Most of the people I talk to are not short on AI tools. They are short on a reason to keep any of them. A new one launches, a friend swears by it, the free tier is right there, and it gets added to the pile next to the three tools that already do roughly the same thing. Six months later nobody on the team can say which one is the system of record, and the monthly card statement is the only complete inventory.
The rule I use, and teach, is small enough to hold in your head. Every job on your desk already has something doing it. A new tool does not get to move in alongside. It has to take the slot.
Why tools pile up
Anxiety about AI is a collecting anxiety. When the ground is moving, buying the next tool feels like doing something, and a subscription is easier to justify than an afternoon of practice. So people collect tools when the thing they actually need is a skill. Skills survive layoffs, pricing changes, and the day a product gets acquired and quietly shut down. A tool you never learned to use well survives none of that.
The One Slot Rule is a forcing function against that reflex. It does not say never try new tools. It says a trial is a contest with a defined job, a current holder, and a winner.
The four questions
Ask them in order. Each one only makes sense once the one before it has an answer, and most challengers lose on the first or second.
- What is the job? Name it in one plain sentence, the way you would explain it to a new hire. "Tell me what to fix next on the website" is a job. "Give me the best keyword intelligence available" is a wish, and a more expensive one.
- Can it do that job? Not a demo, not a feature list. Run the real job with the real inputs and look at what came back. If you cannot verify the output, it did not do the job.
- What does it touch that is hard to undo? Your calendar, your customer list, your codebase, a patient record, a payment. Read and report is one kind of tool. Write and send is another, and it needs a much stronger case.
- What would have to be true for the challenger to take the slot? Write the condition down before you decide. If the condition is not true yet, the holder keeps the slot and the challenger goes on a list with a date to re-check.
Every job gets one slot. A new tool does not get added. It has to take that slot from whatever holds it now, and if it cannot win the slot outright, it is Cut.
Two worked examples
The public version of this rule lives on Last Tool Standing, where every verdict is either Still Standing or Cut and the Cut ones get published too. Two rows show how the questions land in practice.
OpenSEO versus Semrush, for the job of telling me what to fix next on a site almost nobody has found yet. OpenSEO holds the slot. It answers the only question the site has right now, and it reads and reports without any write access to the site. Semrush was tried on the free tier and put down on price, which is a verdict about the job at this size, not about the product. The condition for a rematch is written down: enough search volume that the difference between good keyword data and excellent keyword data changes a decision.
OnSpace AI versus Lovable, for the job of getting an MVP with auth, a database, and payments from a prompt. Lovable holds the slot. OnSpace went into testing and the verdict was deliberately not called, because question two had not been answered with a real build yet. A tool in testing is not a tool in the stack.
How to run it on your own stack
- List every recurring job, one line each. Most people find eight to fifteen.
- Next to each job, write the tool that holds the slot today. If two tools hold the same slot, that row is already a contest you have been avoiding.
- For every tool with no job next to it, cancel it or give it a job by Friday.
- Keep a short list of challengers with the written condition for each. Re-check on a date, not on a mood.
The tool count goes down, the skill per tool goes up, and the next launch announcement stops feeling like homework.
First published in Artificially Designed, Issue 11, The Slot on August 24, 2026. The newsletter runs a One Slot verdict every week.