During the conference, I looked at questions from several project organizers, and some of the questions are quite insightful.

The relationship between AI development and Token Economy. Many crypto projects like to directly translate AI business models into tokens, using tokens to incentivize supply, capture network value, govern, and settle transactions.
The problem is, most token designs address fundraising issues, not transaction issues. Token Economy will not disappear, but most projects will lose because they haven't created real demand.
In the next 2–3 years, the market will increasingly reject the idea of "having an agent means needing a token." Because what users truly care about is whether this agent can help them complete tasks, whether the results are valuable, who will compensate for the service, and why this call must use your token instead of stablecoins or fiat currency.
If tokens cannot solve these issues, they are just old-school crypto plays under the guise of AI.
In the AI Agent era, if tokens cannot be embedded into real workflows, they will only become spectators' assets even faster than before.