This is a pretty poorly researched and written article, but the core idea of switching to a trust-the-individual decentralized network is a solid one. It seems

This is a pretty poorly researched and written article, but the core idea of switching to a trust-the-individual decentralized network is a solid one. It seems like we'll need a couple more advances before it's reasonable, but if you can move to a system where there is no ground truth, but rather a record of successful transactions or proof of stake/reputation in an identity then you only need to reach out to the global state to see if there are outstanding deals with an identity or broken promises. The basic idea I've been floating is something like the following: 1) Alice generates a private/public key pair 2) Alice performs some work on their public key (e.g. low hash) 3) Alice finds Bob and promises to send 0.1 BTC to Bob's public address IFF bob performs publically verifiable service X (e.g. provides boolean assignment that satisfies some constraints or rejects the problem as unsolveable) 4) Both parties transmit the unsigned contract 5) Both parties reach out to trust brokers that tell them if either party has been involved in a broken promise, or has outstanding contracts greater than their current credit. Don't sign contract if thats the case 6) Both parties sign the contract (outstanding problem here, this needs to be atomic) 7) If contract goes well each party appends this to their record 8) If contract goes poorly offended party broadcasts this to the network and gains some portion of the lost reputation. There are many outstanding problems that are not fundamental, but serious obstacles to writing this today. The ones that I know of off the top of my head are: 1) Need well understand exchange rate of reputation -> contract value (might be doable if you can buy reputation) 2) Need non-Sybil trust brokers (e.g. trust broker with rep >> contract value) 3) Atomic contract signing 4) contracts need to be publicly verifiable. Any problems I am missing?