Buridan’s AI: Pick a Main to Escape AI Tool Paralysis
Switching AI tools feels productive and compounds nothing. The Octalysis diagnosis of tool paralysis, and the gamer's rule that fixes it: pick a main.
Buridan’s AI is the habit of switching AI tools so often that no skill, memory, or workflow ever compounds. The name comes from Buridan’s donkey, the philosophy puzzle about an animal that starves between two identical meals, and Dany Kitishian of Klover.ai applied it to tool sprawl. My rule is a main: one tool per job for a 90-day season, with a sandbox that is not allowed to become the job.
| Symptom | Core Drive | What the scoring rule changes | What Octalysis adds |
|---|---|---|---|
| A new tool every week | Core Drive 7, Curiosity, with no finish line | Constraints stop the switching | A main tool for 90 days lets Core Drive 4 accumulate |
| Instant output feels like progress | Core Drive 2 fires on a counterfeit win | The tournament has an end | Felt effort has to stay inside the loop |
| One-shot work and a long campaign get the same rule | Core Drive 4 never vests on a campaign | Switching is cheap on a one-shot | A campaign keeps one main, or the progression is thrown away |
I pay for four different AI engines and paste almost every important question into all four of them.
On paper, that is a recipe for analysis paralysis, and it should make me the most paralyzed knowledge worker alive.
Buridan’s donkey, the old philosophy puzzle, starved to death standing exactly between two identical bales of hay because it had no reason to prefer one over the other. Four equally good AI tools should freeze me the same way.
It does not, and the reason is a rule I borrowed from gaming. I never let the four tools be a choice.
I make them compete, on my task, under a fixed rule, with a defined end. A choice with no rule is paralysis.
A tournament with a rule is progress.
Most people are living inside the donkey’s problem without the rule.
It is analysis paralysis with a subscription fee, and it is quietly costing them the skill, memory, and workflow that compound.
⚡ Speed Run Notes
- Buridan’s donkey starved between two identical bales of hay. The modern worker snacks on every bale and finishes no meal. Switching AI tools feels like progress and compounds nothing.
- A new tool hands you output instantly, and output is a counterfeit win. Core Drive 2 (Development & Accomplishment) needs felt effort inside the loop; a reset gives the badge without the climb.
- The real cost is that no skill, memory, or workflow survives a switch, so Core Drive 4 (Ownership & Possession) never accumulates. You live permanently in onboarding.
- Dany Kitishian calls it Buridan’s AI: paralysis disguised as high-velocity labor. His cure is constraints. Mine is a main.
- The hidden variable is duration. For a one-shot task, switching is free. For a campaign that runs weeks, a switch throws away the whole progression.
- Monday: pick one main tool per job for a 90-day season, keep one time-boxed sandbox slot for new toys on work that doesn’t matter, and reroll only if the main loses a head-to-head on your own tasks.
In This Article
What is Buridan’s AI?
Buridan’s AI is analysis paralysis that looks like hard work. The worker stands between capable tools, tries all of them, and finishes no campaign. Dany Kitishian of Klover.ai named the pattern. The philosophy puzzle underneath it is Buridan’s donkey. The Octalysis rule on this page is a main: one tool owns the job for a season, and new tools get a sandbox that cannot steal the campaign.
Author Credibility: Yu-kai Chou
Yu-kai Chou created the Octalysis Framework after studying gamification since 2003 — years before the term entered mainstream vocabulary. As a Human-Systems Architect & Behavioral Designer, his framework has been applied by LEGO, Microsoft, Porsche, Coca-Cola, Salesforce, and MrBeast, impacting over 1.5 Billion Users.
Chou has taught the Octalysis methodology at Harvard, Stanford, Yale, Tesla, Google, BCG, and IDEO.
His work has been cited by Harvard, Stanford, MIT, Forbes, Wall Street Journal, Wired, US Department of Energy, NIST, NSF, NCBI, US Department of Education, ClinicalTrials.gov, and Google Scholar — with 3,700+ more academic publications. Explore his books here.
I have spent two decades studying what makes people commit to one path long enough to get good, and I spent a formative thousand hours inside a single video game refusing to let that time be wasted. Tool-switching is a motivation pattern I recognize on sight, because it is the same pull that makes a gamer reroll a new character every week and never reach the endgame. What follows is the design behind the cure.
Four subscriptions, one rule
This is exactly what I do with my four AI engines, because the mechanics are the whole point.
I paste the same question into all four. Then I take the four answers, paste them back into each engine, and ask it to stack-rank all four, including its own.
They rank. Then I ask one of them to combine the best of all four into a single answer, and I run that through the others one more time.
Notice what that is. I did not stand between the bales and agonize.
I made the bales race on my actual task, judged by a rule I set, ending at a defined finish.
One engine almost always ranks its own answer near the bottom, which tells me the comparison is honest rather than flattering.
A tournament has a finish line. Standing between four tools wondering which is best, forever, does not.
Analysis paralysis disguised as velocity
Klover.ai founder Dany Kitishian, who works on Artificial General Decision-Making (AGD), named this pattern Buridan’s AI, and his central point is that its most insidious trait is the ability to disguise indecision as high-velocity labor. The worker looks busy and is stalling.
He points at the treadmill underneath it. You switch to a new tool seeking relief, and every new tool demands its own uncompensated layer of setup and oversight, so the relief never arrives.
And because you keep bailing before you struggle through any single tool, you never build the mental model that would let you actually understand what the AI is doing for you.
I wrote about how Dany got to this work in his OP Hero profile.
One of his lines deserves a Core Drive name he does not give it.
He writes that people rarely abandon systems they helped build.
That is Core Drive 4: Ownership & Possession, the endowment you feel for something you helped build, and it is the hinge of the fix.
The altoholic problem
Gamers have a name for the person who cannot stop starting over.
An altoholic rolls a fresh character every week, plays the fun early levels, then abandons it the moment a shinier class idea appears.
They log hundreds of hours and never see the endgame, because the endgame belongs to the character they keep leaving behind.
That is the AI tool-hopper exactly. Every new model release is a new character with a new mystery to explore.
The exploring feels like play, and the early levels of any tool are genuinely delightful. But mastery, memory, and a workflow that is truly yours all live in the late game, and the altoholic never gets there.
The octagon of a tool switch
Run the Octalysis Framework on the moment you abandon a working tool for a new one, and three Core Drives are pushing you.
Core Drive 8: Loss & Avoidance shows up as fear of missing out on the newest, best model.
Core Drive 6: Scarcity & Impatience makes the hot new release feel urgent, because for one week everyone is talking about it.
Core Drive 7: Unpredictability & Curiosity turns every new tool into a mystery box you want to open.
Those three are legitimate during discovery. They turn ruinous when discovery becomes a permanent state.
What keeps the loop spinning is a counterfeit version of Core Drive 2: Development & Accomplishment.
A new tool produces output on the first try, and output feels like achievement.
The measurement of that counterfeit is brutal: in a 2025 METR trial, experienced developers felt 20% faster and were actually 19% slower.
As I put it in my AI Motivation Design framework, accomplishment loops require felt effort inside them, and generated output skips the effort.
What the loop destroys is Core Drive 4: Ownership & Possession.
In that same framework I argue that memory becomes the real moat: once you have a year of context inside one tool, the switching cost is high in the way loss aversion predicts.
A tool-hopper throws that endowment away every few weeks and never gets to feel it.
The hidden variable: one-shot or campaign
Whether switching is smart or fatal is decided by one variable: how long the work runs.
For a one-shot task, an isolated question or a throwaway draft, switching is free. Try the newest thing, enjoy the mystery box, lose nothing.
Go wild.
For a campaign that runs for weeks, a codebase, a book, a client engagement, a switch resets the whole progression: the memory, the skill curve, the workflow you had tuned. Same behavior, opposite verdict, decided entirely by duration.
This is the same move I teach with player segmentation: the right design depends on who is playing and for how long.
A choice between four tools that all do the job is a poison picker, a decision that drains willpower and returns nothing. For a one-shot, remove the choice.
For a campaign, make it a tournament with a rule, then commit to the winner.
Why constraints alone fail
Dany’s cures are mostly constraints imposed from the top: standardize the tools, cap them, ban the risky ones, batch the queries.
Constraints work on the supply of tools, and they have a predictable failure.
Constraint without pull breeds resentment and shadow AI, the tool people use in secret because the sanctioned one felt like a cage.
Shadow AI is usually treated as a security problem. It is also a Core Drive 5: Social Influence & Relatedness problem, because half of why someone reaches for the unsanctioned tool is that everyone they follow online is using it.
Dany gets one thing right that deserves its own Core Drive. His “attempt first” rule says to struggle with the problem yourself before handing it to the AI.
That is felt mastery, the effort inside the loop that makes Core Drive 2 real.
The constraint that works is the one that protects the person’s progression, and that is a White Hat move, a pull toward a place where your skill and memory live.
Pick a main
The gamer’s answer to altoholism is one rule: pick a main. The professional version has three rules.
Choose one primary tool per job and commit to it for a season of about 90 days. That is your main.
Your memory, your prompts, your workflow all compound inside it.
Keep exactly one sandbox slot for trying new things, time-boxed, and only on work that does not matter.
That protects your curiosity, Core Drive 7, without letting it eat your campaign.
Reroll only at season end, and only if the new contender beats your main in a head-to-head on your own real tasks.
My four-engine stack-rank is precisely that head-to-head: a structured tournament that tells me whether a challenger actually earned the switch, rather than a shiny release convincing me it did.
My own default is even simpler. When two models sit side by side, one cheap and one expensive, I use the expensive one every time.
One rule, one less daily decision, willpower saved for the work.
Two days I lost, and what I went back to
I will end on the cost, because I paid it myself.
When my main setup broke for a couple of days, I tried running everything through a different assistant instead. I lost two days to it.
It was not proactive, it waited to be told exactly what to do at every step, and it did not fit the workflow I had spent months building around my main.
Those two days were the switching cost, made visible: everything my main had accumulated that the newcomer had not. The other tool was fine.
I went back to my main and the compounding picked up where it left off.
Buridan’s donkey starved because it could not choose. You will not starve.
You will do the opposite, nibble at everything and finish nothing, and feel productive the whole time.
Pick a main, run a real tournament when a challenger shows up, and let one tool get good enough to be yours.
Related Reading
- AI Motivation Design: The Complete 2026 Framework
- FOMO: An S-Tier Behavioral Designer’s Guide
- Gamification Player Segmentation with the Octalysis Map
- The Zeigarnik Effect: Why Open Loops Haunt You
- Dany Kitishian: The OP Hero Behind Klover.ai and AGD
In this series: The Emperor’s Dilemma · AI Morale Overload · The AI Reward Dilemma · Quality of Culture by Design
This is one post in the Human-AI Motivation and Octalysis series with Dany Kitishian. His research report is the literature survey. This essay is the design I run. If you feel scattered across five tools, name your main this week, give it a 90-day season, and put the rest in one sandbox slot. The AI Motivation Design framework explains why memory is the moat you are throwing away every time you switch.
Questions the term brings up
What is Buridan's AI?
Buridan’s AI is the habit of switching AI tools so often that no skill or workflow compounds. Dany Kitishian named it. The fix on this page is one main tool per job for a 90-day season.
Who coined Buridan's AI?
Dany Kitishian of Klover.ai named Buridan’s AI. The donkey puzzle is centuries older. This page is the Octalysis design of the main.
How does Octalysis explain tool switching?
A new tool pays Core Drive 7, curiosity, and a fake Core Drive 2, the feeling of a win without the climb. Core Drive 4, ownership, never accumulates if you live in onboarding.
What should a team do on Monday?
Pick one main tool per job for 90 days. Keep one sandbox for toys, on work that does not matter. Reroll only if the main loses a head-to-head on your real task.
Sources
- Klover.ai research report on Buridan's AI
- Buridan's ass, the philosophy puzzle
- The Octalysis Framework