Let's break down what exactly has changed with @RallyOnChain updates, explained with specific logic.

Old state: 1,000 accounts participated in an activity, of which 600 are bots or spam accounts, 300 are casual spam posts, and only about 100 are genuinely thoughtful. The reward pool is diluted among 1,000 submissions, so those who put in real effort receive a very small proportion, resulting in a poor return on investment.
After the new state: The Minimum Sorsa Score threshold directly filters out a large number of low-quality accounts; bots and farm accounts cannot participate. Manual bans further clean up any remaining bad actors. Max Winners Per Period limits the total number of winners; for example, if the cap is 50, rewards are only distributed to the top 50 scoring submissions.
With the same reward pool, the quality of participation improves, the number of winners decreases, and each earnest creator’s expected reward increases significantly.
The AI scoring criteria are public: content relevance, information accuracy, originality, interaction quality, with each weight visible on the platform. You know why you passed or failed, so you can target your improvements accordingly.
On-chain settlement means the distribution process is transparent and auditable; there are no black boxes. Funds are held in a custody contract, and once AI scores are calculated, the distribution is triggered automatically without manual intervention.
The result of this mechanism design is: the positive correlation between your content quality and your reward expectation has been strengthened. On most platforms, these two aspects are almost decoupled.