Ole Hylland Spjeldnæs
@spjoleh
Followers
987
Following
326
Media
22
Statuses
319
Explore trending content on Musk Viewer
النصر
• 939615 Tweets
Washington Post
• 319409 Tweets
Durk
• 232551 Tweets
Virginia
• 200507 Tweets
Bezos
• 159844 Tweets
Beyoncé
• 130997 Tweets
WaPo
• 95554 Tweets
LA Times
• 83546 Tweets
Edwin Santos
• 70249 Tweets
Forest
• 69447 Tweets
Democracy Dies in Darkness
• 63120 Tweets
Boulos
• 57556 Tweets
Zendaya
• 51986 Tweets
#荒野7周年空前の超感謝祭
• 46365 Tweets
#史上初金車確定無料ガチャ
• 46286 Tweets
Marçal
• 44440 Tweets
Chicó
• 39435 Tweets
連載マンガ
• 37768 Tweets
#الاهلي_الاخدود
• 32187 Tweets
レジェンド車両
• 27549 Tweets
全員最大金枠アイテム5つ
• 27485 Tweets
Anitta
• 25088 Tweets
Tether
• 24245 Tweets
بن زكري
• 20373 Tweets
Amazon Prime
• 20113 Tweets
Leicester
• 19673 Tweets
Albon
• 17864 Tweets
SÃO PAULO IS COMING
• 16451 Tweets
Phil Lesh
• 16245 Tweets
ANIM NA TAONG SAMAHAN
• 13409 Tweets
DJ Clark Kent
• 12361 Tweets
Dudu
• 10043 Tweets
Pinned Tweet
delta's design is exceedingly simple. To get there, you can start with a rollup stack and just delete stuff:
- Remove the need for bridges and 3rd party interoperability services by having all the "rollups" (domains) share state
- Integrate the execution layer with the base
7
6
48
ASS Is All You Need
App-specific sequencing (ASS) refers to applications themselves ordering the transactions touching their state.
Appchains and app-specific rollups trivially have ASS, but the lack of global state introduces new problems
Let's do better 👇
8
20
130
Braid was recently introduced by
@MaxResnick1
to improve Ethereum’s censorship resistance by incorporating multiple concurrent proposers (MCP).
This thread will present a system which makes MCP a first-class citizen and solves all of its current problems.
4
8
89
I will write a post on why
@deltadotnetwork
is great etc later, but first I want to mention how proud I am of the team we have built and are building, including our Zurich-based engineering core led by
@iorulezz
, and the US-based
@riabhutoria
,
@MylesOneil
and
@0xFunk
Having
10
2
49
Hey what do you think will happen if we just take some consensus protocol and... remove stuff...
7
2
39
In this article
@riabhutoria
and I explore the global financial system of the (near) future, as some tbd very cool tech innovations get out there
1
6
35
pod and [redacted] announcing different consensusless systems roughly at the same time is the blockchain equivalent of humans in independent parts of the world all developing agriculture in a span of 1000-ish years
Entering a new age
Did you know that consensus is not needed to solve the double-spending problem?
15
13
96
2
4
28
10 years of intense research, and tendermint is still pretty much sota for byzantine agreement
here we introduce an approach to build permissionless systems which are not affected by the inherent limitations of consensus
physics
0
4
26
Related: Systems which scale by adding more machines are better than those that only scale by upping hardware reqs
e.g. rollups scale execution in this way
narwhal-style consensusless protocols scale data throughput by validators adding workers
near scales storage w partitioning
on rollups, L1s, interop, safety, censorship resistance:
rollups scale better than fast L1s because they can have a smaller replication factor than L1s.
less redundancy = more room to go faster due to less communication overhead between block producing nodes. verifying nodes
28
65
394
0
5
26
The three pillars of the future internet will be SNARKs, Byzantine reliable broadcast and CRDTs
All are newer than TCP/IP and have become feasible only in the last 5-10 years.
1
2
25
@alexisohanian
@sriramk
admins owning their subreddits as (potentially fractionalized) nfts which give them claim to some % of ad rev, which they share with the mods
Incentivizing good, unbiased moderation, rewarding good community members
2
0
19
I suspect that most large-scale distributed systems could be rewritten to not require global coordination, and that the world will move in this direction
CRDTs feel like permissionless consensus 10 years ago and zk 5 years ago
1
1
19
Thinking about Unichain, appchains, and the new Paradigm L2
Article coming early next week
2
0
19
How married is Solana to PoH?
Should be doable to pair strict access lists with something like Mysticeti and make use of the fast path
I guess it'd no longer be mempoolless but meh is that really such a big deal
@0xMert_
@aeyakovenko
3
0
18
Customizability 2
Finally, with only a few additions to our current construction, we can allow apps to use any programming language and fully customize their execution environment
We will leave the exploration of this to a future post
@repyhlabs
will soon go public 🫡
/end
1
0
18
We get trustless interop by having a fixed set of global laws which each participant must respect
This comes at the cost of sovereignty, and the trade-off between interop and sovereignty is defined by the scope of these global laws
The core concept of sovereignty vs interoperability is super simple from a human analogy: one can’t do stuff on a whim and also expect others to go along. There is a required compromise and well designed systems pre specify the interfaces ***ahead*** of time: this is what I agree
3
6
50
1
3
17
MCP -> ASS
We previously introduced a new high-level architecture which has MCP by default (see thread for details) — we clearly see that letting apps themselves form the proposers for txs touching their state achieves ASS
Braid was recently introduced by
@MaxResnick1
to improve Ethereum’s censorship resistance by incorporating multiple concurrent proposers (MCP).
This thread will present a system which makes MCP a first-class citizen and solves all of its current problems.
4
8
89
1
0
16
Big chain/small chain is a much better dichotomy than modular/monolithic
MegaETH and Solana should belong to the same category
Similarly it makes more sense to compare the Cosmos ecosystem to the Superchain than to think of individual appchains as monolithic blockchains
2
1
15
If zkVMs continue to improve at the rate of 2024 we'll have kappa=10 by christmas
2
1
13
The reports of seamless interoperability have been greatly exaggerated
So in order to have defi on Ethereum we need:
- mainnet for settlement layer
- optimism superchain
- uniswap chain
Can we maybe do it, idk, more elegantly?
73
9
391
2
1
13
Interoperability
Unlike rollups and appchains, which lack global state, apps now have strong (even synchronously atomic, see tweet below) interoperability with no intermediaries
Users can switch between apps near instantly, w/o bridges or third parties
@guiltygyoza
@_weidai
@0xRainandCoffee
They would basically coordinate on block building -- the idea is a rough generalization of "k-shared asset transfer has consensus number k" from this paper
1
0
3
1
0
13
Customizability
The functionality of an application such as an AMM largely depends on its sequencing rules. Giving the app itself control over sequencing opens up for new kinds of apps.
Apps now have a clear revenue source due to MEV
More context:
If you're suddenly interested in app-specific ordering because it now has a funny abbreviation (ASS), then you should check out some of the writing and presentations I've done. Will also list some other good stuff.
3
8
44
1
0
13
Slightly embarrassed to admit that I only realized just now that we should assume real-time proving for our system
3
0
12
what did he know
The end game of blockchain scaling is zk-execution + an analogous ~ O(1) da solution
@CelestiaOrg
‘s O(sqrt(n)) DAS is an excellent first step
0
0
5
0
1
11
Thinking about using PoW for an epochless blockchain(-like thing, no total ordering)
This is impossible to achieve with PoS, as you need to periodically reconfigure the validator set
Tbf epochless would make a lot of stuff really painful for us so might not be desirable
1
0
10
@yangl1996
This would be much easier for people to accept if you had a recursive SNARK that verified everything up to and including the current block.
The optimistic "since these txs are older than the challenge period we trust their validity" does not feel equivalent
4
1
10
Needs exploring: staking reward independence from amount staked
When rewards are proportional to stake, users delegate to the big validators
If they're independent, we get a diseconomy of scale: incentive is to delegate to smaller validators, thus stake dist ~ uniform
2
1
9
hey yeah i'm sorry my camera is off and there's german hip-hop in the background, i'm trying to work on my overhead press
0
0
8
The state changes introduced in this thread are beyond what Ethereum would or could feasibly implement in the foreseeable future, however at
@repyhlabs
we are building what can be, unburdened by what has been. Stay tuned
0
0
9
Scale
This design pushes every required ordering action to the apps, who together build the dependency DAG
Therefore the validators only need to run a super lightweight consensus protocol, allowing apps to consume bandwidth on demand as with AWS (more on this in the future)
2
0
9
@aaronjmars
@deltadotnetwork
I recommend this article that I wrote a while back which dives in deeper on marrying CRDTs with blockchain tech
It describes delta and how its properties differ from e.g. Ethereum
0
1
8
zk-TLS question: Is it considered impossible to limit the interaction between the server and the client/prover, with the latter just sending a proof to the verifier afterwards as in SNARKs?
This seeming need for MPC or similar is so annoying
4
1
8
The thing that no one seems to be able to do is making shared infrastructure as minimal as possible without sacrificing canonical (bridgeless) interop
1
0
8
Global ordering was a mistake
1
1
8
@0xMert_
@aeyakovenko
Mysticeti is Sui's new consensus, like Narwhal but mempool+consensus integrated iiuc
yeah it seems to be the main selling point of poh. my sense is that sui's stuff would be much better wrt network utilization and also solana's state model could take advantage of fast path
1
0
5
@kyled116
@aeyakovenko
@EliBenSasson
@ukolodny
Are there advantages to having any dapps/smart contracts/activity directly on the L1 as opposed to doing everything on zk-rollups?
1
0
8
The blockchains of the future could have plausibly been developed in a world without bitcoin
2
0
6
@MaxResnick1
This scheme relies on a deterministic ordering rule — consensus-based ordering would bound the total throughput of the k proposers, including duplicate txs, at the current level. In practice, the throughput of Ethereum would decrease relative to today.
1
0
7
@LefterisJP
There‘s an important distinction between social media cancelling and governance voting in that the former gives the pro-cancel group disproportionate power
voting him out is fine; trying to coerce his colleagues into forcing him out less so
1
0
6
On today's episode of o̶p̶t̶i̶m̶i̶z̶i̶n̶g̶ removing things that shouldn't exist: unproven state diff lists, and therefore also SDL reversals, domain slashing, general complexity
Slightly embarrassed to admit that I only realized just now that we should assume real-time proving for our system
3
0
12
0
0
7
Follow
@repyhlabs
as we continue toward our full announcement of what we’re building
Here is the full article:
1
0
7
@divine_economy
@ShivanshuMadan
@enzo_gte
@deltadotnetwork
@Infinity_VM
@riabhutoria
Conceptually I think it's reasonable to think of domains as supercharged appchains with better interop, customizability and scale, but yeah they're not independent state machines so it is inaccurate in the literal sense
2
0
6
SNARK aggregation via the bilinearity of pairings being less efficient than the brutalist "just make proving the verifier super fast" feels like a crime against god
0
0
6
I guess you can do something like, compute the true state root and accompanying proof for the block and send both to verifier who compares proven and original roots
But the zkVM can't compute an actual fraud proof
1/ Zeta X will be the first DeFi L2 on Solana.
Imagine the speed & convenience of your favorite CEX, fully on-chain. ZX is designed to elevate trading to the next level.
To achieve this, we leverage Sovereign SDK: the most flexible rollup framework. Built by
@sovereign_labs
.
🧵
31
45
340
5
0
5
The Glocal Web relies on three core technologies:
1. A shared state CRDT;
2. A minimal, leaderless, orderless consensus protocol;
3. SNARKs (not covered in this piece, stay tuned for future articles to see how they fit the rest of the puzzle).
1
0
6
Doing pure math is cool in part because you always have lots of really useful mental models in your RAM for when thinking about stuff
1
0
6
This is especially true in a space where every new feature or property is added by building directly on top of v1.0 and complexity is non-decreasing with every product iteration
1
0
6
The following two movements are very different reactions to the issues of the current internet:
1. Web3: global state with limited autonomy / customizability for apps
2. Local-first internet: no cloud dependency, also no global state
1
0
6
ASIC:VPU::custom circuit:zkVM
Today, we are thrilled to announce that
@1kxnetwork
have co-led a $33m Series A for
@FabricCrypto
, a company dedicated towards accelerating the adoption and proliferation of advanced & programmable cryptography, including zero-knowledge (ZK) and fully homomorphic encryption
11
15
183
1
0
6
Blockchains = global coordination only
Makes sovereignty impossible and adds massive unnecessary overhead to most interactions
Needed: minimal coordination infra for sovereign actors
So much MEV talk on the timeline…
Can we bring back sharding to the ethereum roadmap so we can have decentralization again?
Global consensus has been a fun experiment, but we’ve now understood that it is susceptible to capture.
Let’s look at MEV from first principles.
MEV
6
9
112
1
0
5
The end game of blockchain scaling is zk-execution + an analogous ~ O(1) da solution
@CelestiaOrg
‘s O(sqrt(n)) DAS is an excellent first step
0
0
5
The unbundling of global and local is the next step of permisssionless systems
we've been witnessing a long march of unbundling of the global consensus
- unbundling of global execution to local execution with rollups, driven by scaling
- unbundling of global ordering to local ordering, with all private order flow solutions, driven by MEV management
I
1
4
24
1
0
5
@EricRWeinstein
@elonmusk
Imagine looking at the dominance of the private over the public sector in space tech and be mad at the former
0
0
5
The European urge to wash down Taco Bell with one of those Whole Foods probiotic sodas
0
0
5
Is the zero-knowledge property not much stronger than necessary for most (all?) privacy applications?
cc
@Zac_Aztec
@_weidai
@kobigurk
3
0
5
Thinking more about the Solana vs Sui approach to data prop
Solana: mempool in principle O(1) in validator count by maximizing leader-dependence, decrease load on leader w turbine
Sui: make protocol maximally uniform to utilize resources optimally
opposite ends of the spectrum
0
0
5
We will now introduce a state model which solves the three issues above and is optimized for MCP both for cr and scaling: Although state is still global, users can now send their transactions to only one proposer.
This user→proposer mapping is part of the global state.
4
0
5
1
0
5
Wei is a phenomenal thought partner for early-stage technical founders, speaking from personal experience
4/ If you are working on something that fits that description--whether it's rigorous academic research, novel application of existing tech, or even moonshot ideas that simply compel you--please don't hesitate to reach out. DM's open.
7
1
21
1
0
5
@ayyyeandy
How do you see this playing out for rollups? Afaict not a single interop stack really satisfies your requirements
3
0
4
In bed realizing that we can likely kill a component which has taken months of engineering work to build
Unironically no better feeling 😃
0
0
4
Today, the state of the web is relies on centralized clouds: data and compute are owned and controlled by a few large providers (AWS, Azure).
This is clearly bad for many reasons, I won’t elaborate further here
1
0
4
Like good luck getting to the Raptor 3 solely by making changes to the 2 while it is continuously in use
1
0
4
@MaxResnick1
has mentioned two further problems with Braid:
1. The effects of a transaction can be known only after the full transaction set has been constructed.
2. The same tx can be included by multiple proposers, preventing linear throughput increase.
1
0
4
Integrated systems are nice in part because we can create disincentives for parasitic behavior by anyone
It's funny to me when ppl say "L2s are parasitic to the L1"
imo it's the other way around. A lot of L1s are parasitic to L2s, trying to trap them into their "all value must flow to the base layer" mindset
The base layer is there to enable builders, not extract from them
12
6
65
0
0
4
More interestingly, no additional ordering is needed: Everyone with the same set of blocks has the same dependency DAG, and this is sufficient to determine the canonical state. This is very similar to Narwhal
1
0
4
We had Galois theory as part of our second year abstract algebra course and pretty much every class I was just like holy shit
Peaked beautifully with the fundamental theorem
If you put the rationals and the m-th roots of unity together, then quotient by the rationals, you get the group of the m-th roots of unity
Math is fun. Ask a toddler on the street.
4
3
46
1
0
4
Many thanks to
@AlanaDLevin
@different_mj
@Rob_Al_Ghul
@jessewldn
@lsukernik
and the
@repyhlabs
team for excellent reviews and suggestions!
1
0
4
@MaxResnick1
In Braid, validators run k copies of the chain, k proposers each propose a block, and an unordered transaction set is formed as the union of all transactions contained in all the k blocks.
1
0
4
Happy to see Sui usage going up, feels like good tech being rewarded
0
0
4
Thinking about Voltaire's "I disagree with what you're saying, but will defend with my life your right to say it" toward the end of the French monarchy/the beginning of European democracy
Something something full circle
0
0
4
Have to assume GPU, but still hugely impressive and close to the numbers
@IrreducibleHW
posted for their FPGA in April
Hopefully the conclusion is not that GPU perf approaches custom hardware perf
1
0
6
@jon_charb
The do everything on bitcoin to enable net new interesting things at the cost of scalability approach is cool
The let's use bitcoin as a gimmick (settlement layer) with external da and be like any other rollup approach does not interest me much
0
0
4
@divine_economy
@jon_charb
@deltadotnetwork
I mean validators are using their CPUs when verifying proofs, is that sufficient to call it execution?
Probs makes more sense on ethereum where zkr settlement means that validators actually execute smart contract code
0
0
3
2
0
3
@_weidai
@guiltygyoza
Can be sync+atomic if the proposers coordinate (with no dependency on external protocols) s/o
@0xRainandCoffee
1
0
3
@MaxResnick1
Arguably the main issue is the need for a synchronization/finality gadget, which introduces timing games between the proposers — seeing others’ blocks before proposing your own is advantageous.
1
0
3
We believe taking the best of the two — what we’re calling the Glocal Web — achieves the desirable properties of both, and more (such as throughput going from being global to local, allowing apps to scale as needed)
1
0
3
Benefits:
1. Customizability, flexibility, and value capture for each app;
2. Secure, atomic composability without external deps;
3. Web2-like on-demand scale;
4. Sovereign, censorship-resistant users.
2
0
3
SNARK -> SNARG: Do we need knowledge soundness?
This is meant purely wrt. proving validity of program execution, say zk-EVM.
The K in SNARK is there to make sure that the prover has a valid witness. In our case, that they correctly executed some list of transactions.
Really,
0
0
3
Thank you
@megaeth_labs
for onboarding the next billion people to Rahul Pandharipande's monstrous moonshine lectures 🙏
2
0
3
@guiltygyoza
@_weidai
@0xRainandCoffee
They would basically coordinate on block building -- the idea is a rough generalization of "k-shared asset transfer has consensus number k" from this paper
1
0
3