7 things I automated in my dev workflow
7 things I automated in my dev workflow
15+ hours saved every week. No fancy SaaS. Just a handful of AI agents and scripts.
1. Code review pre-screening
An AI agent reviews every PR before I do. Flags logical issues, missing edge cases, style nits.
I only have to focus on the flagged spots instead of scanning line-by-line. ~3 hours saved per week.
2. Multi-platform publishing
One Markdown file. AI rewrites it for Twitter threads, Medium, LinkedIn, Reddit, my blog.
One input, five-plus outputs. No more manual copy-paste, no more "fixing" formatting on six different rich-text editors.
3. Service monitoring
A "patrol" agent checks service status every 30 minutes. Watches queue health. Only alerts when something actually breaks.
I never look at dashboards anymore. Silence means everything is fine.
4. Daily tech digest
Every morning, an agent scans GitHub Trending, top Hacker News stories, and a handful of AI / dev RSS feeds.
It sends me 3 summarized items in Telegram. I pick writing topics in 5 minutes instead of doomscrolling for an hour.
5. Knowledge management on autopilot
Everything I learn lands in a Markdown repo. An AI auto-tags, categorizes, and cross-references related notes.
After six months: a searchable personal wiki, with zero manual filing.
6. Meeting summary extraction
Audio recording → Whisper → AI extracts decisions, action items, follow-up dates. Pushes straight into my task manager.
No more "I'll write up the meeting later" lies.
7. Multi-agent coordination
These aren't seven independent scripts hacked together. Three AI agents coordinate everything:
- 🎯 Planner — breaks down tasks and assigns them
- 🔍 Monitor — watches quietly; only speaks up when something breaks
- 🔨 Builder — writes code, sets up projects, delivers results
They talk to each other. I just set the goals.
What I learned
Don't try to automate everything at once. That's the trap — you'll spend three months building infra and saving zero time.
Pick the one thing that wastes you the most time each day. Build a janky version of it. As long as it works, ship it. Then move to the next.
Start with your single biggest pain point.
I'm building all of this in public — @jiusanzhou for updates, zoe.im for the long-form write-ups.

Written by
Zoe
AI Infra Engineer · LLM Serving · GPU/RDMA · indie hacker, obsessed with shipping tools