Give us your brief. See where a second model takes your ideas.
Most people ask an AI for campaign ideas once and take the list. Fractl Agents does that too: it asks Claude for forty ideas from your brief. Then it writes forty more a different way. A second model throws out forty one-line starting points, and Claude turns each one into an idea for your brief. Every idea is then grouped by angle, the move it makes to get attention, and you see where each batch put its ideas: the angles both tried, and where one leaned in.
Keep working with these ideas in Fractl Agents
We email you a link. It opens a free Fractl Agents workspace with $200 of credits, and your ideas open in the chat with a spreadsheet of the full list: every idea, which batch it came from, the repeats merged into it and the model's shortlist.
Send yourself the link within 24 hours; after that the link stops working. Your brief was deleted when the run ended, and the copy on this page is deleted after seven days.
What Fractl Agents does during the run
It checks that this is a marketing brief
A quick read: is there an organisation or product, and something it wants? If not, nothing runs and nothing is counted against your two runs.
Two batches of ideas are written side by side
Claude writes forty ideas from your brief in one request, the way most people would ask. At the same time a second model, DeepSeek, writes forty one-line starting points, and Claude turns each into an idea for your brief, dropping the ones that do not fit.
Repeats are merged
One strict pass groups ideas that share the same mechanism and the same subject. A different angle on the same subject stays separate.
Every idea is rewritten into the same short shape
A title, two sentences and a proof point, so you judge the idea, not how it was written.
Every idea is grouped by angle
One more pass sorts all the ideas by the move each one makes to get attention, so you can see which angles each batch went after and where it put its weight. A model also picks the ideas it would pitch.
You read the result
First a chart of where each batch spent its ideas, then examples of the difference, then the other ideas a model would pitch and every idea with the repeats merged into it.