Product discovery
The first research question is whether agents can identify and evaluate product work from an outcome. The test should compare hypotheses before code is commissioned.
Research
Longer-term ideas stay labelled as research.
Autonomous product organisations combine software delivery, product discovery, evaluation, economics and governance. The early engineering evidence is real. The broader institutional claims are proposals to test.
Agent evaluation
An evaluator must be checked against known cases and human review. A model score is not a calibrated probability simply because it is written as a percentage.
Organisational design
Roles, budgets and decision rights need their own evidence. More agents, more completed issues or less human involvement are not enough.
Economic questions
The organisation should account for model cost, delivery delay, verification effort and the opportunity cost of not trying a different intervention.
Competing organisations
Competing teams may help when evidence is weak, but a small LinkedIn extension does not justify a standing ecosystem by itself.
Institutional evolution
Changing governance rules is a public-interest research problem. Agents should not silently rewrite the authority structure that governs them.