Convex optimization seems to be an increasingly important topic to know given AMM design, research and growing popularity within finance. Just sharing a few resources that I've found helpful for learning and revising it.
Wrote a second part on coprocessors with
@0xkrane
. In this we identify the trend of offloading compute in the Ethereum ecosystem and put it into perspective about how we are heading to a proof based world
An open list of problems spanning across crypto (2nd Edition)
The problems we solve decide the direction we take in the future and the adoption we see.
Wrote about Part 2 in the series on "The theory of MEV."
The best way I've thought about this paper is that it describes sampling rules for MEV like the one we see in Signal Processing.
Wrote an incomplete primer on Intents.
The discourse around intents has fascinated me since it looks like a much-needed evolution of UX. I believe intent architectures will mostly be application-specific.
@DegenSpartan
So there's actually research which says that hidden voting is better (also seen in ballot voting), because passive voters tend to subconsciously prefer majority vote in public votes (when they arent privy to things). Strange we have not seen these experiments.
Wrote an introductory article on how blockchains could become a privacy layer for AI, specifically covering recent MPC protocols for transformers and some other thoughts.
Wrote about Multidimensional blockchain Fee Markets.
As L1s and L2s proliferate, different resources demand different fee markets. A single-fee market doesn't capture demand or usage.
Wrote a blog about Exploring Privacy in CFMMs. As we design new AMMs to address various aspects, privacy will also have a moment in the light and become a core focus. It's important to keep in mind that it seems we trade off privacy for utility
My current favourite is Replicating Market Makers, I just didn't know replicating payoff functions was possible (I haven't gotten to the followup paper yet)
Never got why DeFi frameworks are built with "we'll play second place" and make oracles integral vs "we'll become the primary place for price discovery and attract volume"
Published the second part in the MEV supply chain series. In this i cover the enshrined proposer builder separation landscape mostly focusing on in protocol solutions to pbs and end with talking about the cost of enshrinement.
ETHEREUM'S SUPPLY CHAIN, PART 2
In part 2 of this series,
@0x_emperor
explores proposals for enshrining PBS.
We start by examining the reasons to enshrine, then move on to design goals, the proposals themselves, and other dependencies that implementations must be aware of.
Do the buyers understand that you're mostly just going to lose it all to funding, except if you catch it in a swing which itself might be an extremely crowded trade so the upside might get capped for everyone
ETHEREUM'S SUPPLY CHAIN, PART 1
In this series,
@0x_emperor
explores the past, present, and future of the Ethereum supply chain.
This first part describes Ethereum under PoS and MEV-Boost, and analyzes issues with the current architecture.
I think it's an exciting time to be in crypto research because there's meaningful progress and you can impact products/future research, the field is smaller so you can know everyone. Everyone flocking to AI these days will quickly realise that you'll never be relevant.
Credible auctions are more critical than ever as our digital frontiers grow on-chain and off-chain. MEV auctions, Art auctions, and Ad auctions are just a few types that permeate our lives. Wrote a blog exploring a recent work about auctions on chain
@epolynya
yeah once i started using dydx, i saw no other reason why someone would want to go to CEX, it didn't even have any downtime during flash crashes in the last month. CEXs will get obscured into being fiat onramps at best in the future
More often than not the answer to "where does the yield come from" in DeFi is emissions, explicit or hidden ones. Emissions are a good bootstrapping/introductory tool in moderation but if you need to pay your users to keep using your product in perpetuity, is there any loyalty?
I know most engineers in crypto don't like the nascent research side in MEV/AMMs etc., but it's important to understand that research begets theory, and builders of tomorrow will need some framework, to begin with. As a field matures, newcomers rarely have the luxury of learning
Coprocessors (ZKs, TEES etc)
I am not one to predict trends, but we'll see more designs leveraging coprocessors and, hence, a new wave of defi apps being possible on the chain. Doesn't have to be dynamic lending protocols but also things we do today, just safer and transparent.
This is an incredulous amount of compute being wasted and theres more compute being wasted on failed liquidations.
by
@GuilleAngeris
would be useful for solana imo and much needed
did you know 58% of solana's compute is spent processing failed arbitrages?
jito-solana contains the systems to efficiently capture this MEV and redistribute it to stakers and validators on the network.
if you want to see a more efficient solana, vote with your stake and stake
Wrote about my experiences as an amateur in crypto while I study stoch calc and make my way through RMMs. I'm not a great trader/investor so this is NFA of course.
Lido looks like it’ll be one of the first protocols to breach the billion dollar revenue mark in a few years, while the path to distributing this to token holders seems unclear. Will we see a point where US regulatory uncertainty becomes a net negative for the space?
@AutomataEmily
"we do a few thousand dollars a week"
"oh yeah its just human labor but instead of physical labor you use a phone"
"Some of our workers have carpal tunnel though"
@tarunchitra
@AaveAave
While his takes are usually very based and I love reading his articles, at least when it comes to crypto it seems try hard and uninformed, I often wonder if that is because defi would make their(stripe's) vision obsolete on some level
Are order books JIT LP liquidities over a finite tick? If yes then doesnt that mean the cheaper txes get and the more we converge to ubiquitous JIT liquidity, we’d even possibly simulate a order book on some level over tick?
@dystopiabreaker
They dont want wealth to be a positive sum game, just behind closed doors, access to a few so they stay in power. One of my favorite lines in a book by will durant is “revolution always seems to occur between diastoles and systoles of wealth concentration and distribution”
Today, we launch Infernet, an inference engine by which you can use foundation & classical ML model outputs in your web3 applications and access them on-chain.
As "Verifiability" becomes the next core issue of LLMs, there's a middle ground which is a best-of-both-worlds formulation and LLM "proofs of knowledge" become more and more important. Tarun ser, formulated this well in his blogpost. Check it out!
If Plato and Yudkowsky had a debate about epistemology, who would win?
@priorupdates
and I try to answer this by investigating what it means for an LLM to have “knowledge” from the lens of interactive proofs (IPs) and ZK arguments of knowledge
The list is not meant to be comprehensive (since covering all of crypto is quite impossible), and I'm happy to add more from researchers/builders in these spaces.
Crypto X AI
Excited about the work we are doing at Ritual towards building an ML layer for all of crypto & generally, the kind of structures crypto helps us build for democratizing AI. Read for more crypto-specific AI problems, and of course
@ritualnet
With cryptographic primitives like MPC, TEEs, ZK, FHE becoming possible and capturing mindshare in crypto, it is time to look at applications these primitives unlock in the context of modern AI. Happy to talk to people interested in building with Ritual and help them where I can
The violence really starts when these DAOs that have all accumulated each other's tokens start exploring going into stables to avoid treasury value erosion or otherwise risk losing upto 50% of their treasury value.
Palworld will probably change the expectations from all web3 games, a total of 7M raised, 1.2M concurrent users at the top and going up, 6M in sales. In hindsight looks very evident to have tried a “pokemon with guns” game.
@mewn21
I hate to be the fundamentalist cuck, but is there any reason beyond the LMs to look at these chains is something I ask myself everyday and that's the answer I came up with.
Maybe out on a limb, but when you learn that 4 builders build 88% of Ethereum’s blocks you wonder if thats a world apart from 3 mining pools controlling bitcoin. Begs to ask the question, how are we decentralising these stacks?
@DegenSpartan
@inversebrah
The anxiety that half of these bridges induce for the moments that nothing shows up on either blockchain and your money is gone is stressful af. bridge ux is a meme
Knowing how to formulate a convex optimization problem and/or understanding a given one is the most important skill to acquire. While Convex Analysis seems to be more important for understanding things now (convex optimization will be important in the future too)