Lately I have been thinking about what it means that OpenLedger funded a $5 million Cambridge research program the same month mainnet launched.
Most projects allocate research budgets after traction. Before traction usually signals the team knows the hardest problem is still unsolved.
The hardest problem in verifiable AI attribution is not storing records on-chain. That part is solved. The hard part is the accuracy of the attribution calculation itself. Determining how much a specific training example genuinely influenced a specific output is an open research problem at scale.
A mainnet running on influence-function approximations while the foundational accuracy question is being researched in parallel is worth understanding clearly before forming a view on what the attribution rewards actually represent.