'Beyond Blockchain: Convergent Consensus' by Mike Anderson (Strange Loop 2022)

Despite considerable hype over recent years, blockchain technology has often failed to live up to the promise of providing cheap, accessible, high performance decentralised computing.

We can do much better: A novel approach to consensus inspired by CRDTs and category theory can make decentralised computation efficient and scalable. Immutable persistent data structures and digital signatures help keep everything secure.

In this talk we explore the Convergent Proof of Stake consensus algorithm and demonstrate how it can be used to implement efficient real-time decentralised computation using Convex Lisp, a functional language inspired by Clojure and the lambda calculus.

Mike Anderson
Managing Director, Convex Foundation
@mikera

Over a varied career Mike has been CTO at several companies as well as a strategy consultant at McKinsey & Company where Powerpoint and Excel were the development tools of choice. Since the age of eight, Mike has had a passion for writing software and developed a strong interest in open source technology. He is a contributor and maintainer of open source code in the Java and Clojure ecosystems. In a vague attempt to find balance in life, Mike also volunteers his time as a modern jive dance teacher.

------ Sponsored by: ------

[MarkRuvald]

Thanks for the presentation. Very enlightening. It just strikes me that every callable transaction, the execution payload is tiny. It should seem that we need a way to drop the established stateful history, since some older nearest common ancestor.

[drdca]

I missed how this is supposed to avoid Sybil attacks? He mentioned proof of stake, but up until 31 minutes doesn’t seem to have gone into that.
“we use a lisp as the language for it” is cool I guess, but I feel like explaining details of that could go after the “why is this secure” bit?

[virdvird]

Less protected against sandwich attack then regular PoS or PoW (if we check only tx consistency, not tx result consistency).
Or doesn't allow multiple users interact same smart contract at same moment (if we check tx consistency and tx result consistency)

[rtBen]

Has he realised 1. The novelty of crdt over unconfirmed blocks has no benefit over single large blocks, because txs are already gossiped and the block is a resource constraint; 2. For everything else, Ethereum…?

[nilskp]

On Bitcoin: "it's certainly not scalable, you couldn't run a real global system on this"

Lol, Bitcoin has run globally for 13 years, and without any 51% attacks, even with the enormous incentive to do so.