What makes a good assignment in the age of AI?
Socrat is built around a single question. Students are going to use AI; the assignment should turn that into the work and show you what they actually understood.
Built from inside the AI shift, and inside the classroom.
Socrat's founding team watched AI transform day-to-day work at Meta and Microsoft: what gets written by hand, what gets delegated to a model, how good engineers re-learn their own craft.
The same shift was visible from the other side, while teaching at Columbia. Students had access to tools more capable than the assignments they were being given. Instructors were left guessing what was a student's thinking and what was a model's.
Socrat came out of those two vantage points. The tools change what good work looks like, and assignments should change with them. The instructor sees how every student is thinking, and the student leaves better at learning.
A focused team, holding a few things tight.
- Conversation as the assignment. The thing the student does is the thing you grade. No essay middleman; no authorship guessing.
- Ground the AI in your course. Your readings, your syllabus, your standards: the model answers from what you teach.
- The opportunity cost of data. So many conversations happen about a student's understanding, and only the final product gets turned in. Instructors are often unclear on what a student actually knows until test day.
- Privacy is the floor. Built FERPA-aligned. Student data stays scoped to your class.