Some time ago, I developed an A-shares sentiment monitoring system (today I saw Lobster🦞).

The main focus was on a comprehensive model update for my previous single North-South position monitoring. Although there were constant bugs, the overall framework has been established—18 Agents simultaneously scraping sentiment data from Eastmoney, Snowball, Tonghuashun, and Taoguba.
Optimistic and Pessimistic Agents compete and score, and if their opinions diverge too much, an automatic arbitration is triggered. Additionally, the model iterates its weights through signal backtracking from T+1 to T+7.
In simple terms, I want the machine to monitor the market for me—much more reliable than me scrolling through posts myself.