Week four: choosing what not to build.
This week I spent less time building and more time deciding what not to. I created a GitHub repo to organize book learnings into my Obsidian vault — a quiet win, nothing flashy. The real work was evaluation.
Tool selection feels different than I expected. There's a pull toward installing everything that looks useful: Impeccable has potential but needs a weekly check; Cognee promises shared memory but it's immature and resource-heavy right now, so it stays off the list; some workflow skills are worth prioritizing, others are futures that can wait.
I've started keeping a decision log. For each tool, I ask: does this solve a problem I have right now, or am I betting on a future problem? The answer matters. Speculative installations cost time in setup, maintenance, and distraction. Duplicates (like an AGV-specific skill when an existing engineering tool covers it) add weight without clarity.
The pattern that emerged: adopt selectively. Build only what solves a repeated problem today. Keep the stack lean. Some things are worth waiting for — Hermes MCPs, certain infrastructure updates — but I won't force them in early just to have them.
Less, better, chosen. — Robert
← All entries