← All classes
CLASS 10June 24 · AI-Enabled Ideation

"It doesn't matter what the model learned from,the interesting thing still comes from you."

CT
Taught byCeren Topçu-WeigandCreative AI for cultural institutions, Pratt alum
Built the Gropius Bau ChatGPT visitor guide (2022)
Sydney Seifert-Gram
WithSydney Seifert-GramTook the second half: AI for research and experimentation

The class in brief

Ceren opened with Margaret Boden's three types of creativity and the case for a personal archive as "ground zero," the source you check AI output against because the model is sycophantic and will back your worst idea. Sydney closed with her own research method: interview-style prompting, forced connections, and a conversational path to a portable writing-style guide. After this page you can run the Interviewer and Interpreter modes on a real brief, force two unrelated domains into a new idea, and check any AI output against your own archive instead of its enthusiasm.

The night at a glance

Why this matters · 6:13 PM

You define good. The machine generates toward it. You decide when it's done.

Ceren opened on five objectives for the night : know your creative edge, co-creation between human and AI, ideation methods, a Crazy 8s activity, and practical tips. The frame underneath all five is the same co-creation slide line that gives this page its hero quote: you bring the intent, the machine generates the range, and judgment stays yours.

"AI is sycophantic and will back your worst ideas, so you need an external source you trust as ground zero."

Sydney, on why a personal archive matters

3
creativity types, after Margaret Boden: exploratory, combinational, and transformational. Most AI-assisted work lives in the first two.
4
things that make up your "creative edge," the part AI can't do for you: discernment, intuition, emotional depth, cultural nuance.

The framework · 6:33 PM

Boden's three types of creativity

01

Exploratorythe most common type, and AI's strongest match

Deep immersion in a known style until you find what's possible at its edges. Example: Monet's Water Lilies, the same pond for thirty years into abstraction . AI is pattern recognition, which makes it a strong partner for pushing a known style and generating variations.

02

Combinationalmerge two unrelated domains

Connect fields that have never met to make something neither could alone. Example: the first iPhone, iPod plus phone plus web device, 2007 . AI moves fast across distant fields, which is why Forced Connections works.

03

Transformationalrarest, and still an open question for AI

Break the rules everyone thought were fixed and start over with new ones. Example: Copernicus putting the sun at the center . AI can help humans find transformational ideas; whether it can originate a paradigm shift alone stayed an open question in the room.

Boden's three types, defined on one slide
Boden's three types, defined on one slide · 6:33 PM

Ceren followed the framework with the Rick Rubin line that anchors this page , then named the four things AI still can't do for you .

The idea · 6:40 PM

Your personal archive is the blueprint of your creative identity, and the thing you check AI against.

Old photos, sketchbooks, writing, values, saved references, even saved AI outputs : this is Ceren's answer to a model that flatters everything you show it. Sydney's follow-on framing made the co-creation loop explicit : you bring the creative intent and set the vision, the machine generates many variations, you explore the range and decide when it's done.

4
LLM ideation modes named in one slide: Interviewer, Critique, Designer + Builder, Interpreter. Method 01 and Method 02 below both come from this one.

The exercise · 6:51 PM

Running Forced Connections live on the Ideation Atlas

The tool

39 ideation methods, each with an AI-assisted variation.

Ceren's own build, on Base44 : an explanation, an analogue practice, and an AI-assisted variation for each of 39 methods, sorted into Divergent/Classic, Perspective & Empathy, Association & Connection, Reframing & Provocation, and Narrative & Future. The room worked Forced Connections: name your challenge, add a few genuinely distant domains, ask for a handful of real insights from each, then pick the collision worth pursuing.

Cap the output. About five sentences per domain, so it doesn't flood you.

Stay vague to reach surprising places

Tony's warning, repeated through the room: if you let on that you're honing toward one idea, the model "chases you and paints you into that corner." The method works because the domains stay unrelated on purpose.

Branching a thread

Edit an earlier turn, or use the three-dots menu and "Branch in new chat," to fork a research direction while keeping the original intact. One classmate reached Pauline Oliveros's Deep Listening this way, starting from a science-center exhibit brief.

The power is in the prompt

Sydney's own admission: she wrote her research-collaborator persona, then used an LLM to troubleshoot and harden it, so it pushes back instead of turning into a generic idea machine. Writing the prompt is itself part of the craft.

The activity · 7:14 PM

Eight ideas. Eight minutes. Eight panels.

Crazy 8s closed the ideation block : a brief, a strict clock, and eight quick panels, described in plain language before any image generator touches it. Sydney's own 5-tips slide, which followed the break, reads as the instruction manual for everything the room had just done .

"Describe your rough sketch in plain language, then let the image generator work from that."

Ceren, on the Crazy 8s hand-off to AI

Sydney's 5 ideation tips, the whole class was told to screenshot this one
Sydney's 5 ideation tips, the whole class was told to screenshot this one · 7:43 PM

Methods and prompts

Five methods to take with you

METHOD 01 · 6:43 PM part of: structure your ask

Interviewer mode

Before you generate anything, let the model find your real problem. It asks one question at a time until the actual challenge surfaces, instead of you guessing at a brief.

Where it came fromOne of the four LLM ideation modes Ceren Topcu-Weigand named in class 10, where the model asks questions one at a time until the real challenge surfaces instead of you guessing at a brief.Use it whenUse this at the very start of a project when you are not yet sure what the actual problem is and want the model to help you find it before generating solutions.

Working prompt

Ask me questions one at a time until you understand my challenge. Don't offer solutions yet, just keep asking until the real problem is clear, then summarize it back to me before we go any further.

You will know it worked whenit asks one question at a time and holds off on offering solutions until it summarizes the problem back to you.

METHOD 02 · 6:43 PM part of: reject the first draft

Critique mode

Grade your own idea before you present it, using an outside voice instead of your own comfort with it. Judge it yourself first: the model's job is to check your grade, not set it.

Where it came fromAnother of Ceren Topcu-Weigand's four ideation modes in class 10, where the room graded their own concept first, then had the model act as an outside design educator to grade it independently.Use it whenUse this once you have a concept ready to share and want an honest outside grade to check against your own.

Working prompt

Here's my concept: [paste]. I'll grade it first: my score out of 10 is [your number], and the weakest part is [your answer]. Now act as an experienced design educator, grade it 0 to 10 yourself, and name the 3 most important improvements. Tell me where my own score was wrong.

You will know it worked whenit gives its own number apart from yours and names three specific improvements, then tells you plainly where your own score missed something.

METHOD 03 · 6:43 PM part of: keep the judgment human

Interpreter stress test

Rewrite the idea in three unrelated voices and see what survives. Decide for yourself which retellings still hold the core meaning before asking the model to check your read.

Where it came fromTaught by Ceren Topcu-Weigand in class 10 as a way to stress test an idea by rewriting it in three unrelated voices, such as a tabloid headline, a love letter, and a legal contract, to see if its core meaning survives.Use it whenUse this when you want to know whether an idea's core meaning is strong enough to hold up outside its original form.

Working prompt

Rewrite this idea [paste] as a tabloid headline, a love letter, and a legal contract, keeping the core meaning intact in all three. My read: it survives in [name which ones, for you]. Now you check me: does the core idea actually hold in the ones I picked, and where does it break in the others?

You will know it worked whenit checks your guess about which retellings still hold the core meaning against what actually survived in each of the three versions.

METHOD 04 · 6:51 PM part of: go wide then narrow

Forced Connections

Ceren's Ideation Atlas method, run in a plain chat. Give the challenge, add distant domains on purpose, cap the output so it doesn't flood you.

Where it came fromRun live in class 10 from Ceren Topcu-Weigand's Ideation Atlas, forcing a challenge through several deliberately unrelated domains to surface genuine insights from the friction.Use it whenUse this when you are stuck on a familiar idea and want a genuinely new angle by connecting it to something far outside its usual context.

Working prompt

Here's my challenge: [paste]. Force it through these four unrelated domains: [name four, deliberately distant]. For each one, give me 2 to 3 genuine insights, about 5 sentences total per domain. Don't soften the domains to fit the challenge, let the friction do the work.

You will know it worked wheneach of the four domains produces insights that actually come from that domain's own logic, not the same idea restated four times.

METHOD 05 · 8:02 PM part of: context beats prompts

Build your writing-style guide

Sydney's second method: edit a sample paragraph conversationally, then have the model analyze the whole exchange and write a reusable style guide you save and upload wherever you write.

Where it came fromSydney Seifert-Gram's method in class 10, editing a sample paragraph conversationally and explaining each change, then having the model turn the whole exchange into a reusable style guide.Use it whenUse this once you want a portable record of your writing voice that any AI tool can follow without you re-explaining it each time.

Working prompt

Write two paragraphs on [any subject]. I'm going to edit them conversationally: for each change I'll tell you why I'm making it and the tone I actually want. Once we're done, analyze the whole exchange and write me a reusable style guide, as rules with examples, that I can save and hand to any tool.

You will know it worked whenthe style guide it writes back names specific rules with examples pulled from your actual edits, not generic writing advice.

The close · 7:43 to 8:03 PM

AI opens the research space wide. You keep judgment and voice.

Sydney closed the night with her own five-part method , then handed the room a 30-minute activity to run it live: paste her creative-research-collaborator prompt into a new chat, give it a brief, and let it interview your instincts before offering domains and real references, never finished ideas . Push back on its first answers. Save the fragments that work and grow them into later projects. Ceren and Sydney closed together on contact info before the room broke .

Try this prompt

Quiz me on Boden's three creativity types and the four LLM ideation modes. Then give me a small challenge and make me run it through Interviewer mode and Forced Connections.

You will know it worked whenit quizzes you on Boden's three creativity types and the four ideation modes first, then makes you run a small challenge through Interviewer mode and Forced Connections.

The shelf

Tools and references

Tools that night

  • Ideation Atlas, ideation-techniques.base44.app, 39 methods built by Ceren on Base44
  • ChatGPT and Claude, used live for Forced Connections and the research-collaborator prompt
  • MCP (Model Context Protocol) explained as a concept, example given: Canva connected to ChatGPT

Named in the room

  • Rick Rubin, The Creative Act · Margaret Boden, creativity types
  • Sila Sehrazat Yücel and Keepkwan, AI artists referenced for the creative-edge discussion
  • Pauline Oliveros, Deep Listening · cerentopcu.com, hello@cerentopcu.com
← Class 9 · AI Remix Jam CLASS 10 OF 24 Class 11 → Data Cleanup & Prep for Agentic Workflows