Giving the answers more context
The first answers were too general. I added research from recent Reddit and X discussions through Perplexity, giving the fictional customers more information to respond to.
Socrados is a prototype built by five people at a Copenhagen hackathon. It creates AI personas who give feedback on an idea and discuss it together. I helped build how they respond and interact.

Testing every small idea with real people can be expensive for agencies. We wanted to explore whether AI could help them prepare for that research. The agency owners we spoke to had one main concern: could they trust the answers?
I worked with a teammate on generating the fictional customers and their feedback. My parts included individual chats, group discussions, web research, and suggested edits to the idea.
After the weekend, I kept working with another teammate for a week. I added more varied personalities, replies in the user’s language, a page to track AI usage, and a landing page.



The first answers were too general. I added research from recent Reddit and X discussions through Perplexity, giving the fictional customers more information to respond to.
I built discussions where each fictional customer reads the others’ replies before answering. They use different AI models, and an extra round asks them to challenge the group’s view.
Socrados stayed a prototype. We never fully answered whether its feedback could be trusted. We had a good time, shared a few laughs, and did what we all love: building things, trying out ideas with testers, and watching how they used our app.