In 1975, Brian Eno and Peter Schmidt had the same problem every creative person hits at some point. Not running out of ideas. Running out of different ideas.
The same instincts kept firing. The same solutions kept coming up. So they made a card deck.
113 cards. Each one a short, cryptic instruction designed to interrupt how you’re already thinking:
- “Honor thy error as a hidden intention”
- “Faced with a choice, do both”
- “Use ‘unqualified’ people”
They called it Oblique Strategies. Musicians pulled cards when they got stuck in the studio. When a random instruction meets a specific problem, that’s where the unexpected ideas come from.
I built a Claude Code skill that does exactly this for product and project ideas.
Why I Built It
I kept hitting the same creative wall. Not “I have no ideas.” More like “I keep having the same ideas.” Every brainstorm landed in the same five places. And asking an AI to help made it worse. AI is trained to be helpful, which means it goes for the most likely answer. And the most likely answer is usually the one everyone already has.
Oblique Strategies works because it doesn’t try to be helpful. It gives you a random constraint and makes you work inside it. The strangeness is the point.
So I wrapped all 113 strategies into a skill and taught Claude how to apply them.
How It Works
You describe your topic. The skill picks a strategy at random. Not the one that fits best, because that would bring back the same patterns you’re trying to escape. Then it gives you two ideas:
- Literal: a concrete, buildable idea that follows the strategy applied directly to your topic.
- Abstract: a stranger take on the same strategy. Something you might not immediately know how to build, but that opens an angle you’d never have found by thinking normally.
You always get both. Some will feel too weird or impractical. That’s fine. The goal isn’t a good idea every time. It’s to break out of the pattern that’s been keeping you stuck.
What It Looks Like in Practice
Idea #1: “iOS app to learn handstand”
Strategy: “Mechanicalize something idiosyncratic”
Literal: Everyone finds their balance point differently — a personal micro-adjustment of the wrist, a slight forward lean unique to their body. The AI identifies your specific idiosyncratic balancing pattern and turns it into a trainable, repeatable mechanism. Your weirdness becomes your technique.
Abstract: The app studies how you fall — your personal, consistent collapse pattern — and builds a machine-like counter-sequence to reverse-engineer it. You don’t learn a generic handstand; you learn the precise antidote to your own body’s failure mode.
Strategy: “Ghost echoes”
Literal: Overlay a semi-transparent ghost of your best previous attempt on the live camera feed as you train. You chase your own past self, not a stranger’s ideal form.
Abstract: The ghost isn’t yours — it’s built from thousands of anonymous users who fail at exactly the same point you do. A composite specter of collective failure at your specific stage. You’re not chasing your past self; you’re training against the averaged ghost of everyone who got stuck where you are.
Idea #2: “iOS finance management app for Gen Z”
Strategy: “Honor thy error as a hidden intention”
Literal: Instead of flagging overspending as failures, the app reframes every budget mistake as signal — a pattern that reveals what you actually value vs. what you think you value. Errors become the most honest data in your financial profile.
Abstract: The app has no budget targets at all. You just spend. Every month the AI shows you only your “errors” — the purchases that broke your own past patterns. No judgment, no correction. Just a mirror of your inconsistencies, which turns out to be more revealing than any budget ever was.
Strategy: “Use ‘unqualified’ people”
Literal: All financial advice in the app comes from other Gen Z users who were in your exact situation 6 months ago — not certified advisors, not influencers. Peer-to-peer guidance from people close enough to your context to actually be useful.
Abstract: The app’s AI was trained exclusively on the financial decisions of 19-year-olds — not economists, not analysts, not adult finance gurus. It reasons the way a broke but resourceful 19-year-old would. The “unqualified” perspective is the product.
Idea #3: “iOS language learning app using AI”
Strategy: “Cut a vital connection”
Literal: The app removes translation entirely. No definitions in your native language, no bilingual flashcards — just the target language explaining itself through images, audio, and context. You can never look something up in your mother tongue. The crutch is cut from day one.
Abstract: The AI cuts the connection between you and your progress metrics entirely. No streak, no level, no percentage complete. You have no idea how far you are or how fast you’re moving. The only signal is a single felt sense: did you understand it or not?
Strategy: “Only one element of each kind”
Literal: The app teaches exactly one word per day. Nothing more. No lessons, no grammar drills, no streaks — just one word, delivered at the right moment, with one example sentence. Radical reduction as the entire design philosophy.
Abstract: The AI picks one person for you to learn the language from — one native speaker whose content, speech patterns, and vocabulary become your entire curriculum. You don’t learn the language; you learn how one specific person speaks it.
Idea #4: “AI video prompt manager for pro film teams”
Strategy #37: “Do something boring”
Literal: The prompt manager’s core feature is a standardized prompt template that every team member fills in the same way — shot type, subject, motion, lighting, duration, negative space. No freeform text allowed. Boring, rigid, mandatory structure. The creative work happens inside the constraints, not in how you write the prompt. The team always speaks the same language to the model.
Abstract: The tool deliberately makes prompt crafting tedious. You can’t type — you select. Every element of the prompt is chosen from dropdowns built from your production’s own past outputs. The friction is the feature: slowing down makes you deliberate. The “boring” input process forces the filmmaker to commit to exactly what they want before the model ever runs.
Strategy #44: “Faced with a choice, do both”
Literal: When a team member submits a prompt, the tool automatically forks it — sending two variations to the model simultaneously. One follows the prompt exactly as written, the other introduces a subtle directorial shift the system infers from the production’s style history. The team always reviews two outputs, never one. Choosing is the review process.
Abstract: The tool refuses to let the team converge. Every time a prompt reaches consensus — everyone has approved it — the system splits the project into two parallel timelines, each carrying a different version forward. The production exists in superposition until the director collapses it. The decision isn’t made until the last possible moment.
How to Steer a Session
Once you have ideas, you can keep going:
- “another”: new random strategy, fresh pair of ideas
- “go more abstract”: skip the literal, go weirder
- “go more literal”: skip the abstract, make it actionable
- “develop this”: turns the last idea into a short brief: the idea in one sentence, who has the problem, what makes it different, and what you’d build in a weekend to test it
- “synthesize”: after three or more rounds, finds the pattern running through all your ideas and surfaces the strongest combination you didn’t see coming
If you want to try the skill, send me a DM and I’ll share it with you directly.