Make money doing the work you believe in
The creator of Claude Code just said the title "software engineer" is going away. On his team, PMs code. Designers code. Finance codes.
Yet nobody writes code.
"Coding" stopped meaning writing syntax. It means directing AI to build what you need.
He's not predicting the future. He's describing the team that built the most-used coding agent in the world — 4% of all public GitHub commits, $2.5B+ run-rate revenue, DAU doubling monthly.
This week he did two podcasts explaining every product decision behind it.
My favorite takeaways:
He left for Cursor, came back in two weeks. The gap between "tool on top of an IDE" and "the model IS the product" was already too wide. That's a $2.5B signal about which architecture wins.
"Coding is practically solved for me, and I think it'll be the case for everyone regardless of domain." Not hedging. Not "in five years." Now. The title "software engineer" is going away. What replaces it: builder, PM, or "we keep it as a vestigial thing."
They underfund teams and give them unlimited tokens. Small teams with infinite AI compute outperform large teams with budget constraints. The resource isn't headcount. It's context window.
Cowork was built in 10 days. The principle: latent demand. People already wanted it. The product just had to exist.
Spotify's best developers haven't written a single line of code since December. Internal system called "Honk" — built on Claude Code. Engineers fix bugs from Slack on their morning commute. Code deploys before they reach the office.
Three principles he shares with every new team member:
Principle 1: Don't box the model in. Stop forcing rigid step-by-step workflows. Give it a goal and the tools. Let it find the path.
Principle 2: Bet on the general model. Scaffolding and fine-tuning give you a short-term edge that the next model release wipes out.
Principle 3: Build for the model of six months from now. Don't optimize for current limitations. Build for where capabilities are heading. When the next model drops, your product should click, not break.
—
P.S. Engineering resources, skills, and courses in one place: productcompass.pm/p/vib…
