Article 006: Social Media
URL: https://thelongrun.work/articles/006-from-keystrokes-to-supervision
Medium
Tags (5 max)
- Technology
- Future Of Work
- Software Engineering
- Leadership
- Artificial Intelligence
Subtitle
“The developer’s role becomes more supervisory. This is not a loss of agency.”
LinkedIn Posts
Post 1: Publish-day (2026-02-23)
Hook: The role flip
I opened my IDE this morning to approve, not create.
70 files changed overnight. Agent work. Refactoring, tests, documentation. I didn’t write any of it. But I spent three hours evaluating it.
Approved 90%. Rejected 10%. Explained why. Directed the next iteration.
This is what supervision looks like. Not management. Deep technical judgment. Spotting what’s wrong without having written it yourself. Knowing which abstractions fit and which don’t. Deciding what ships and what gets reworked.
The ratio flipped. Used to be 80% typing, 20% reviewing. Now it’s inverted.
Does that make me less of a developer? Or is the craft just evolving?
New article: From keystrokes to supervision [link]
What’s your ratio these days?
Visual idea: Screenshot of a large PR with 70 files changed, or a git diff showing overnight agent work
Post 2: Insight (2026-03-02)
Hook: The value of “no”
Agent generated 200 tests. 95% coverage. Looks great.
I rejected 30% of them.
Not because they were wrong. Because they were brittle. They tested implementation, not behavior. They’d break on refactoring even when behavior stayed the same.
Knowing what NOT to do requires as much skill as knowing what to do. Maybe more.
The agent can generate code. It can’t (yet) evaluate whether code should exist.
That’s supervision. That’s judgment. That’s the work.
Visual idea: Screenshot of test coverage report, or a PR comment rejecting tests with explanation
Post 3: Reflection (2026-03-16)
Hook: The progression has always existed
Since writing about supervision, I’ve been thinking: this isn’t new.
Senior developers have always moved toward supervision. You review PRs. You guide architecture. You decide what’s good enough. The supervisory role isn’t new.
What’s new: the scale and speed.
You reach that role faster. And you’re supervising output that would require a team 10x the size of what you’d manage with humans.
The agent executes at 10x speed. You evaluate constantly.
The open question I’m sitting with: how much supervision can one person effectively do?
Visual idea: Before/after comparison showing team size vs output volume, or a simple graphic showing the supervision scale shift