Ironically the OP is called A_D_E_P_T - ADEPT was the first "proper agent project" I got involved in, back in 1996! I just looked at a couple of deliverables &

Ironically the OP is called A_D_E_P_T - ADEPT was the first "proper agent project" I got involved in, back in 1996! I just looked at a couple of deliverables & tidied a demo as it was closing then, but then I did more work on lots of old MAS projects for years after. Here's the thing - we failed. Why? Three things. - service providers like Expedia, Lastminute.com, Zwillow did the data integration by hand. They were able to afford to do this (the bit we didn't expect) because they were able to attract adverts and charge a transaction fee. They had a business model. - Specific applications like the HMRC self-assessment were able to offer a user experience that gave close control and easy completion. The agents we built were able to do easy completion but they took control off the user and could not perform at 100%. Users were not confident and so didn't use them. - We did most of the work pre iPhone, so we didn't have the right interface. Of course that has changed... but the Appstore also changed things and user expectations are sky high now. It was also fersum difficult to program proper MAS. The issue is creating a system based on the expectation of behavior of many other systems which is of course a combinatorial mind bender. We mitigated this a bit with ideas from economics and game theory, but I think we were a long way off solving it. LLM's do provide extra capability for solving some of these problems, and you young people are so clever and hard working that maybe you can just blitz through some of the others. I can't get at Manus yet, but I don't really expect that it will be a proof of this.