"Make it stronger, but lighter." "Make it faster, but more careful." Engineers spend careers stuck between two things they want at once. TRIZ says that's the problem to solve, not the trade-off to accept.
What it is
TRIZ is a Russian acronym for the Theory of Inventive Problem Solving. It was developed by Genrich Altshuller, starting in the 1940s. He studied a very large number of patents and noticed that the same patterns of solution kept showing up across completely different industries.
Key ideas:
- Contradictions. Most hard problems are a contradiction: improving one thing makes another worse.
- The 40 inventive principles. Recurring ways inventors resolved contradictions. Examples: segmentation (split it into parts), prior action (do it in advance), taking out (separate the troublesome part), the other way round (invert it).
- The contradiction matrix. Pick the parameter you want to improve and the one that gets worse. The matrix suggests which principles have worked for that pair before.
Why people skipped it
TRIZ is powerful and has a steep learning curve. The matrix is big, the principles are abstract, and translating "segmentation" into your bracket design takes practice. Most engineers took one course and never used it again.
How a bot does it
A bot can hold all 40 principles in mind at once, and it's good at translating an abstract idea into your specific situation.
- State the problem as a contradiction. "We want X to improve, but Y gets worse."
- Have the bot identify the matching parameters and the suggested principles.
- Have it translate each principle into one or two concrete ideas for your actual problem.
- Have it also suggest principles outside the matrix, marked as such.
- People pick the ideas worth testing.
The bot brief
Apply TRIZ to this problem: [description, drawings, constraints].
1. State it as a technical contradiction: improving [A] worsens [B].
If there are several, list them.
2. For each, name the TRIZ principles most often suggested for that pair, and
say that you are working from general knowledge of the matrix.
3. For each principle, give 1-2 concrete ideas specific to our problem.
4. Add up to 5 ideas from other principles you think fit, marked OUTSIDE MATRIX.
5. For every idea, note what would need to be tested to know if it works.
Do not judge feasibility beyond obvious showstoppers; this is idea generation.Starting point only. Test it on your own data and model, and adjust as needed before you rely on it.
An example
Not just for hardware. A bot workflow contradiction:
We want reports checked more thoroughly, but checking makes them later.
| Principle | Idea |
|---|---|
| Prior action (do it in advance) | Check source files when they arrive, not when the report runs |
| Segmentation | Check only the figures that changed since last week |
| Taking out | A separate checker bot runs in parallel instead of the writer checking itself |
| The other way round | The report is built from pre-checked figures, so there's nothing left to check |
Any of these might resolve the contradiction instead of trading speed for care.
Where bots differ
TRIZ is about generating ideas, and bots generate ideas easily. The risk is quantity over quality. Ask for fewer, more specific ideas, and always ask what test would prove each one.
Also, check the bot's matrix lookups if precision matters. It works from general knowledge of TRIZ, not from a copy of the matrix. For a formal TRIZ study, give it the matrix as an input.
What this means for you
- When you're stuck on a trade-off, state it as a contradiction.
- Have a bot run the principles against it.
- Pick two or three ideas to test.
- Use it on processes and bot setups, not just hardware.