AIYING | PO × AI | Episode #001
What If AI Had Square Wheels?
Series:
PO × AI — Think the Unthinkable
Method:
The Intermediate Impossible
Source inspiration:
Edward de Bono,
PO: Beyond Yes and No
Artificial intelligence is becoming increasingly good at removing effort.
It writes, explains, calculates, summarizes, designs, codes and recommends.
The smoother the experience, the better the technology appears.
But here is a question worth exploring:
This is the first case in our PO × AI series, where we use Edward de Bono's thinking methods to generate new possibilities for the age of artificial intelligence.
The Original PO Inspiration
Edward de Bono developed the word PO to provide a deliberate signal for provocative thinking.
Instead of immediately deciding whether an idea is right or wrong, we temporarily suspend judgment and explore where the idea might lead.
In PO: Beyond Yes and No, de Bono introduces the Intermediate Impossible as a way of using an apparently impossible or impractical idea as a stepping-stone.
The unconventional wheelbarrow illustration provides our starting point. A familiar arrangement is altered, inviting the thinker to investigate what changes when an accepted configuration is disturbed.
The wheelbarrow illustration and Intermediate Impossible method belong to de Bono's work.
"AI should have square wheels," the Thinking Clutch and the AI experiments in this article are contemporary AIYING adaptations.
Our objective is not to claim that square wheels are better than round wheels. It is to use an unusual arrangement to discover possibilities that ordinary thinking may overlook.
1. FOCUS — The Disappearing Thinking Effort
A student and an impressive presentation
Imagine Maya, a student preparing a presentation about climate change.
She asks AI to generate the outline, organize arguments, write the slides and anticipate audience questions.
Within minutes, she has a polished presentation.
During the presentation, a classmate asks: "Why did you choose that argument?"
Maya struggles to explain.
The presentation was completed, but her ability to defend its reasoning may not have developed to the same extent.
The hidden assumption
We often assume: The less effort required from the user, the better the AI experience.
This is frequently true for routine work. But is it always true for learning, judgment, creativity and capability development?
Our focus is therefore:
How might AI accelerate work without eliminating the thinking that people need to develop?
2. PROVOKE — Give AI Square Wheels
Imagine an AI that does not immediately provide the complete answer.
Instead, it might:
- Ask the learner to attempt the problem first.
- Request a prediction before showing the result.
- Reveal a hint rather than a full solution.
- Ask the user to defend a choice.
- Introduce an error for the user to detect.
- Ask for an independent explanation at the end.
At first glance, this seems to make AI worse. It adds steps to a process that technology could complete instantly.
But PO asks us not to reject the provocation immediately.
We must first move from it.
3. MOVE — From Obstruction to Productive Friction
This is the decisive stage. Without movement, the provocation remains merely a strange statement.
Movement 1 — Change the position of effort
In a conventional AI interaction, the system may do most of the reasoning before the human engages with the result.
What if the human makes an initial attempt and the AI then provides guidance?
The work is not eliminated. It is redistributed.
Movement 2 — Examine the interruption
A square wheel interrupts smooth movement.
In a learning interaction, a well-designed interruption could make the learner pause, predict, explain or reconsider.
The interruption becomes an opportunity for deliberate thinking.
Movement 3 — Extract the principle
We should not confuse the two.
Unnecessary delays, confusing interfaces and arbitrary obstacles are generally undesirable.
But a carefully chosen challenge can serve a learning objective.
Movement 4 — Find a workable arrangement
We do not need to make AI inconvenient for everyone, all the time.
Instead, we can give the user control over the amount and kind of thinking effort required.
This leads to our practical discovery.
4. DISCOVER — The AI Thinking Clutch
Imagine an AI assistant with two operating modes.
⚡ Smooth Mode
Purpose: Complete routine tasks efficiently.
AI provides direct answers, drafts, summaries and solutions, with appropriate checks.
Measure: Quality, speed and task completion.
🧠Square-Wheel Mode
Purpose: Develop human capability.
AI asks for attempts, predictions, explanations and corrections before revealing complete solutions.
Measure: Independent understanding and skill development.
Why call it a Thinking Clutch?
A clutch changes how power is transferred.
The proposed Thinking Clutch changes how cognitive work is shared between the human and AI.
Sometimes the AI performs the task. Sometimes it helps the person develop the ability to perform the task.
An AI system should be able to optimize not only for the task completed, but also for the human capability developed.
Where could it be useful?
- Students: Guided problem-solving instead of immediate homework answers.
- Professionals: Decision preparation that asks users to identify assumptions and trade-offs.
- Programmers: Debugging hints before complete fixes.
- Writers: Argument development before full draft generation.
- Everyday AI users: Learning new skills while using AI for practical tasks.
5. JUDGE — When Is Friction Useful?
PO generates possibilities. Judgment determines whether, where and how they should be used.
| Criterion | Assessment |
|---|---|
| Originality | Reverses the assumption that frictionless AI is always best. |
| Feasibility | Can be prototyped through ordinary AI prompting. |
| Potential value | May encourage independent explanation, reasoning and practice. |
| Evidence | The specific design needs testing; benefits should not be assumed. |
| Main risk | Unnecessary friction can frustrate users or obstruct important work. |
| Essential safeguard | User control, accessibility, context sensitivity and immediate override when needed. |
| Success measure | Improved independent performance, not merely increased time spent. |
Where should we avoid it?
Deliberate friction should not delay emergency assistance, urgent safety information, accessibility support or other time-critical needs.
It is primarily a proposal for learning, practice and reflective work, not a universal rule for AI design.
6. ACCOMPLISH — The Five-Minute AI Experiment
You can test the Thinking Clutch using an ordinary AI assistant.
Experiment A — Smooth Mode
Choose a concept you want to understand. Ask AI to explain it completely.
Experiment B — Square-Wheel Mode
Choose a different concept of similar difficulty.
This time, ask AI to make you contribute to the reasoning.
Compare the results
After each task, close the AI conversation. Without looking at the answer, try to:
- Explain the concept in your own words.
- Apply it to a new example.
- Identify one possible misunderstanding.
| Measure | Smooth | Square-Wheel |
|---|---|---|
| Independent explanation (0–5) | ___ | ___ |
| New-example application (0–5) | ___ | ___ |
| Error identification (0–5) | ___ | ___ |
| Time taken | ___ | ___ |
| Perceived effort (1–5) | ___ | ___ |
This is a small exploratory exercise, not a controlled scientific study. Different task difficulty and prior knowledge can affect the results.
A stronger test would use multiple matched tasks, alternate the mode order and assess performance again after a delay.
Three Further Innovations
The same provocation can generate several distinct possibilities.
Innovation 1 — The AI Difficulty Dial
Let users choose how much independent thinking they want to perform.
- Level 1: Direct answer.
- Level 2: Answer with explanation.
- Level 3: Hints before answer.
- Level 4: Guided problem-solving.
- Level 5: Independent attempt with feedback.
The dial could be adjusted to match the learner's goal and ability.
Innovation 2 — Understanding Checkpoints
Before a user accepts an important AI-generated recommendation, the system could offer a short understanding check.
For example:
"Would you like to review the assumptions and trade-offs behind this recommendation before deciding?"
The purpose is not to create an obstacle, but to make understanding available when valuable.
Innovation 3 — The AI Capability Dividend
Today, AI value is often discussed in terms of time saved, output produced or cost reduced.
What if we also measured what people become able to do independently after using AI?
Possible measures include: independent task completion, improved explanations, skill retention and transfer to unfamiliar problems.
This is a proposed measurement concept, not an established empirical result.
The Eight-Second PO Short
The companion YouTube Short communicates the central provocation through a cinematic visual transformation.
Format: 8 seconds, vertical 9:16, energetic instrumental music, no narration.
| Time | Visual | On-screen text |
|---|---|---|
| 0–2s | A luminous circular wheel spins smoothly and transforms into an instant AI answer. | AI MAKES IT EASY. |
| 2–4s | The wheel morphs into a square. The mechanism jolts. | PO: GIVE AI SQUARE WHEELS. |
| 4–6s | Each jolt becomes a spark in a glowing human brain. | FRICTION CAN SPARK THINKING. |
| 6–8s | The square wheel transforms into a luminous lightbulb. | WHAT SHOULD AI MAKE YOU THINK? |
Sound: Smooth electronic rhythm, synchronized percussive interruption, uplifting resolution.
Branding: AIYING — Accomplishing with AI.
The Short is currently represented by the storyboard above. Its YouTube link can be embedded here after publication.
Reflection — What Should AI Make You Think?
We began with a deliberately strange proposition:
We moved from an unconventional mechanical arrangement to a question about how effort is distributed between people and AI.
We discovered a possible design: an AI assistant that can switch between task completion and capability development.
We then proposed a simple experiment to investigate whether deliberate thinking friction is useful.
The larger possibility
AI need not be judged solely by how much effort it removes.
In the right contexts, we might also judge it by the abilities it helps people build.
Your PO challenge
Choose one activity where AI currently makes your work effortless.
Now ask: What useful human capability might be disappearing from that effortless process?
Design one small interruption that could help preserve or develop that capability.
Share your experiment or an unexpected interpretation in the comments.
The objective is not merely to consume another idea. It is to contribute a possibility.
Source and Methodological Note
Original source: Edward de Bono, PO: Beyond Yes and No.
Methodological foundation: The Intermediate Impossible, a deliberate provocation used to generate movement toward new possibilities.
Illustration: The unconventional wheelbarrow, provisionally catalogued in our source inventory on printed pages 90–94.
AIYING original contributions: The square-wheel AI provocation, productive-friction interpretation, Thinking Clutch, Difficulty Dial, Understanding Checkpoints, Capability Dividend and practical AI experiment.
Editorial qualification: The original wheelbarrow illustration is paraphrased, not quoted. Its precise wording and printed page range should be checked against the source scan before treating this as a fully verified bibliographic reference.
