The AI-Era Playbook

01AI writes the code. You still own it.

Every line that ships is yours, whoever or whatever typed it. That makes technical depth more valuable, not less: the job shifted from writing code to reading it fast and catching what the model got confidently wrong. A team that ships unreviewed AI output is a team that finds out in production.

02The language wars are over. AI ended them.

The old case for a slower, familiar stack was developer velocity. AI erased that tax: the boilerplate, the learning curve, the "we can't hire for it" fear. So pick the runtime the problem actually deserves. This site is one Rust binary for exactly that reason. And hire people who are curious across stacks, because the tool they knew last year matters less than how fast they absorb the next one.

03Anyone can build now. Focus is the moat.

When a working prototype costs a weekend, the product decision is no longer "can we build it" but "which one thing do we refuse to dilute". A small team with a sharply focused product beats a bigger team with a sprawling one, and AI just made small teams enormous. Product judgment is the scarce skill of this bubble.

04Ask customers why, never what.

Always talk to customers. But your job is not to collect feature requests; a customer's "what" is a guess wrapped in their vocabulary. The "why" underneath it is data. When someone asks for an export button, the interesting question is what they do with the file at 9pm on a Thursday.

05Pop the why stack.

Ask why five times. The first answer is a symptom, the second is a workaround, and somewhere around the fifth you hit the need your product actually has to serve. Stop early and you build the workaround. Every strong product I have worked on served the bottom of the stack, not the top.

06Users are the hard part now.

The noise floor rises every day, fast and loud, because everyone else got the same building superpowers you did. Distribution is the constraint. If your plan for getting users is "launch and see", the plan is that nobody finds out you exist. Decide how people will discover you before you write the first line.

07Build fans, not installs.

Installs churn. Community compounds. Followers, fans, people who wear the shirt and defend you in comment sections: that is infrastructure, as real as your database, and it can't be copied by the next AI-built clone. It gets built the slow way, by showing up and being worth following.

08Serve a fundamental need. Wear a real personality.

When features are cloneable in a week, only two differentiators are left: a product that serves the underlying need genuinely well, and a personality people recognize and like. Need plus brand. Everything else about your product can and will be copied.

09Marketing is storytelling now.

Memes and viral hits get eyeballs, and eyeballs don't convert. What converts is story: entertain people or teach them something, and do it in your own voice. The ad reads as an ad and gets skipped. The story gets finished, remembered, and forwarded.

10Brand is a bank account.

Every time you entertain or educate, you make a deposit. Every time you disappoint, or talk mostly about yourself and your product, you make a withdrawal. Most companies discover the balance only when they need to ask for something, and find the account empty. Deposit for months before you withdraw.

11Regret the things you did.

It's better to regret something you have done than something you haven't.

Risk-taking is the true engine of innovation. Betting on a new technology early, shipping bold and opinionated branding, publishing the thing that might flop: the upside of those bets is far larger than the comfort of playing it safe. Safe is also the most crowded strategy in the market.