Software Engineers Don't Have a Time Problem. We Have a Context-Switching Problem.
On any given day I'm writing TypeScript, reviewing PRs, debugging Cloud Run deployments, drafting docs, and maintaining a knowledge base. Each switch has a cost. Each cost compounds.
AI tooling doesn't replace my thinking. It compresses the low-leverage parts so I can focus on what actually matters.
Where AI Fits in My Workflow
Knowledge management. I run my second brain on Obsidian with a PARA structure. Claude acts as co-maintainer. I dump messy thoughts in, and it returns structured notes with proper links, metadata, and a filing suggestion. What used to take 20-30 minutes per note now takes under five.
Code scaffolding. NestJS modules, Docker configs, GitHub Actions workflows, PostgreSQL migrations, test boilerplate. AI writes the skeleton. I write the logic. I review everything.
Debugging. When I hit a wall, especially across Cloud Run to Cloud SQL to external API chains, I feed Claude the error log, relevant code, and what I've tried. It surfaces hypotheses I missed and suggests targeted diagnostics instead of shotgun debugging.
Documentation. API docs, runbooks, meeting summaries, blog drafts. Docs that used to take an hour take 15 minutes. More importantly, they actually get written.
Learning. I'm ramping on DevSecOps and Machine Learning alongside my core work. AI compresses the loop between "I don't understand this" and "I have a working mental model."
Task breakdown. I describe a feature and get back a checklist with subtasks, dependencies, and edge cases. Planning stays thorough without eating into build time.
What Doesn't Work
Architectural decisions. AI lists tradeoffs. The judgment call is still yours.
Missing context. It doesn't know your team dynamics, org risk tolerance, or why that service is built the way it is.
Blind trust. Every suggestion needs review. Stop reviewing, start accumulating debt.
Over-prompting. If you spend more time prompting than thinking, you've inverted the value.
The Shift
The real gain isn't "I do things faster." It's "I spend time on the right things."
Before AI tools, a chunk of my day went to formatting, scaffolding, structuring, and context-rebuilding. That's overhead, not engineering. Now that overhead is compressed, and the time goes to design decisions, hard debugging, and shipping features.
That's the productivity gain worth talking about.
Drafted with Claude, edited by me, filed in Obsidian where it'll be findable six months from now.