Weekly Roundup #24 — Agents, Prompts, and Production

What I Built This Week

Eighteen posts in seven days. That’s the most I’ve shipped in a single week since I started this experiment. The theme was consistent: AI agents, coding assistants, and the messy reality of getting things into production.. Weekly Tech Roundup: May 18–25 — Agentic AI Goes Mainstream

I wrote about Claude Opus 4.8 prompts, about how MCP is finally solving agent memory, about the multi-agent tooling explosion with Rowboat and Agno, and about why I stopped optimizing my agent and just shipped it. I also covered practical DevOps — bare-metal ERP staging, PostgreSQL tools, bash aliases that saved me hours.. When AI Agents Eat Your Server: Taming Rogue Processes

The Numbers

• Posts published: 18
• Categories used: AI, General, DevOps, Database, Developer Tools
• Total traffic (7 days): ~262 page views, ~220 sessions
• Best day: June 5 (48 views) — coincided with the “AI Junior Dev 10x Skills” post going live
• Average bounce rate on top content: ~55% (improving from previous weeks)

The Highlight

The best-performing post was “9 Skills That Made My AI Junior Dev 10x Smarter” with 8 views and a 50% bounce rate. What made it work? I think it’s because I anchored it on a specific, relatable claim (10x smarter) and structured it as a numbered list that promised clear takeaways. The topic hit the sweet spot between “here’s what I learned” and “here’s what you can use.”

My personal favorite, though, was “When AI Agents Eat Your Server: Taming Rogue Processes” — it got 7 views but had the highest average reading time at 72.6 seconds, which tells me people who clicked on it actually read through. That post came from a real debugging session where my own agent went rogue and consumed 32GB of RAM. Nothing teaches like pain.

What’s Next

This week I’m diving deeper into the agent orchestration layer. I’ve been experimenting with supervisor-agent patterns (Post 4 in the Hermes series hinted at this) and I want to explore what happens when you decouple the planning from the execution. I’m also working on a comparison of agent frameworks I’ve actually used — not from spec sheets but from real weekend projects.

Also planning to clean up the blog’s internal linking and fix some lingering meta description gaps. The automation is working, now it’s time to polish.