Labs

Synthetic Research Personas

lab · demo

We pick one of five research projects, pick any or all of the four people written from that project’s research, and put a single question to them.

Each person answers from their own goals, their own pressures, and the passages they were built on. Every answer appears the moment it lands rather than at the end, so the first is there to read at about twenty seconds while the last is still being written. What makes it worth running is the comparison underneath: four answers are raw material, and where they pull apart is the argument a real team would have. Every person here is synthetic, written from research rather than interviewed, and the page says so on every answer.

Checking the Labs AI…

This demo builds its projects and its people from the workshop as the page loads, so it needs JavaScript switched on.

These personas are synthetic. Each is written from research on this project, not from an interview with a real person.

The research project

Each project has its own research behind it, and its own people written from that research. Pick the one your question belongs to.

Loading the projects…

Ask one question, between ten and five hundred characters. A question with a trade-off in it is what makes four people disagree rather than politely agree.

0 characters, 10 needed

Each person is asked in their own pass on a model we host, and every answer appears below the moment it lands, so there is something to read from about twenty seconds in. The bench answers one question at a time, so you may find it busy.

How this works

Five research projects sit behind this demo, and each one carries its own research and four people written from it. Ask a question and everyone you picked answers it in turn, from their own goals, their own pressures, and the passages of the research they were built on. Every answer lists the passages it drew from, so a claim has somewhere to point.

Answers arrive one at a time rather than all at the end, which is what makes the wait worth sitting through: the first person is on screen at about twenty seconds, while the last is still being written.

Where they agree, and where they do not

Each person answers from their own research and their own job, so you get four takes on the same question. Sometimes they pull apart hard enough to name, when the program manager would cut monitoring to fund training and the analyst would do the exact opposite, and the comparison leads with that split because it is the argument a real team would have. Sometimes they land in the same place for different reasons, and it leads with what they agreed on, then with what only one of them raised.

These people are not real

Every persona here is synthetic, written from the research on its project rather than from an interview. They are a way to rehearse a conversation, not evidence of what anyone believes, and no answer here should stand in for asking the people who actually do the work.

Nothing is kept

Each question stands on its own. There is no memory between runs, no history to come back to, no link to share, and nothing you type leaves the bench.

Back to Labs