--- title: "The Daily Prompt: Ask for Ten Terrible Ideas Before You Ask for a Good One" dek: "Creative block usually means you're guarding the first idea too closely. Today's prompt makes the model go wide, bad and weird, so you can end up somewhere good." author: "Jonah Kim" author_title: "Columnist, The Daily Prompt" category: "Creative AI" series: "The Daily Prompt" format: standard tags: ["Prompting", "Design", "Creators", "Writing"] published: 2026-11-07T16:13:07.284Z updated: 2026-11-07T16:13:07.284Z reading_minutes: 3 url: "http://localhost:3000/article/daily-prompt-ten-terrible-ideas" ai_disclosure: assisted ai_disclosure_label: "AI-assisted" ai_notes: "Jonah ran the prompt and its variations with two chat models; the Crumb & Co examples are model outputs, trimmed for length." members_only: false summary: - "Asking for good ideas gets you the predictable middle; asking for terrible ones pushes the model out to the edges." - "Labelling how each idea is bad turns the list into a map of creative directions." - "The key step is asking the model to find the good idea hidden inside two of the bad ones." - "Treat the rescued idea as a starting point, because your own second pass is where it gets good." publisher: "Ideabro" --- # The Daily Prompt: Ask for Ten Terrible Ideas Before You Ask for a Good One *Creative block usually means you're guarding the first idea too closely. Today's prompt makes the model go wide, bad and weird, so you can end up somewhere good.* A copywriter I know, Bea Lindqvist-Obi, keeps a jam jar by her monitor labelled TERRIBLE. Whenever a brainstorm stalls, she writes the worst idea she can think of on a scrap of paper and drops it in. "The jar is always fuller than the notebook," she says, "and the notebook is mostly stolen from the jar." It's an old trick in creative rooms. Comedy writers pitch the hackiest possible joke just to get it out of the way. Designers sketch the ugliest logo first. Being allowed to be bad loosens your grip on the first idea, which is usually the most obvious one. Today's prompt plays the same trick with a model. ## Today's prompt ``` I'm working on: [describe the project in one sentence]. Give me 10 deliberately terrible ideas for it. Make them bad in different ways: too literal, too expensive, too weird, wrong audience, wrong decade, too boring, and so on. Label each idea with the way it's bad. Then pick the two terrible ideas that hide a good idea inside them, and explain that good idea in one sentence each. ``` ## Why bad ideas work Ask a model for "good ideas" and you'll get the middle of the road: competent, familiar and forgettable. The model reaches for the most likely answer, which is usually exactly what you were already thinking. Asking for terrible ideas changes the target. To be bad in ten different ways, the model has to explore the edges, and the edges are where interesting ideas live. Labelling *how* each one is bad turns the list into a map of directions you could take. The final step does the most important job. It asks for the good idea hiding inside the bad one, which is a skill human creatives use constantly. It also takes the pressure off you. Nobody gets precious about idea number seven when it's labelled "too weird." ## Try it: a worked example I used it for a fictional bakery, Crumb & Co, that wanted a social media campaign for its new sourdough. Here are a few of the ten: > 3. **Too literal:** A 45-minute livestream of dough rising. Nothing else. > 6. **Wrong decade:** A 1950s-style public information film warning citizens about the dangers of bad bread. > 9. **Too expensive:** Hire an orchestra to play outside the bakery every time a loaf comes out of the oven. And here are the two hidden good ideas it pulled out: > - From #3: A short time-lapse of a single loaf's 24 hours, from starter to slice, posted as a calming weekly ritual. > - From #9: A small bell the bakers ring whenever a fresh batch is ready, turned into a neighbourhood signal: "the bell just rang." That second one is genuinely lovely, and it costs about $12. ## Three variations **For designers**, get bad visual directions: ``` Give me 10 terrible visual directions for [project]. Bad in different ways: clashing colours, wrong mood, unreadable type, too minimal, too busy. Then find the two with a usable idea inside. ``` **For writers**, get bad openings: ``` Write 10 terrible first lines for [story or article]. Each one should fail differently: clichéd, confusing, too slow, spoils the ending, wrong tone. Then rewrite the two most promising into first lines I'd actually use. ``` **For musicians**, get bad hooks: ``` Suggest 10 terrible song titles and one-line hooks for a song about [theme]. Bad in different ways: too sentimental, too on-the-nose, wrong genre. Then tell me which two could become a real chorus, and why. ``` ## The rule of the second pass This is the part that matters most: don't treat the model's "good idea" as the answer. Treat it as a starting point and do a second pass yourself. In my experience, the model's rescued idea is usually a 6 out of 10, and your own second pass, prompted by it, is often an 8. The model's job here isn't to have the idea. It's to get you past the blank page quickly enough that you can have it yourself. --- Source: [The Daily Prompt: Ask for Ten Terrible Ideas Before You Ask for a Good One](http://localhost:3000/article/daily-prompt-ten-terrible-ideas) — Ideabro. AI-assisted: Written by a person with AI help for research, outlining or editing.