I've thought some about trust minimization over the past couple years as I've delved into things blockchain. To me, there are different levels of trust that have to be analyzed and reproduced in order for two parties to come together. This is a somewhat different point than what Szabo is making here as I am looking at the social-interaction level.

At a very basic level, there is the trust of competence. This is maybe 10% of what matters. If you are a software developer and someone is looking to hire you, then some kind of objective display of knowledge is beneficial.

The other 90% though is a much more personal level of trust. Is this person someone I can rely on? Will we get along? Is this person competent, but a complete jackass?

To that end, I think we will see evolving a domain-level, blockchain-based reputation system. It will be like LinkedIn's system where others vote for you, but within different domains of knowledge. Person A might say you are a skilled mathematician. Person B might say you know your craft beers. Person C might say you stick to your commitments.

Does Person D want to business with you? D can check your scores and make a determination.

It's not a truly trustless system, but it decentralizes trust from each person and spreads it to the network so you aren't relying solely on one person's representation of himself.