Brainstorming with AI

I wanted to write a blog post and asked Claude to help me brainstorm. But where do you start? Just typing “hey Claude, help me brainstorm” felt like walking into a meeting with no agenda.

I had tried something better before. I’d tell Claude: “Ask me questions. Propose different solutions. Walk me through this.” That worked better than free-form chat. The structure helped. But I still felt like I was missing something—like there was a better way to do this that I haven’t discovered yet.

Then I found the brainstorming skill.

What It Actually Is

The brainstorming skill is a structured framework for collaborative design. It has three phases: Understanding, Exploration and Design. It uses specific tools, like structured question-asking. It’s systematic but flexible.

The key shift: you move from “chatting with AI” to “collaborating with structure.”

And I thought: what better way to test this than using it to design a blog post about itself?

The Session

Here’s what happened when I used the brainstorming skill to plan this very blog post.

Phase 1: Understanding

Claude started by doing exploration. No barrage of questions. Instead, it explored what I’d shared, formed a model, and said: “Based on what you’ve told me, here’s how I think this should work…”

Then it asked one targeted question: What’s my primary goal? Not twenty questions about audience, tone, length, platform, and publication date. One question. I answered: “Share my experience using the skill.”

Then the next question: Who’s my audience? I said: broader audience, not just developers. Another question: What’s my key insight? I said: it’s a super helpful discovery process.

No wasted motion. Each question filled a real gap. Claude gathered what it needed: purpose, constraints, success criteria.

Phase 2: Exploration

Now Claude proposed three approaches:

Before/After Story - Show a frustrating free-form chat, then the same conversation with structure
Live Commentary - Walk through the session with running commentary
Discovery Journey - Traditional blog structure: intro → explanation → example → conclusion

Each came with trade-offs. From these three above, Claude recommended choosing Live Commentary. I ended up choosing Discovery Journey. We moved on.

Notice what happened: Claude didn’t dump all three options as equal. It had a preference and explained why. I could agree or redirect.

Phase 3: Design

Then we built the structure piece by piece. Opening: start with the problem, add my earlier attempt to improve it, then introduce the systematic approach. The example: show our actual session. Closing: explain why structure helps.

At each step, Claude checked in: “Does this work?” I confirmed or adjusted.

The meta moment: you’re reading the result of that session right now.

What Makes This Approach Different

Notice what didn’t happen:

Claude didn’t ask me 42 questions upfront
It didn’t dump a complete outline without checking
We didn’t skip exploring alternatives

We built it together, piece by piece.

This is what the brainstorming skill does. It creates a systematic process for collaborative discovery. You don’t waste energy figuring out HOW to collaborate. The structure handles that. You focus on WHAT you’re building.

Why It Matters

The systematic approach is a discovery process. Having structure doesn’t constrain creativity—it creates space for it. You’re not stuck wondering “what should I ask?” or “did I cover everything?” The framework guides you. You discover what you’re actually trying to build.

And that’s the insight: structure doesn’t limit you.

This blog post used the brainstorming skill to design… a blog post. But the same process works for technical challenges:

New features - “I want to add user authentication to my app” → explore OAuth vs JWT vs sessions, database schema, security considerations
Code architecture - “This module has become a mess” → explore refactoring approaches, examine trade-offs between patterns
API design - “I need an endpoint for this” → explore REST vs GraphQL, data structures, error handling strategies
System design - “We need to scale this service” → explore caching strategies, database choices, deployment approaches

The phases stay the same: Understanding (what are we solving?), Exploration (what are our options?), Design (how should it work?). The systematic approach works whether you’re designing prose or code.

If you work with AI, try it. Use the brainstorming skill next time you need to design something. See what you discover.

How to Use It

The Easy Way: Claude Code

If you use Claude Code, install the Superpowers plugin:

Open Claude Code
Run /plugin install superpowers
Restart Claude Code
Ask Claude to brainstorm with you

The plugin lives at: https://github.com/obra/superpowers

The Manual Way: Any AI Tool

Copy the brainstorming skill prompt from https://github.com/obra/superpowers/blob/main/skills/brainstorming/SKILL.md and paste it into your AI tool. Tell the AI to follow it.

This works with ChatGPT, Claude on the web, or any other AI assistant. You lose the automated tooling, but you get the systematic framework.

Credits

The brainstorming skill and the Superpowers plugin are created and maintained by Jesse Vincent, who blogs at https://blog.fsck.com/. His blog is excellent—he writes about software development, AI collaboration patterns, and building better tools. If you found this useful, check out his work. He’s super cool.

From Structured Ideas to Structured Products

The same way a framework helps you think clearly with AI, structure also matters when building with it. Many founders start with AI-generated MVPs that work just enough to get traction, but quickly become fragile and hard to scale. That’s why we built Spin by fryga. It’s a consultancy that helps founders bring stability, clarity, and structure to AI-built products, so they can ship faster, fix smarter, and scale without fear of things breaking.

How I learned to brainstorm effectively with AI: A structured approach using Claude