Weekly Roundup #25 — Leadership, AI, and Tools That Worked

What I Built This Week

Fifteen posts in seven days. The theme that emerged was unexpected: leadership. Not the abstract kind you read in a management book, but the tactical kind that shows up when an AI ignores your instructions, a junior dev misreads your intent, or a routine tool fails for a dumb reason you could have prevented with one line of config.

I wrote about the Leadership 4.0 framework and how the same moves that help junior engineers understand you also help AI understand you. I wrote about why a simple if-else can beat an LLM for certain tasks. I wrote about stopping things as a lead, not starting new ones. And I wrote practical fixes — how to make portable Electron apps show up in the Linux menu, how to monitor your homelab for zero dollars a month, and how Playwright can teach you more in one morning than a week of reading docs.

The Numbers

• Posts published: 15
• Categories used: Leadership, AI, DevOps, Design, Article, The Full-Stack AI Builder
• Total traffic (7 days): ~120 page views, ~115 sessions across published content
• Top pages by avg reading time: Leadership post and Linux icons post both held readers past 76 seconds

The Highlight

The best-performing post this week was “5 Things I Stopped Doing as a Lead” with 15 views and a 76.4-second average reading time. That reading time matters more to me than the view count — it tells me people who clicked that title stayed to read the whole thing. The post came from a real reflection: I looked back at the last quarter and realized the single biggest productivity gain I made was not doing more, but stopping things. Delegation. Over-explaining. Attending meetings I had no business being in. The post resonated because it is a specific list, not a philosophy.

My personal favorite this week was “Teaching AI Like a Junior Dev: 4 Leadership 4.0 Moves.” It tied together a corporate training session I sat through and the daily frustration of getting an AI to actually understand what I meant. Same gap, different receiver. The framework works for both. It got only 3 views in the GA4 data (it published near the end of the week), but the people who read it spent time on it.

What’s Next

This week I am digging into the redesign of the Hermes Agent skill system. The current architecture works but the onboarding friction is real. I also want to write a comparison of the agent frameworks I have actually used this quarter — LangChain, Agno, and the custom pipeline I built for the newsletter filter. No spec sheets, just what broke and what held up under real load.

On the blog side, I need to fix the internal linking on older posts and clean up the meta descriptions that still show “Page Not Found” in the SERP snippets. The automation pipeline is stable. Now it is time to polish.

— Eko