Environments, not advice

"Environment" is the lab's preferred unit of design. Not "tool," not "platform," not "system", environment.

An environment is a place where work happens. The work belongs to the person, not the environment. Advice tells you what to do. An environment gives you somewhere to do it.

This ethic shows up across every project in the lab:

• Synapse is a map you wander, not a recommendation you receive.
• DartWorld is a 3D world you walk through, not a quiz that ranks you.
• CDL is a room full of materials and a 15-person cohort, not a lecture.
• NBD is a 90-minute workshop format, not a deliverable you receive.
• Threshold is a toolkit you assemble for your own practice, not a curriculum you follow.

The choice matters because of how each one fails when it goes wrong. Advice fails by being ignored or wrongly taken. The student either follows it (and may be wrong) or doesn't (and you didn't help). Environments fail by being empty, the student walks in and doesn't know what to do. That failure is more visible and more correctable. The student is doing their own work, so the failure mode is in the design of the room, not in their compliance with your suggestions.

There is a second reason. Advice presumes you already know what the person needs. Environments do not. The same environment can support a student who's months ahead and a student who hasn't started thinking yet, because the user works against the room's affordances, not against the system's model of who they are.

This is the harder design problem. It's also why most career tools are advice tools. Advice is easier to build, easier to evaluate ("did they take the advice"), easier to justify to funders. Environments require a different evaluation logic, measure what the user produced in the space, not whether they followed instructions.

Open thread. The temptation to ship a recommendation is constant. Users ask for it. Funders ask for it. AI makes it easy to produce. The discipline is to keep designing environments that surface evidence and let the user do their own deciding, even when the environment looks, on the surface, like a tool that should just tell them the answer.