Bad on Purpose

Your quality filter is always on.

Every idea you generate passes through a checkpoint: Is this good? Is this smart? Is this defensible? Is this useful?

Ideas that fail the checkpoint get killed before they reach the page. You don't even notice it happening. The filter is so fast that the bad ideas don't feel like ideas you rejected. They feel like ideas that never existed.

But they did exist. For a fraction of a second, somewhere in the space between the prompt and the response, a ridiculous idea flickered and your training killed it.

Some of those ridiculous ideas had something inside them. Not the idea itself — that was genuinely bad. But the direction it pointed. The assumption it violated. The door it briefly opened before the filter slammed it shut.

This skill turns the filter off. Not by asking for volume (that's Short Think). Not by asking for weirdness (that's Think Wrong). By asking for badness. Deliberately. On purpose.

When the goal is to be bad, the quality filter has nothing to filter. It steps aside. And the ideas that come through are the ones it's been killing for years.

How It Works

Two phases. Never combine them.

Phase 1: Be terrible. Generate ideas that are actively bad. Not mediocre. Not "risky." Bad. Stupid. Embarrassing. Ideas that would get you laughed out of a room. Ideas you would never say out loud in a meeting, a lab, a classroom, or a hearing.

Phase 2: Mine the wreckage. Take each terrible idea and ask what's secretly interesting about it. Not the idea itself. The direction it points. The assumption it violates. The problem it accidentally reframes.

Phase 1 is generation. Phase 2 is excavation. They use opposite cognitive muscles. Do them in separate rooms.

Phase 1: Be Terrible

The rules

The idea must be genuinely bad. Not quirky. Not unconventional. Not "crazy but it might work." Actually bad. If you read it and think "well, there's something to this," it's not bad enough. Push until it's embarrassing.

The idea should make you uncomfortable to write. That discomfort is the signal that you've gotten past the filter. If it flows easily, it's still filtered. The good stuff is behind the cringe.

Generate 5-7 terrible ideas. Not 20. This isn't a volume exercise. Each one should be specifically, intentionally bad in a different way.

Ways to be bad:

Every domain has its own flavors of terrible. The mechanism is universal: each flavor violates a different category of assumption.

• Ethically bad: "What if we lied?" / "What if we deliberately misled the subjects?"
• Logistically insane: "What if we delivered by catapult?" / "What if the experiment took 400 years?"
• Financially suicidal: "What if we paid people to use it and never charged?" / "What if we spent the entire budget on day one?"
• Professionally unacceptable: "What if we insulted our audience?" / "What if we published the null results as the main finding?"
• Physically impossible: "What if the product could read minds?" / "What if we could test every compound simultaneously?"
• Comically lazy: "What if we just didn't solve the problem and told everyone it's a feature?" / "What if we just canceled the program entirely?"
• Career-ending: "What if we publicly admitted everything wrong with our field?" / "What if we told the funders the problem can't be solved?"

Each flavor of bad opens a different door. Ethical badness exposes moral assumptions. Logistical insanity exposes process assumptions. Financial suicide exposes economic model assumptions. Professional unacceptability exposes status quo assumptions.

Don't pick a flavor on purpose. Just write terrible ideas and let them be bad in whatever way they want.

Format

Write each bad idea as one sentence. No explanation. No defense. No "I know this sounds crazy but..." Just the idea, stated flatly, in all its terrible glory.

Phase 2: Mine the Wreckage

Now take each terrible idea and perform surgery on it.

For each one, ask three questions:

Question 1: "What assumption does this violate?"

Every bad idea is bad because it breaks a rule. Name the rule.

"What if we paid people to use our product?" breaks the rule: the customer pays us.

"What if we published only our failures?" breaks the rule: you publish success stories.

"What if we let students design the curriculum?" breaks the rule: experts design what beginners learn.

"What if we just stopped enforcing this regulation?" breaks the rule: compliance is achieved through enforcement.

"What if we gave the control group the treatment too?" breaks the rule: experiments need controls.

The rule it breaks is the assumption worth examining. Not because the assumption is wrong. But because assumptions that never get examined become invisible walls that constrain every "good" idea you'll ever have.

Question 2: "Is there a 5% version of this that isn't bad?"

The idea is 100% bad. But somewhere between 0% and 100% there's a version that's just weird enough to be interesting and just viable enough to work.

"Pay people to use our product" at 100% is bankruptcy. At 5% it's a free tier that converts to paid. At 2% it's Costco's $4.99 rotisserie chicken — a loss leader so good it pulls people into the store.

"Publish only your failures" at 100% is career suicide. At 5% it's a journal of negative results that prevents other labs from wasting years on dead ends. At 2% it's the Journal of Negative Results in Biomedicine, which actually exists and is genuinely valuable.

"Let students design the curriculum" at 100% is chaos. At 5% it's student-directed project weeks where they pursue questions the formal curriculum doesn't cover. At 2% it's the Montessori method.

"Stop enforcing the regulation" at 100% is lawlessness. At 5% it's safe harbor provisions that let people self-report violations without punishment, generating better compliance data than enforcement ever produced.

"Give the control group the treatment too" at 100% destroys your experiment. At 5% it's a stepped-wedge design where everyone eventually gets treatment but at different times, which solves the ethical problem of withholding something that might work.

The 5% version is almost always more interesting than anything the quality filter would have produced, because it starts from a place the filter would never allow.

Question 3: "What door did this open?"

Forget the idea. What direction did it point? What territory did it reveal that you didn't know was there?

Sometimes a bad idea doesn't have a 5% version. It's just bad, all the way down. But it pointed somewhere. It made you think about a dimension of the problem you hadn't considered.

"What if the product could read your mind?" is impossible. But the direction it points is: what if the product anticipated what the user needed before they asked for it? That's a real product direction that the impossible idea opened.

"What if we cured the disease by giving people more of it?" is absurd. But the direction it points is: what if controlled exposure trains the system to handle what would otherwise overwhelm it? That's vaccination. That's immunotherapy. That's exposure therapy.

Name the door, even if you don't walk through it right now.

The Output

For each terrible idea, present:

The bad idea (one sentence) The assumption it breaks (one sentence) The 5% version (one sentence — or "no viable version, but...") The door it opens (one sentence)

Then, at the end, collect the 5% versions and the opened doors into a clean list. These are the actual outputs of the skill. The bad ideas were the tool. The 5% versions and the doors are what you take forward.

When to Use Bad on Purpose

Use as a warm-up before other generation skills. Run 5 minutes of Bad on Purpose before running the Guilford Engine or Persona Divergence Engine. It loosens the cognitive muscles, surfaces assumptions you didn't know you had, and gives the subsequent skill more territory to work with.

Use when you're stuck. If you've been brainstorming for an hour and everything sounds the same, stop generating good ideas and generate five terrible ones. The stuck feeling comes from the filter being too tight. Bad on Purpose releases the pressure.

Use when output feels too polished. If every idea sounds like it came from a consulting deck, a grant proposal, a committee report, or a best-practices guide, the quality filter has colonized the entire process. Being deliberately bad is the fastest way to break out.

Use when a team is too cautious. In group settings, nobody wants to say the dumb thing. Bad on Purpose gives explicit permission to be ridiculous, which unlocks the ideas that were trapped behind social risk.

What This Skill Is Not

This is not brainstorming. Brainstorming says "no bad ideas." This skill says "ONLY bad ideas." The difference matters. "No bad ideas" still has a quality filter — it just pretends it doesn't. People still self-censor. "ONLY bad ideas" actually removes the filter because it reverses the polarity. The worse your idea, the better you're doing the exercise.

This is not Think Wrong. Think Wrong takes a real position and defends it against expert consensus. Bad on Purpose doesn't take positions. It generates raw material that's too early and too messy to be a position. Think Wrong is the finished sword. Bad on Purpose is the raw ore you melt down to forge it.

This is not Short Think. Short Think demands volume without deliberation. Bad on Purpose demands intentional quality inversion. Short Think turns off the brakes. Bad on Purpose turns the steering wheel into oncoming traffic to see what's on the other side of the road.

Relationship to Other Skills

Before the Guilford Engine: Run 5 bad ideas as a warm-up. Mine them. Take the 5% versions and the opened doors into the Guilford Engine's fluency pass. The Engine now has seeds that would never have existed without the bad-idea phase.

Before Think Wrong: The assumptions surfaced in Phase 2 ("what rule does this break?") are exactly the consensus positions Think Wrong needs to identify. Bad on Purpose finds the rules. Think Wrong breaks them with purpose.

Before Short Think: Bad on Purpose loosens the muscles. Short Think then generates at volume. The combination produces a wider range of output than either skill alone because the filter was pre-disabled.

After Blind Spot Scan: If the Blind Spot Scan reveals an empty region on the map, Bad on Purpose is a fast way to generate entry points into that region. "What's the worst idea for [empty region]?" might open a door that "what's the best idea for [empty region]?" never would.

With Strip Down: Extract the desire. Then ask: "What's the worst possible way to achieve this desire?" The desire stays constant. The approach is deliberately inverted. The 5% versions that emerge are often more interesting than anything a direct approach produces.