@jomo @dw_innovation @Bellingcat Can you link your talks? I would be interested.

Some issues I could think of here is the use of ChatGPT, you would probably want this a little more in-house. Also to be dependable for OSINT, you would need some sort of fuzzy matching, where not all features are a must or could deviate from what you described. Lastly some sort of inclusion of historical changesets may be great, since places change all the time :)