design
The angry fantasy goes like this: eliminate the managers. Be ur own CEO. Sign ur own work.
Here's what u actually own now:
Sales - Nobody's knocking on ur door. Eventually u need customers
Marketing - Not just shipping code anymore. Now ur shipping the story too
Legal - Incorporation. Tax stuff. Yeah that's u now
Capital - Hosting costs. API tokens
Design - Apply ur own authentic taste consistently
The pattern: every layer u eliminated as a solo dev? Those layers did work. Valuable work. Work that doesn't generate code but generates revenue.
Hard truth tho: the best engineer who cant sell works for someone who can.
The actual play isn't eliminating middlemen. It's becoming the middleman. The indie hacker thesis. 35% of 2024 startups had solo founders. One person, one product, one signature line.
Begin by building a copycat services for yourself. Not a shiny new idea that requires validation and hope customers will love. Build the first product for yourself, be it www search, custom AI agent or anything else that u use daily.
the takeaway
What matters now
The farmers and programmers analogy holds: one operator with heavy machinery replaces a hundred shovel-wielding workers. Question isn't whether to use the machine. It's whether u know enough to notice when it's digging in the wrong direction.
The insanity isn't that the world changed. It's that we're pretending it didn't.
U will build shitload. Whether u sell and capture $? That's the game.
Just remember: in food industry neither farmers nor gardeners earn CEO level payouts.