samlaf
@samlafer
Followers
1,675
Following
1,294
Media
306
Statuses
1,820
Eigenseal @Eigenlayer . Maintainer eigensdk-go. My tweets are my own.
Montreal
Joined January 2010
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
Rafah
• 1277204 Tweets
#ŞimşekMebeMülakatsız68Bin
• 439504 Tweets
Fenerbahçe
• 135351 Tweets
Egypt
• 107972 Tweets
Superman
• 95031 Tweets
#الاهلي_الهلال
• 72141 Tweets
Qatar
• 68435 Tweets
Manchester United
• 56617 Tweets
Crystal Palace
• 39479 Tweets
مالكوم
• 39397 Tweets
لودي
• 30851 Tweets
Evans
• 27191 Tweets
Bernard Pivot
• 24838 Tweets
David Corenswet
• 24102 Tweets
Olise
• 22835 Tweets
#KONvFB
• 20862 Tweets
Casemiro
• 19968 Tweets
#CRYMUN
• 19097 Tweets
Hades II
• 18478 Tweets
ميشيل
• 16209 Tweets
Cavill
• 14762 Tweets
محرز
• 13716 Tweets
Hades 2
• 13641 Tweets
Dzeko
• 12917 Tweets
جيسوس
• 12474 Tweets
ŞAMPİYON GALATASARAY
• 12473 Tweets
Pat Riley
• 12070 Tweets
Cengiz
• 10270 Tweets
James Gunn
• 10135 Tweets
Pinned Tweet
Check out our new work. Been thinking about "illuminating the dark forest" of Narwhal based chains for a while, and
@0xtrojan_
the powerhouse was just the perfect fit for this task. Was a blast to work with him.
1/ MEV 🤝 DAG: As a part of the
@eigenlayer
research fellowship, I worked with
@samlafer
on analyzing the landscape of MEV on chains that use Narwhal DAG-based consensus mechanism ♾️🧵
Here is what we found 👇
3
20
125
3
2
30
Just joined
@eigenlayer
! Super stoked and looking forward to learning from, contributing to, and pushing forward with one of the top teams in the space. Let's BUIDL!
29
9
206
Ethereum started with a mempool and flashbots removed it.
Solana started with no mempool and jito added one.
15
12
112
This past month I got the chance to participate in
@wintermute_t
x
@encodeclub
's MEV hackathon.
@0xK1000
and I just submitted our project "MEV mitigation using modular DAG-based consensus protocols" (). Would love some criticism/feedback!
2
10
70
We wrote the first (?) btc lightclient on starknet at the starknet hackathon in Amsterdam. Check it out:
Thanks to
@GuthL
for the constant moral and technical support, and my awesome teammates
@maxgillett
and BrDu
#1111
3
13
58
Did I just read that Sui has a way to enforce NFT royalties onchain?
IMO this is one of the best security features of Sui's native object ownership--classic wallet drainer attacks (e.g., giving a token/NFT approval to a smart contract that later gets hacked) are impossible by construction. If coins or NFT's are owned by address A, it is only…
12
52
192
9
3
50
For those having trouble understanding the connections between the different parts in
@svlevine
's (awesome) control as inference tutorial, I made this diagram synthesizing the major relationships.
If you want to know how probabilistic inference can be tied to optimal control, I just put up a new tutorial on control as inference:
This expands on the control as inference lecture in my class:
6
280
808
0
10
49
@0xmisaka
It certainly has blocks. It's a DAG-based mempool, not unlike that of Avalanche, where every validator can "propose" blocks every round of the DAG, all of which get totally ordered by a consensus running on top of the DAG. See my recent thread:
This past month I got the chance to participate in
@wintermute_t
x
@encodeclub
's MEV hackathon.
@0xK1000
and I just submitted our project "MEV mitigation using modular DAG-based consensus protocols" (). Would love some criticism/feedback!
2
10
70
1
0
22
Saturday morning brain dump of web2 auth, smart contract wallets EIP-4337, zk auth, and eigenlayer AVS/operator registration.
What we want is **fine-grained**, **programmable** authorization and authentication.
1
2
23
Do you view PoW and PoS as sybil resistance mechanisms, or as economic incentive-alignment schemes? Or are those two sides of the same coin?
1
2
20
My team and I won USD$6700 at
@ETHGlobal
's hackathon last weekend. We built an accountability
@Superfluid_HQ
vault that helps you stay OnTrac with your most important tasks. Check it out.
1
4
19
@VitalikButerin
spending a lot of time on your blog lately. Would appreciate if you could fix all the broken links (pretty much all links) that are still pointing to your old domain.
5
1
18
Wait so... 37.4% of ethereum nodes are not even keeping up with the head of the chain?
6
4
17
Just realized, SSLE is JIT leader election, whereas Narwhal based consensus can retroactive leader election (see ). At the cost of slightly extra latency, you can get perfect (!?) leader DDoS prevention.
1/ MEV 🤝 DAG: As a part of the
@eigenlayer
research fellowship, I worked with
@samlafer
on analyzing the landscape of MEV on chains that use Narwhal DAG-based consensus mechanism ♾️🧵
Here is what we found 👇
3
20
125
1
0
17
Based preconfs are all the rage in my circles, but not sure how far out they've actually made it out. New FAQ with
@0xtrojan_
to help newcomers understand what's going on with them and why they are interesting. Check it out!
1/
@samlafer
and I have been looking 👀 into based preconfs and working on building an FAQ. 🧵 Huge thank you to
@drakefjustin
for his thorough review and feedback. 🫡
Here are the key takeaways👇
3
4
25
1
0
16
How is metamask swap getting away with charging users 0.85% for each trade, and still getting 120M volume in the last month...??
31
0
8
tendermint experts: how difficult would it be to fork tendermint and make it natively dual-staked, in that the state-machine would pass it 2 different votings powers (one per token) and tendermint would only finalize after reaching 66% threshold on both votings. (see modular…
2
0
15
Do any of the L2s implement what VB describes in as "Direct read-only L1 access inside the L2"?
This would funnily enough makes L2s an "L1-cache" for the L1 (which becomes an "L2-cache" in this scheme). Aka I could read from a contract in L2, and if…
2
0
15
+1. We don't need pBs, we need pBLGTQ+s. Builders that not only build, but also create merkle proofs for stateless clients that have sufficient bandwidth and demand low latency, verkle proofs for those that have low storage+bandwidth and don't mind waiting longer. And that also…
@IanSNorden
@txsequencer
Then go through and brick selfdestruct. Give ample warning and refund the 100k USD that is irrevocably lost from the EF funds.
PBS exists so might as well use it at this point, and this lets gives us verkle trees and stateless validation. But it doesn't need to be in-protocol.
3
1
7
4
0
14
Many thanks to
@KevinAftermath
for his deep knowledge and insightful discussions about Sui and MEV, and
@SashaSpiegelman
for his patience distilling Narwhal topics for us. Also super grateful to
@buffalu__
and
@ThogardPvP
for many discussions that helped shape our thoughts about…
1
2
13
So "pre-confirmations" can be either "pre-execution preconfs" (aka inclusion preconfs) or "post-execution preconfs". Both of these could be built on
@eigenlayer
.
Reminds me of the difference between shared sequencers (inclusion preconfs) vs fast finality layers (post-exec…
@CantelopePeel
apparently different things...
the term "pre-confirmations" (or "pre-confs", or "0-conf") originally started with Bitcoin as confirmation of Tx execution before the next block was mined
I use the term to describe a confirmations from the proposer that the Tx executed, including…
1
0
3
0
1
13
Lot of interesting parallels between ILs and async execution like monad and solana's endgame (), as well as to "lazy rollups" ().
Notice how eip-7547 deems an IL invalid "if the inclusion list transactions are not executable at the…
happy wednesday :-) sharing a short doc™ – "Inclusion lists: execution, consensus, & engine spec overview."
this doc tries to:
(i) convey the relative simplicity of the changes,
(ii) highlight the decision points of the design.
10
32
182
3
2
13
@0xdoug
Interesting that solana is going through a very similar period currently, with 98%+ of txs being failing bot txs. Current scheduler can basically be seen as randomized ordering (txs randomly sent to one of 4 parallel threads). New scheduler is supposed to make ordering actually…
2
1
12
This, imo, is the thing that will make or break the modular thesis
The real alignment problem is alignment of incentives across components of the modular stack
5
2
33
1
0
11
I'll also be at SBC with Soubhik and the gang from monday to friday. Looking forward to meeting lots of people and discussing
@eigenlayer
or anything else. Please say hi :)
AVS/middleware/infra teams headed to Stanford SBC next week, we’d love to chat
@eigenlayer
. If interested, send me a note to connect! I will be there 8/27-9/1.
2
4
32
0
0
12
New awesome use case for coprocessors on
@eigenlayer
just dropped.
I personally find it unfortunate that we’re stuck with this dual chain model, but if that’s what we have might as well make the best of it.
StakeNet is really cool!
Today, it lets
@jito_sol
redistribute SOL among validators based on their performance over the last n epochs. But it can be generalized meaningfully.
So why has no one done this before?
Ethereum has historically been the only chain where LSTs truly…
11
14
88
0
0
11
If given the choice, I would always bet on the heterogeneous architecture.
cosmos over polkadot
suave over anoma
eigenlayer avs over avax subnet
3
0
9
Career update: I was fired from
@BlockchainAG
due to bear market conditions 3 week ago. Decided to use the skills I had learned to give PvP liquidations a try. So currently looking to start or join a small MEV team. Please DM me if this sounds interesting, let's talk. :)
2
4
11
Personal commit-reveal experiment: I *commit* here to spending the next few days researching and understanding DAS and to *reveal* my "findings" by Sunday.
1
1
11
Narwhal is the culmination in a series of throughput improvement optimizations, that work by taking things off of consensus' critical path (hence increases latency)
(low lat, low tp) -------> (higher lat, higher tp)
single tx -> blocks -> hashed blocks -> Narwhal
1
2
10
Short thread on the history of MEV related consensus changes in Avalanchego. Do you know why
@avalancheavax
has a private mempool? Do you still think it's a leaderless protocol? This thread is for you. Thanks to
@wardbradt
for the inspiration:
2
1
10
"we're creating blockchains to prevent brokers from frontrunning user transactions"
...
"you probably want to use flashbots protect if you don't want the mempool sharks to frontrun your transactions"
oops
0
2
10
Ethereum stack v2023: CL-EL-EL
ConsensusLayer-EigenLayer-ExecutionLayer
1
0
10
Inspired by
@PatrickAlphaC
's multicall tutorial, I've decided to benchmark the different ways to batch evm calls, and found some interesting alternatives along the way (static multicalls!!). Started a repo if anyone is interested in contributing ()
2
0
10
Thought experiment: imagine a rollup with a two-tiered staked decentralized sequencer. A rollup node then needs to wait for 4 layers of confirmation.
It's layers all the way down. Maybe
@eigenlayer
should start talking about quadruple-staking
(image from…
3
0
10
@gakonst
What about ganache? :P
I've actually recently had to switch to ganache because anvil's RLP encoding for eth_getProof is broken so we can't generate risc0 fraud proofs.
This also made me realize the state dump/load functionality on ganache is actually much more advanced/involved…
2
0
5
The monolithic vs modular debate is getting repetitive, and kind of forgets one central aspect: once you've broken down your blockchain into "modules", how do you order them?
There are multiple blockchain pipeline philosophies. 🧵
(by pipeline I just mean ordering of stages, not…
4
1
9
New Distributed Systems reading group kickoff on May 7th. Come discuss the PACELC framework with us!
Am finally shipping this with
@samlafer
, our first session will be Tuesday May 7th at 7pm EST on . DM to join if you haven't already contacted, I will reach out to everyone who filled the form out already!
9
3
39
0
3
23
Sybil Resistance is the antithesis to Censorship resistance in some sense. Sybil Resistance protects the blockchain from the army of pseudoanonymous users. Censorship resistance protects every individual user against a (potentially colluding) blockchain lieutenant.
3
0
9
@DavidTiczon
@SuiNetwork
Except for execution layer experts it seems
@dubbel06
@0xMert_
As for BlockSTM, I am happy you like it. We like it too😻
You can see that it is widely adopted by the industry (e.g., Polygon and Sei).
Honestly, I think Sui chose static parallelism only because it is easier and they don't have in-house expertise for something like BlockSTM.
2
1
20
2
0
7
The major idea is to separate the mempool protocol (Narwhal) from the consensus protocol (Tusk, Bullshark, Hotstuff, Tendermint, etc), so as to achieve network speed throughput (200k+ on testnets), at the cost of slightly higher 2-3s latency.
1
0
8
Is running suave as a rollup on ethereum even possible? How would ethereum validators run the builder code if it needs to run inside that builder's sgx?
1
0
8
Question for the p2p experts: how much extra bandwidth does each extra discv5 topics and their related subnets cost to the avg eth node (both those that are participating in new subnet, and those that don't)?
Thinking what would happen if we were to bunch all the "alt-mempools"…
4
0
7
How do builders keep their edge on Suave if they have to reveal their builder source code to the world as an MEVM-solidity program?
1
0
8
I just contributed to the 🕯️ KZG Ceremony 🕯️ using 0xfb1…5d1a to help scale Ethereum
↓ Add your own randomness ↓
0
1
7
@Thyborg_
Cosmos is in a pivotal transition state right now, with Skip trying to lead it down the same path flashbots has lead Ethereum, and Jito is currently leading Solana. This. is. a. gigantic. mistake. There are SO many other designs we must explore before settling on MEV auctions...!
1
0
7
@zmanian
@cometbft
@iqlusioninc
"Each tx floods across the network being rebroadcast many times."
yikes.. why not use libp2p gossipsub like most blockchains? Are the params hard to get right?
2
0
7
polygon: hold my beer
people that think synchronous composability is a great feature have never tried building a system where the sender always blocks while waiting for the receiver to respond.
it does not scale, period.
asynchronous composability will become the default, the narrative will adjust
34
41
264
1
1
7
Probably stupid thought experiment: what happened if we admited price discovery never happens on blockchains, so we should operate them like ATS/dark pools, aka first tx in each block must be an oracle price update, and all trades in a block happen at that price?
3
0
8
One of the best talks I've seen so far on multi-proposer consensus protocols. TLDR: if we're already pipelining for efficiency and rotating leaders for fairness, why not pipeline while making everyone propose?
0
0
8
Should it be mandatory for core devs to disclose the protocols they have invested into?
2
0
8
Is there a good writeup somewhere on epochs in blockchains. Why is solana's epoch 2 days, cosmos chains' epoch 1 day, and ethereum's 6 min?
Once ethereum gets SSF via something in , do we still keep 32 slot epochs? If not, what happens?
2
3
6
@anomanetwork
do you have a diagram like
@CelestiaOrg
to add a "intent gossip layer" to the modular blockchain picture? Would be nice to compare to other mempool/private order flows solutions.
@cwgoes
@adrian_brink
@awasunyin
2
1
7
Block producers earn money from block rewards, tips, and MEV. Protocols like
@alkimiya_io
,
@OilerNetwork
allow hedging some of this volatility. Doesn't fee redistribution solve most of these and reduce hedging down to just token price volatility though?
2
1
6
The whole DAG-rider, Narwhal, Tusk, Bullshark saga of papers and implementations started as part of Facebook's Libra turned Diem project, and is now being integrated into major L1 projects such as
@Mysten_Labs
,
@AptosLabs
,
@sommfinance
,
@CeloOrg
, and
@anomanetwork
(afaik).
2
0
7
Interesting breakdown of trading on aggregators like 1inch, matcha, and paraswap vs trading on Rook, which redistributes MEV profits with the user instead of validators.
Ever wondered how a billion dollar wallet trades on-chain?
This thread follows one anonymous whale through a complex $110M trade, and boils everything down into strategies that can help you guys save on gas and get better execution.
🧵 🐳
81
368
2K
1
2
7
Is anyone "illuminating the dark forest" of sui or aptos? Would love to see some mev-explorer, or chat with someone who has analyzed data such as % of tx duplicates on narwhal blockchains.
1
0
6
Bring your idea, or not. We’ll build something out. have a question? we answer it. too many questions? we hop on a call. want to build the whole thing out? we’ll connect you to vcs. enjoy working with the team? We have full-time roles. this fellowship is the deal of a lifetime.
📣 Announcing the EigenLayer Research Fellowship (ERF)
The EigenLayer Research Team currently has bandwidth to work with 5 fellows deeply on a topic that they care about and interests us 🤝
13
42
350
2
1
7
From the PoV of the proposer, a mempool can be seen as a decentralized sentry node. Very useful for home validators, but a pretty lame one for people with proper infra. If you know what you're doing, share your ip address and have nodes forward to you directly.
1
0
6
And of course the creators of this beautiful theory
@alberto_sonnino
@SashaSpiegelman
@LefKok
@GDanezis
who also happily answered some questions on CT. Please be the ultimate judges of this worker architecture change 🙈🙈
1
0
6
Is it completely stupid to call merkle proofs SNARKs? They are a succinct (logn instead of n), non-interactive, proof of tx inclusion in a block, that reveal the tx (so SNARK and not zk-SNARK)... maybe?
1
0
6
My new blog post is out! I thought it was about time someone went through Vitalik's posts and made a graphical depiction of ethereum scaling solutions. Any ideas on how to improve this diagram?
0
0
6
So what IS a blockchain exactly? It's a ledger (read db) with certain properties. Eg. according to:
@IOHK_Charles
(): timestamped, immutable, and auditable
@sreeramkannan
(): censorship and reorg resistance, DA, and validity.
@llllvvuu
@MTorgin
@musalbas
@eigenlayer
There are four properties we want from a blockchain: censorship resistance, reorg resistance, DA and validity. A light node can a) sync headers to defend against reorgs, b) perform DAS for DA and c) ZK validity check to defend against the last three attacks.
2
2
7
1
0
7
priority fees on Solana and Arbitrum are used by exactly the opposite set of users.
Solana pfs are for users to jump ahead of the 98% of cheap spam txs by bots (spam is cheap)
Arbitrum pfs are for bots to jump ahead of the timeboost queue containing real user txs (spam is $$$)
3
0
7
Eth2 has a p2p network both at the consensus-layer (for blocks) and at the execution-layer (for txs aka mempool). Cosmos has a single shared p2p network for blocks and txs at the consensus layer (tendermint core). What are the pros and cons of both approaches?
2
0
7
@anomanetwork
and
@osmosiszone
are working on as an alternative, similar to what
@project_shutter
is doing on ethereum. No need to say, I would love to help if any of the above teams are hiring. :) Good day to you, frens.
1
1
7
Let's look at equity market microstructure and make analogies to what could happen in a FSS blockchain. These are just some preliminary thoughts following my reading of this blackrock report
3
1
6
1
0
7
@aeyakovenko
Eigenlayer inherits 3 properties from ethereum: economic, decentralized, and proposer inclusion trust ().
AVSs with objectively slashable faults can tolerate 100% collusion and still resist attacks up to the total slashable amount.
AVSs with subjective…
0
2
7
Could STARKs use verkle proofs instead of merkle proofs in their FRI protocol? Probably not what they want because they want to remain quantum-secure, but does it even make sense (just learned about verkle trees)?
2
0
7
zk provers currently have a ~1M blowup cost in computation efficiency
fraud proven approaches have a 100-1000x blowup (only way to reliably get fraud proofs is to incentivize a watcher network that replicates the computation)
Can zk ever bridge the gap?
2
2
7
The endgame for restaking layers like
@eigenlayer
and AVSs is complete smart-contract level reification () of the underlying p2p, consensus, DA layers.
@theyisun
talks about
@axiom_xyz
and coprocessors enabling introspection and reflection, but this is…
0
0
7
@0xMert_
@SashaSpiegelman
@wgrieskamp
@LoganJastremski
@EconiaLabs
agree. onchain routing for dex aggregators is the only application I've heard of that is hard (impossible?) on solana, that optimistic execution makes possible (don't need to specify dependencies ahead of time).
2
1
6
God daaaaamn solidity's module system is retarded. How can anyone work with this crap?
1
0
6
To address the communication complexity problem (probably will get shot by academics for using this word here but..), there are 4 approaches that I've seen so far:
- come to consensus on hash of txs instead of txs (approach pioneered by
@SeiNetwork
afaik?)
-
@SuiNetwork
and…
1
1
6
0
0
5
Accurate?
The tradeoff that PBFT protocols make for having 66% resistance to BFT attacks instead of 51% is that they can get censored (lose liveness) starting from 33%.
3
0
6
There are broadly-speaking 2 approaches to MEV: flashbot's/optimism style MEV auctions, and
@chainlink
/
@arbitrum
"fair sequencing" approach (threshold decryption + time-based ordering). Both approaches have virtues, but we chose FS since Narwhal's DAG is a natural fit for it.
1
0
6
@kostascrypto
I love how you actually tried to grey out my name but forgot 1 out of 3 places 😆
1
0
5
Prediction markets elicit (using a reward sys) and aggregate (using an algo) info (predictions) to produce a reliable forecast of *future* events.
Blockchains elicit (using inflation+fees) and aggregate (using a consensus protocol) votes to come to agreement on *past* events.
1
0
5
First tweet after adding "my tweets are my own": social consensus is overrated.
2
0
6
Can I please get off of BSC as soon as possible. Is it BNB now? Even Metamask if confused and doesn't show all of my assets. Can Binance just show me the chainID instead of a network name.. after their constant rebranding I have no idea what chain they're talking about.
16
1
1
Our question was: why stop the modularity at these 4 still pretty coarse layers, and not continue stepping-into? Can MEV hide in more fine-grained places?
1
0
6
@alinush407
@0xMert_
@b1ackd0g
Oh snap my bad! Mixed you up with
@alberto_sonnino
for the narwhal implementation, who is at mysten labs. Sorry for forgetting about you
@alberto_sonnino
:(
0
0
6
It just clicked for me. Avalanche’s p-chain is a bft zookeeper/etcd/consul. It is to subnets what etcd is to k8s services.
0
0
5
@SashaSpiegelman
@dubbel06
@0xMert_
Saying it's trivial is probably downplaying it..? I heard the scheduler is the hardest part and they're having a lot of problems with it, and rewriting huge parts of it:
My uneducated guess is this scheduling problem is similar to register allocation,…
3
0
6
My brother!
New speaker announcement!
@plafer2718
is a Rust Engineer at
@Informalinc
and contributes to ibc-rs, the Rust implementation of the IBC protocol.
Grab your tickets and mark your calendars to hear Philippe speak on April 28!
🎟️
2
6
11
1
0
6