The Monthly Stack Review Ritual

The stack you built will drift, on schedule

Here's the uncomfortable truth about everything you've built across this track so far: the clean three-role stack from lesson 777, the portable notes layer from lesson 778, the one working automation from lesson 779, the searchable prompt library from lesson 780 -- none of it stays clean by default. Sprawl wasn't caused by one bad decision back in lesson 775. It was caused by small, individually reasonable additions, accumulating slowly, with nobody ever stepping back to look at the whole picture. That mechanism doesn't stop existing just because you fixed it once. A new trial subscription, a colleague's recommendation, a "just try this for a week" that never gets revisited -- the same forces that built the original sprawl are still active every single month.

The fix isn't vigilance. Vigilance doesn't scale across eighteen months of small decisions -- good intentions never do. The fix is a recurring ritual, short enough that you'll actually run it, that catches drift before it becomes a second round of sprawl.

Four checkpoints, one recurring hour

Audit is a re-run of lesson 775's method -- the same one-week usage log format, now familiar enough to take about ten minutes instead of a full week of careful attention, because you already know what you're looking for. Consolidate checks specifically for new overlap: did a tool creep in over the past month that covers a role your stack already had filled? Apply the one-gate rule from lesson 777 the same way you did the first time. Prune means actually cancelling what's flagged idle or overlapping -- and doing it before the next billing date, not after, since an idle subscription costs real money regardless of how small the monthly charge looks. Document means updating your stack write-up with one paragraph: what changed this month, and why. That document isn't a separate task -- it's the same living record that becomes your capstone deliverable in the next lesson.

Notice what's absent from this list: no step asks you to evaluate whether a completely new category of tool might be worth exploring "just in case." That's deliberate. The review's job is maintaining what you have, not expanding it on a schedule.

Every tool gets exactly one of four labels

The mechanism that makes "consolidate" concrete, rather than a vague gut check, is a four-way decision framework applied to every tool in your current stack.

Keep requires real weekly use on a role nothing else covers -- the bar doesn't lower just because a tool has been in your stack for a while; longevity isn't evidence. Cut applies to anything idle for thirty or more days, or anything that's come to overlap a role another kept tool already fills -- this is usually the fastest, clearest decision in the whole review. Add is reserved for a role that's genuinely uncovered, not for "this looks interesting" -- the same discipline from lesson 777 applies here every single month, not just at initial setup. And Trial is the honest middle ground: a candidate for Add that hasn't yet proven itself, capped at thirty days, after which it has to convert to Keep (having earned it with real evidence) or Cut (having not).

The Trial label exists specifically to stop hype from becoming a permanent subscription. A tool that looks genuinely promising still has to clear the same bar every other kept tool cleared -- weekly use, on a role nothing else covers -- before it gets to stay past its trial window.

Set the recurring date now

The single highest-leverage action from this lesson isn't running the review once -- it's putting a recurring monthly reminder on your calendar right now, before you move to the capstone. A review you mean to run "sometime" doesn't run. A review with a fixed date, treated with the same seriousness as any other recurring commitment, is what actually keeps the stack from drifting back to where you started.

It's worth being honest about what happens if you skip a month. Nothing dramatic -- no tool breaks, no cost spikes overnight. What actually happens is quieter and more dangerous: one skipped review becomes two, a new trial subscription goes unexamined past its natural TRIAL window and drifts into being treated as a permanent KEEP by default, and six months later you're running the full lesson-775 audit again from scratch instead of the ten-minute version this ritual was designed to be. The cost of skipping isn't paid immediately. It's paid later, in exactly the sprawl this whole track exists to prevent.