The Quarterly Review: What Stays, What Goes, What's New
Your hex is not a tattoo. It's a living system that changes as tools improve, new options ship, and your work evolves. The quarterly review is the mechanism that keeps the hex current without turning tool evaluation into a full-time hobby. Every 90 days, you spend one focused hour examining your six tools. You make changes based on evidence. Then you close the evaluation window and go back to using the hex for another 90 days.
The cadence matters. Quarterly is frequent enough to catch meaningful changes — a tool that shipped a major update, a new MCP connector that didn't exist last quarter, a shift in your own workflow — but infrequent enough to prevent the chronic browsing and evaluating that the hex exists to stop. If you review monthly, you're spending too much time thinking about tools and not enough time using them. If you review annually, you'll miss the pace at which this space moves. Ninety days is the sweet spot.
The Review Process
The quarterly review has four steps, and the order matters. Do them in sequence, not in parallel, because each step informs the next.
Step one: review the friction log. If you've been keeping a running note of moments where your hex felt limiting — tasks that didn't fit, tools that frustrated, workflows that broke — this is where that data pays off. Go through the entries from the last 90 days. Categorize them: which tool caused the friction, how often, and how severe. Sort by frequency. The friction point that showed up twelve times in 90 days is a real problem. The one that showed up twice is noise. This step takes fifteen minutes and produces a ranked list of your hex's actual pain points — not the ones you imagine, the ones you documented.
Step two: evaluate each tool against its slot. For each of the six tools in your hex, answer three questions. Is it still the best available option for this slot? Has my usage of it changed — am I using it more or less than last quarter? Is it still well-integrated with the rest of the hex? Score each tool: keep (it's working, don't touch it), watch (it's working but something better might be emerging), or swap (it's causing friction or has been outclassed). Most quarters, four to five tools score "keep," one scores "watch," and zero or one score "swap." If more than two score "swap," something went wrong in the initial hex build, and you should revisit the elimination method before making changes.
Step three: scan the landscape. This is the only time you allow yourself to browse new tools. Not Twitter threads, not YouTube reviews — check three specific sources. The MCP connector ecosystem: what new connectors shipped in the last 90 days [VERIFY]? Anthropic's changelog: what Claude capabilities changed? And one curated source you trust for tool news — not a hype channel, but something that filters for production-readiness. Spend twenty minutes maximum on this scan. You're looking for one thing: is there a new tool or capability that addresses a friction point from step one or outclasses a tool you scored "watch" in step two? If yes, flag it for testing. If no, close the browser.
Step four: make changes. If the review produced a swap candidate — a tool in your hex that's underperforming and a specific alternative that's better — execute the swap. One tool at a time. Never swap two tools in the same review cycle, because swapping one tool changes the dynamics of the entire hex and you need to see the effect before making another change. Install the new tool, configure its MCP connector, update your system prompt, and run a quick integration test with the rest of your hex. If it passes, it's in. If it doesn't, either troubleshoot or abandon and keep the current tool for another quarter.
What "Better" Actually Means
The temptation during the quarterly review is to swap a tool because the new option is shinier. Resist this. "Better" in the context of a hex means one of three specific things, and only these three.
Better integration. The new tool has MCP support that the current one lacks, or has deeper MCP support — read-write instead of read-only, more actions available, fewer failure modes. Integration improvements are the most valuable swaps because they make the entire hex faster, not just one slot.
Better core capability. The new tool produces meaningfully better output at the task you use it for. Not marginally better — meaningfully better. If your image generation tool produces images you rate a 7 out of 10 and a new option produces consistent 9s, that's a meaningful improvement. If it produces 7.5s, that's not worth the switching cost. The bar for "better enough to swap" should be high because every swap carries a learning cost — the time it takes to develop the same fluency with the new tool that you had with the old one.
Better reliability. The current tool breaks, has downtime, degrades in quality, or gets deprecated. This is the easiest swap decision because the tool is pushing you out, not being pulled by novelty. If your automation platform has been down three times this quarter, or your LLM provider changed their pricing in a way that makes the tool uneconomical, or a tool announced end-of-life — swap without hesitation.
If a new tool doesn't satisfy one of these three criteria, it doesn't earn a swap regardless of how impressive its demo looks. Demos are not the quarterly review's concern. Performance in your actual workflow, as documented by your friction log and integration testing, is the only evidence that matters.
The "Watch" Category
Some tools won't clearly earn a "keep" or a "swap." They're working fine but something in the landscape has shifted — a competitor shipped a major update, an MCP connector that didn't exist now does, or your workflow is drifting in a direction where a different tool would be a better fit. These tools go on the watch list.
The watch list is simple: the tool's name, what you're watching for, and a one-sentence note on what would trigger a swap. "Watching: Image gen tool. Trigger: if [competitor] ships MCP support, test it for one week." That's it. You don't research the competitor further. You don't install it "just to see." You note the trigger condition and check it next quarter. The watch list prevents premature swapping while ensuring you don't miss a genuine improvement.
Most watch-listed tools stay on watch for two to three quarters before either graduating to "keep" (the threat didn't materialize) or being swapped (the trigger condition was met). The watch list is a holding pattern, not a permanent state. If a tool has been on your watch list for four quarters without a trigger event, remove it from the list. The threat was imaginary.
The Anti-Pattern: The Perpetual Optimizer
The quarterly review has a failure mode, and it looks like this: someone who treats every review as an opportunity to rebuild their hex from scratch. They swap two or three tools per quarter. They're always in the "learning the new tool" phase, never in the "deeply fluent with the current tool" phase. Their hex changes so frequently that their system prompt is perpetually outdated and their workflows never stabilize.
This person is doing tool-collecting with extra steps. They've replaced the sprawl of fourteen tools with the churn of six constantly-rotating tools, which produces the same result — shallow familiarity with everything, deep fluency with nothing. The hex is supposed to produce stability. If your quarterly review consistently results in swaps, something is wrong, and the something is usually that you're optimizing for the best possible tool rather than the best possible workflow. The best possible workflow is the one that's stable enough to compound.
The goal of the quarterly review is to change nothing. The best outcome is "all six tools score 'keep,' the friction log is short, no landscape changes warrant attention." That outcome means your hex is working. The review is a check, not an optimization session. Most quarters should end with zero changes. One change per year is normal. Two changes per year is active. More than two, and you're churning, not reviewing.
Tracking the Long Arc
Keep a record of your quarterly reviews — even just a bullet list per quarter noting what you evaluated and what you decided. Over a year, this record tells a story about your hex's evolution that's invisible at any single point in time. It shows which tools have been stable pillars (the ones you've never seriously considered swapping) and which slots are volatile (the ones where you keep finding friction). The stable tools are your hex's backbone. The volatile slots are where the market hasn't settled yet — either because the tool category is immature or because your needs in that area are still evolving.
The year-view also protects against recency bias. When you're evaluating a tool in this quarter's review, it's tempting to weight the last two weeks of experience more heavily than the first eleven weeks. The review history corrects this by showing you the tool's performance across quarters, not just the recent past. A tool that was solid for two quarters and had one bad week is a "keep." A tool that's been on your watch list for three quarters might be a "swap." The long view makes better decisions than the short one.
When the Quarterly Review Says "Your Work Changed"
Sometimes the review reveals something bigger than a tool swap: your work has fundamentally shifted, and the hex needs to be rebuilt, not adjusted. You changed careers. You added a major new responsibility. You stopped doing the thing that justified two of your six tools. When this happens, the right move is not six individual tool evaluations — it's a full hex rebuild, starting from the audit.
This is rare — maybe once every two to three years for most people. But when it happens, trying to fix the hex one swap at a time is like renovating a house by replacing one wall per quarter. The foundation has moved. Acknowledge it, rebuild the hex around your current work, and start a new 30-day trial with the new configuration. The process is the same as the first time. It's just faster because you've done it before and you know what depth with a tool actually feels like.
The Calendar Prompt
Set a recurring calendar event for your quarterly review. Not a reminder — an event, with a 90-minute block. Treat it like a meeting with yourself that doesn't get moved for other things. The people who actually do the quarterly review are the ones who have it on their calendar. The people who "plan to review quarterly" and don't calendar it end up reviewing when something breaks or when they see a shiny new tool — which is reactive, not proactive, and defeats the purpose.
The calendar event should have an agenda in the description: review friction log, evaluate each tool, scan landscape, make changes. That's the whole thing. Ninety minutes is generous — most reviews take sixty or less. But blocking ninety means you don't rush, and rushing the review is how you make emotional swap decisions instead of evidence-based ones.
This article is part of The Hex System series at CustomClanker.
Previous: The 30-Day Trial | Next: Hex Configurations by Role