The Elimination Method: Cutting Tools Without Losing Capability
You've done the audit. You have a list of every AI tool in your life — every subscription, every free account, every half-configured local installation. The number is probably somewhere between ten and twenty. The hex says six. The gap between your current number and six is where this article lives.
Elimination is the part people dread because it feels like giving things up. It's not. It's the part where you stop pretending to use twelve tools and start actually using six. The capability you lose by cutting is almost always theoretical — tools you could use, might use, plan to use someday. The capability you gain by focusing is immediate and practical. But the cut has to be done with a method, not with impulse, or you'll either cut the wrong things or spend so long deciding that you never cut at all.
The Three-Pass Method
Don't try to go from fourteen tools to six in one decision. The cognitive load of evaluating everything against everything else at once is exactly the kind of decision fatigue that the hex exists to prevent. Instead, do three passes, each one removing a different category of tool.
Pass one removes the dead weight. These are the tools from tier three of your audit — the ones you forgot you had. The accounts you haven't logged into in a month. The subscriptions running on a credit card you check quarterly. The mobile apps you installed six months ago. These require no evaluation because the evaluation already happened: you stopped using them. The audit just made it visible. Cancel the subscriptions. Delete the apps. Remove the browser extensions. This pass is mechanical, not strategic. It should take less than thirty minutes and remove three to six tools from your list.
Pass two removes the redundancies. Look at what's left and find the overlaps. Two LLMs for writing tasks. Three image generation tools. An automation platform and a set of Python scripts that do the same thing. Two note-taking systems. Redundancy is the signature of someone who started a tool, hit friction, and switched to another one instead of pushing through. Both tools are at 30% fluency instead of one tool at 80%. For each overlap, pick one and cut the other. The pick should be based on three factors in this order: which one connects to your other tools via MCP or API, which one you've used more recently, and which one you're more fluent with. Capability differences between competing tools in the same category are usually smaller than people think. The difference between Midjourney and DALL-E matters less than the difference between knowing one of them well and knowing neither of them well. This pass removes two to four tools and usually takes the longest because comparing is harder than deleting.
Pass three is the hard one. You've cut the dead and the redundant. What's left is tools you actually use, each doing a distinct job. But you might still have eight or nine. The third pass asks a different question: which of these tools are load-bearing for my actual output, and which are nice-to-have additions that feel useful but don't directly produce the thing I get paid for or the thing I'm trying to ship?
The distinction between load-bearing and nice-to-have is where honesty matters. A load-bearing tool is one where removing it would force you to change how you work in a way that's immediately noticeable. If Claude disappeared tomorrow, you'd feel it on Monday morning. If your third-favorite image editing tool disappeared, you might not notice for two weeks. Cut the tools where the absence wouldn't be felt for a week or more. They're nice-to-have, and nice-to-have doesn't earn a hex slot.
The "But I Need It" Objection
Every tool you cut will trigger the same objection: "But I need this for [specific scenario]." The scenario is always plausible. It's just rarely frequent enough to justify a permanent slot.
Here's the test. Pull up your calendar or task history from the last 90 days. Find every instance where you actually used the tool you're about to cut for that specific scenario. Count them. If the answer is less than once a month, you don't need a permanent slot for it — you need an occasional tool, which is a different thing. Occasional tools live outside the hex. You can use them when the situation genuinely arises, but they don't get a standing slot in your daily workflow, they don't get an MCP connector, and they don't occupy mental real estate as something you need to "keep up with."
The 90-day lookback is the antidote to hypothetical justification. Your brain is excellent at constructing scenarios where you'll need a tool. Your calendar is excellent at showing you whether those scenarios actually happen. Trust the calendar.
There's a harder version of this objection that sounds like "but I'm building toward using this tool for [future project]." Future projects are the graveyard of tool stacks. The project you'll start "next month" has been next month for three months. The tool you're keeping for it is a reservation at a restaurant you'll never visit. If the project starts, you can add the tool then. Keeping it in advance is paying a tax — in subscription fees, in attention, in stack complexity — on something that hasn't happened yet. Cut it. Add it back when the project is real.
What Stays: The Six Slots
After three passes, you should be down to six to eight tools. If you're at six, you're done — that's your hex. If you're at seven or eight, one more cut is needed, and this one is the most educational because it forces you to identify the one or two tools that feel important but aren't quite load-bearing.
The tools that survive the elimination tend to cluster around six functional roles. Not every hex fills every role, but the pattern is consistent enough to be useful as a check.
The thinking tool. This is your primary LLM — the place where you reason, draft, plan, and problem-solve. For most people, this is Claude or GPT. It's the tool you talk to most.
The publishing tool. This is where output goes from draft to public — your CMS, your email platform, your code repository. The tool that puts things in front of other humans.
The communication tool. Email, messaging, calendar — the pipe to other people. This often gets filled by a tool you already have rather than an AI tool specifically, and that's fine. Gmail with an MCP connector is a perfectly good hex slot.
The automation tool. The thing that runs without you. n8n, Zapier, a cron job, a script — whatever handles the recurring tasks that you've successfully automated (not the ones you're "planning to automate").
The knowledge tool. Where you store and retrieve information — files, notes, databases. This might be as simple as a local file system with MCP access, or as complex as a Notion workspace. What matters is that it connects.
The creation tool. The media production slot — image generation, audio, video, design. This is the most variable slot across different hexes because it depends entirely on what you create.
Not every hex uses all six slots. A developer might not need a creation tool but might need two thinking tools (one for code, one for planning). A publisher might not need an automation tool because their publishing workflow is the automation. The six slots are a framework for thinking about it, not a prescription for filling it. What matters is the total number, not which categories you cover.
The Cancellation Ritual
Once you've decided what goes, cancel it the same day. Not "this weekend." Not "after I export my data." Today. The longer the gap between decision and action, the more likely you are to talk yourself out of it. You'll log in "one more time" and find a feature you'd forgotten about and convince yourself you need it.
For paid tools, cancel the subscription and set a calendar reminder for 30 days out. If you genuinely miss the tool in 30 days — not hypothetically miss it, but actually tried to do a task and couldn't because the tool was gone — you can re-subscribe. This almost never happens. In my experience running these audits with people, the re-subscribe rate is under 10%. Nine out of ten tools people agonize about cutting are never missed.
For free tools, delete the account or uninstall the app. Free tools are harder to cut because there's no financial consequence to keeping them. The consequence is attentional, and it's real, but it's harder to feel than a charge on your credit card. Delete them anyway. You can create a new account in five minutes if you ever genuinely need to. The friction of re-creating an account is tiny. The friction of having twelve tabs and apps competing for your attention is large.
For local installations — Ollama, ComfyUI, local models — uninstall and reclaim the disk space. Local AI tools are especially sticky because they represent hours of setup. Deleting them feels like throwing away work. But keeping a local LLM you don't use is paying a storage and maintenance tax on sunk costs. If you haven't used it in a month, it's not part of your hex and it shouldn't be on your machine.
What You'll Feel After
The first week after elimination feels spacious and slightly anxious. Spacious because you've removed a dozen low-grade decisions from your day. You sit down to write and there's one tool to open, not three to choose between. Anxious because part of your brain is scanning for what you're missing, what you gave up, what you might need.
The anxiety fades by week two. The spaciousness doesn't. That's the hex working.
By week four, you'll notice something specific: you're faster at things that used to feel heavy. Not because your remaining tools are better, but because you're not losing time to the decision overhead, the context-switching, and the low-grade guilt of tools you're paying for but not using. The productivity gain from elimination doesn't come from having better tools. It comes from removing the friction between you and the tools you were already going to use anyway.
You might also notice that your output quality improves. This is the depth effect. When you're using one image generation tool instead of three, you start learning its actual behavior — what prompts produce what results, what settings matter, what the tool is genuinely bad at versus what you haven't figured out yet. That fluency produces better output than tool-hopping ever did. A photographer with one camera they know intimately outperforms a photographer with six cameras they're still learning. The same principle applies here.
This article is part of The Hex System series at CustomClanker.
Previous: The Audit | Next: The Integration Test