If Coding Is Solved, Now What?
Agents write the code now. Writing code is solved; building software isn't. What's left of the engineer's job, and why the shift is worth making now.
~/blog/tag/ai
7 posts in this process.
Agents write the code now. Writing code is solved; building software isn't. What's left of the engineer's job, and why the shift is worth making now.
Most engineers start by vibe coding, hit a wall, and decide agents don't work. A field guide to your first week instead: the setup, the right first project, and one clear step a day until you've built something real.
Four deliverables, two gates, and a loop. The process I use to build software with agents, from a plan written in plain English to docs the next agent can trust.
I don't read the code my agents write on my projects. Here's why that isn't reckless, where the approach has limits, and how to spend your attention by risk.
The tools are the easy part. Bringing a team through the shift to agentic engineering: start with a pilot, add guardrails as trust grows, measure the process instead of the typing, and respect what this asks of people.
The typing job is shrinking and the deciding job is holding. What to learn, what to let go of, and what it means whether you're starting out or twenty years in.
The good news: the world wants more software than ever, and building it has never been cheaper to try. Two projects that show how far agents can take you, and a challenge to solve your own problems.