DSS 2026 / Day 2 / Amber Hall
Valid Signatures, Fragile Accounts: PQC on Ethereum
Magnus Woodgate
Source: YouTube automatic caption track · Exported 2026-11-16. Timestamps link to the source video. Cue end times were not supplied.
Transcript
0:02Hello everyone.
0:06>> Cool. So here to talk today about some
0:09postquantum crypto uh specifically
0:11Ethereum's road map and some of the
0:13stuff involved there. Um yeah, I want to
0:15highlight that we're going to cover like
0:16a lot of different stuff. So this is
0:17kind of like a lay of the land rather
0:19than a deep dive into any specific
0:20topic. Um although we'll get into it. Uh
0:22and yeah, the focus here is security
0:24practitioners. So, um, not necessarily
0:26cryptographers, but you guys, auditors,
0:28L1 devs, core devs, that sort of thing.
0:31Cool. So, I'm Magnus Gate. I lead one of
0:33the security engineering teams over at
0:34Asymmetric. Uh, we do a ton of different
0:36stuff. Uh, I focus on kind of the white
0:38hat research and continuous security.
0:40Uh, yeah, Asymmetric, for those of you
0:43who don't know, I think a few people
0:44here do. Um, we work on a ton of
0:47different stuff, not just web 3. Uh, and
0:49we're included with the like embedded in
0:51the protocol teams from the beginning
0:52ideally. uh and we we don't do audits,
0:54we do continuous security, but you can
0:56talk to me about that and check us out
0:57afterwards. Cool. So the you know today,
1:02like I said, I want to talk about
1:03Ethereum's road map for postquantum
1:05migrations and specifically security
1:07practitioners. Um we're going to cover a
1:09lot of different stuff. So the aims here
1:11is that you know some of the terms, some
1:13of the terminologies that you can go
1:14away with and take a look and start
1:15digging.
1:17So brief outline of the talk. Um we're
1:20going to cover some building blocks and
1:21background and I want to focus here on
1:23the implementation risks a brief note on
1:24assurance and limitations but we'll get
1:26into that. So the background one of the
1:30things I want to start with is is
1:31account abstraction specifically. Um so
1:34the question you know that we're going
1:36to be talking about today the most is
1:37the valid signature part that you can
1:38see highlighted here in yellow. Uh the
1:40question is is the state safe to sign?
1:43Do the bytes match? I mean these are all
1:45things that you see in classical crypto
1:46today that we have. Um, but there's some
1:48other parts that this kind of enables
1:50that you might see an in ERC 4337
1:52mentioned there about the is the policy
1:54intended. Um, you know, can this account
1:57or user actually do the thing that you
1:59want it to do? And then does the account
2:00behavior actually behave as intended?
2:03So, you know, does the thing change the
2:05state that you want it to?
2:07So there's obviously a lot that changes,
2:10but I think I want to focus here on the
2:11stuff that stays the same because it's
2:12it's it's, you know, the most relevant
2:14in terms of when you're reviewing this
2:16from a security perspective. So the
2:18existing obligations do not change. You
2:20still need to worry about intent, replay
2:21protection and whatnot. Um access
2:23control, all of that stuff. Uh the
2:25access policies say the same. The thing
2:27that does change is the signature
2:28specific work. Um specifically around
2:30limited use state, but we'll get into
2:32that. Uh and then I want to draw like
2:34last attention to this last point about
2:36uh retiring authorities because that's
2:38going to be a big one uh when it comes
2:40up and there's some EIPs around that.
2:42But yeah, there's a lot of overlapping
2:43risks, but the things that are probably
2:45going to be interesting for you guys in
2:46an operational capacity is what you need
2:48to recheck. So brief reminder here, um
2:52we're not going to go too much into
2:54this, but effectively Shaw's algorithm
2:56lets you efficiently factoriize the
2:59discrete logarithm problem and you can
3:01recover the private key. uh given a
3:03public key point on the elliptical
3:04curve. Um yet you can take a look into
3:07this, but this is kind of why we're
3:09looking to do the postquantum migration
3:10today. Um not there yet in theory.
3:14So yeah, some caveat here. Those of you
3:17who already know, um the public key
3:20isn't actually exposed until something
3:22is signed. So that doesn't mean
3:23necessarily broadcast the Ethereum
3:25network. Um but that could also be an
3:27off-chain signature stored in like a web
3:292 DB. Um, and it's important to be aware
3:31of that because that DB could be leaked,
3:33for example, and you're still
3:34vulnerable. Um, the public key, if it
3:37it's not actually exposed until
3:39something's signed, um, which creates a
3:41slightly different problem. Um, it
3:42doesn't mean it's not exposed, but it
3:44requires an additional layer, uh, to
3:46shores when you're breaking it with a
3:48quantum computer. Um, but this is, yeah,
3:50it doesn't mean you're safe, but there's
3:51a slight extra layer of protection
3:53there. So, you know, how do we kind of
3:57combat this in the long term? uh the
3:59programmable authentication here um
4:02enables the postquantum migration and I
4:04guess the point here is that Ethereum's
4:06road map is very generic uh and this
4:08basically allows us to pick different
4:10algorithms. You can see here ERC 4337
4:13that you guys may be familiar with uh
4:15basically allows a user to do some
4:17intent or operation with agnostic
4:19account validation which may be
4:21different to what the L1 actually
4:22provides today and results in some
4:24execution. Those of you who haven't
4:26looked at the EIP8141, which is
4:28currently uh scheduled for native
4:30frames, basically starts to bake this
4:32intention into the L1 and then we can
4:34start to do programmable validation at
4:36the L1 layer. Uh essentially allowing us
4:38to switch out the cryptographic
4:39algorithm for signing and verification.
4:41Um there's also some other parts around
4:43this in terms of you can choose who pays
4:44for gas. Uh but at the end of the day,
4:46you still get the same result of
4:48execution, but this enable enables uh
4:50programmable validation.
4:52So yeah, back to 4337, we have a similar
4:56structure here where we have, you know,
4:58is the execution authorized? Um, and
5:00that's what we want to focus on today.
5:02But then there's also the aspect of
5:03payment, which is something you want to
5:04be considering. Um, I recommend checking
5:06out 8141 if you haven't yet.
5:11Cool. Resuming from there. Thank you
5:13guys. Um, yeah, I want to focus on
5:16existing migration risks. So this is
5:18something that you know you guys should
5:19be aware of as we start to migrate
5:21towards these approaches. Um again
5:22highly experimental but you know I want
5:24to give some takeaways that you can
5:25focus on uh more generally. So this is
5:28one interesting edge case that maybe you
5:30know is being discussed but maybe you
5:32haven't considered yet around 772
5:34delegations. So let's say that we have
5:35an account or a wallet in this case um
5:38with a postquantum signature scheme
5:39already in place. You may have delegated
5:41that wallet beforehand and that
5:43delegation is controlled by an ECDSA
5:44key. So even though you have this
5:46postquantum uh signature in place, you
5:49are not safe. So you know all it takes
5:51is u the recovering of the private key
5:54of the ECDSA key and yeah the
5:58interesting thing here is like how are
5:59you going to deal with this as a
6:00security practitioner? So one of the
6:02approaches you might think is okay well
6:04I'm just going to move every account uh
6:06over. So I'm going to move all my
6:07assets, all my positions, reassign all
6:09the roles and update all the
6:10integrations. Maybe that works for you.
6:12it might not work for everyone if you
6:13have these big monolithic applications
6:14and whatnot. Um, and I'm assume that
6:17some some of you guys are want to going
6:18to keep the same uh addresses and things
6:20in place. Now, you have to be super
6:22careful around the relationships uh and
6:24authority there and importantly retiring
6:26the old key. Um, so you want to be
6:28thinking about specifically what can the
6:29old key do? Can it do transaction
6:31delegation? Can it do upgrades and
6:33recovery? Could it overwrite all the
6:34code in your protocol? Uh, effectively
6:36you want to be checking every single
6:37entry point and this is a serious
6:39migration risk. Um, yeah,
6:42cool. So, I'm going to skip over a bunch
6:44of stuff very quickly, um, so that you
6:46guys basically get the the building
6:47blocks for this. Um, postquantum
6:50signatures, if you guys haven't seen the
6:51the three big ones that you see people
6:53talk about, MLDDSA, FNDSA, and SHDSA.
6:56Uh, today we're going to be talking
6:57about SLHDSA or hashbased cryptography.
7:00Uh, specifically some of the uh,
7:02Ethereum proposals. Um, although FNDSA
7:05and MLDDSA you'll see on many other
7:06chains. uh or but yeah
7:10to be be super clear here Ethereum
7:12doesn't have a universal postquantum
7:14signature mandate. So they're not going
7:16to you know fully commit to one if
7:18something comes out that you know if for
7:19example if SLHDSA gets broken uh through
7:21some new quantum algorithm then the
7:23benefit of programmable uh account
7:25validation is that we can change it
7:27whenever we want. Cool. So start
7:30starting with what's the winter's
7:31onetime signature scheme uh plus and
7:33plus c you may see a compressed variance
7:35and more secure versions effectively
7:37it's it's a a hash chain that provides a
7:40onetime signature uh this is a building
7:42block for SLXSA that we'll go into XMSS
7:46uh very easy to say extended Merkel
7:48signature scheme basically builds a
7:50Merkel tree with watts keys as the root
7:52here uh and yeah you can basically do
7:55use this to combine trees where you have
7:57an XMSSS tree a watts key that signs the
7:59root of another tree and then that tree
8:01is then basically bound and you can
8:02traverse the whole path.
8:04Falls the forest of random subsets is
8:06one of the final building blocks and
8:08basically sits at the bottom and these
8:11uh where the the the leaves are
8:12essentially secrets and you can reveal
8:14them bit by bit um and then they
8:16consider use but we'll go we'll go into
8:18this.
8:20Cool. So the I spoke about winets
8:22onetime signatures uh fors actually has
8:24a few time signature properties. So if
8:26you do double sign uh if you do reveal
8:29multiple parts about the tree uh then
8:31the it's not immediately game over
8:33unlike watts uh effectively you can see
8:35the bit strength decreases and this is
8:37something you want to be careful with uh
8:39as as you're you know reviewing this
8:41from a security perspective did
8:42something get signed twice and we'll go
8:44into some of the instance and edge cases
8:45around that but it's not game over
8:47immediately unlike with watts cool so
8:50piecing those all together you can take
8:51a look at how SLHDSA works at a high
8:53level you have the fours uh forest
8:55random subsets at the bottom. You have
8:57it connected with a W signature to the
8:59fors route uh to some bottom XMSSS layer
9:02uh and it goes up to an upper XM XSS
9:05layer uh all the way to the trusted
9:06public route. So you can think of like
9:08this massive tree structure that's
9:10really complicated and you can choose
9:12the number of layers and falls
9:16jardin which is you know you guys might
9:18have seen uh one of the proposals here
9:19is to take ethereum's modifi modified
9:22version of SHDSA with a different hash
9:24function uh based off catch to be
9:26optimized for EVM um also known as in
9:29the sphinx minus family and you
9:31basically get this two trade-offs so you
9:33have the stateless spx which is the
9:34hypertree structure you just saw for
9:36SHCSA And then you have compact jardin
9:38signing. Um the idea is that you know
9:41the stateless one you can always fall
9:42back to if you need it but you want to
9:44use the jardin one for dayto-day because
9:46the the proof is significantly smaller
9:47and the signing is significantly faster.
9:50The way it works is basically you use
9:52this hypertree structure to sign uh like
9:55fours instances. So then you basically
9:57have like a pool of signing instances
9:59and then you use you basically every
10:02time you use one of the instances you
10:03you use it and throw it away. So because
10:05of this property of fours uh you can
10:08sign you can reveal more information
10:10about it uh but then the bit strength
10:11lowers and you keep a track of this with
10:14you know in this case q= 1= 2= 3 and so
10:16on um and once you've used the queue you
10:19need you need to get rid of it one of
10:20the experiments here you can actually
10:22see an onchain implementation um with
10:254337 and it combines the ECDSA and spx
10:28so you actually get both so uh in this
10:30case a quantum computer we need to get
10:32to solve the discrete logarithm problem
10:34and you'd have to find an SPX
10:35implementation breach or or
10:36cryptographic break.
10:39Cool. So, the intended life cycle here
10:40is that Q selects a fors instance under
10:42the root. Um, you reserve it, you
10:44persist it, and then you sign. You might
10:46go, well, why do you need to do this?
10:48Uh, it's very similar with atomic
10:49signing. You don't want to double sign.
10:51Uh, so you want to make sure that
10:53instance Q is always persisted even if
10:56you didn't actually broadcast it. If you
10:57think there's a chance that it was
10:58broadcast anywhere, uh, you never want
11:00to rewind Q, otherwise you risk losing
11:02some of the bit strength there.
11:04And to put this in perspective, uh the
11:07current implementation I believe took
11:0847.3 seconds on uh spx stateless signing
11:12on a legend. I know S+. So these
11:14signatures are massive. You don't want
11:16to be using the stateless path uh
11:18unnecessarily. Um yeah, I don't know who
11:20wants to wait 47 seconds for signing,
11:22but hopefully that speeds up. Cool. So
11:24the bulk of the talk uh is
11:26implementation risk. So what do security
11:27practitioners care about now that you
11:29understand some of the basic blocks as
11:30these things roll out?
11:33I've mentioned this already, but the
11:34existing uh obligations still apply. So
11:37if you have account author like
11:39authorization, you know, you have roles
11:41uh per dexes have keepers and things
11:43like that, you want to be checking the
11:44authorization route. How do you verify?
11:46Uh is that route still active? Can some
11:49key somewhere control it? And it might
11:50be in super non-obvious ways. 7702
11:53delegations, permit 2's leaked,
11:55offchain, uh and and and the like. Um
11:58but yeah, all of the existing
11:59obligations still apply.
12:02And the thing I want to emphasize here
12:03is all of the signature families have
12:05their own invariance. Here's some of the
12:06ones you probably already checked for
12:07ECSA and BLS. Uh hashbased, you know, is
12:10is is no different. Um in this case,
12:12we're talking about the proof passing
12:14and and and whether it's stateless or
12:16not. Latis cryptography as well has its
12:18own uh parameters. So you want to be
12:20checking this with whatever you're
12:21working with postquantum or pre-quantum
12:23crypto.
12:25And the thing to that I want to
12:28emphasize around the stateful signing.
12:31So using this Q instance that I spoke
12:33about is that just because the proof
12:35verifies doesn't show that the key was
12:36unused. So you could sign multiple times
12:39where Q equals 1 and it's going to leak
12:41key material and basically make it
12:43potentially easier uh to to forge some
12:46kind of proof. Um the wallet must
12:48preserve the signing history. In that
12:49case, that's the value Q.
12:52So, here's one instance that you may be
12:55thought of with pre uh quantum crypto.
12:57Basically, you have some, you know, uh a
13:00wallet on two devices, device A and
13:01device B. Device A signs with Q= 7. It
13:04has a a crash or a power outage. So,
13:07then you and you can't recover it. So,
13:08you you clone it from device B or an
13:11attacker does uh and you sign with Q7
13:14again. So now you've got two signatures
13:15with Q equ= 7 and you've leaked more
13:17information about uh the four's uh trees
13:20than you wanted to.
13:23So what an attacker would do is they
13:24would sit there collecting all these
13:25openings uh and then basically they
13:27would start to do random offline
13:29searches where they would verify the
13:30randomizer um and they effectively need
13:33to look for some proof that you know
13:36does an action that they want to do. So
13:37that might be bypassing the frame uh
13:39validation layer to do a transfer or
13:41whatever. Uh this is why one of the
13:42reasons that if you are doing a
13:44migration today then maybe the hybrid
13:45ECDSA signature requiring both is a good
13:48approach but it's obviously going to
13:49cost more in gas.
13:52So moving on here's like a hypothetical
13:54other scenario that you want to be
13:56testing for. Um this is you know similar
13:58to classical cryptography. You basically
14:00want to be checking that the you know
14:03mutating you you want to be checking for
14:05failed close cases specifically. So in
14:07this case you want to be checking Q
14:09specifically. uh if you do a roundtrip
14:11signing test, so sign and then verify in
14:13your unit test for your code or
14:14whatever, you want to be checking that
14:16if you do vary Q um or or remove Q uh
14:19that things fail closed instead of uh
14:22silently succeeding.
14:25Similarly, you know, I think it's
14:27happened before, but in in just general
14:28Merkel uh tree implementations, you want
14:30to be checking that the proof actually
14:33verifies the message or authenticates
14:34membership of the message. you want to
14:36be doing the full walk all the way up
14:38the tree whatever you're doing stateless
14:39or stateful um and and checking that
14:42that works. So effectively the thing I
14:44want to highlight here is you want to be
14:45mutating uh inputs and checking that
14:47everything fails closed under all
14:48circumstances
14:51and the scenario here is basically an a
14:53attacker supplied root root um and yeah
14:56if you don't check the full path then
14:58you may uh be vulnerable to this. It's
15:01very similar to existing Merkel tree
15:03implementations.
15:05Cool. And this is similar. You want want
15:08to be thinking about the this mutation
15:10approach when you're looking at frame
15:11transactions as well. Uh there's a
15:13concept where you you could accidentally
15:15approve a frame that you
15:18didn't mean to approve. Um and you want
15:20to be mutating and and basically
15:21checking that you know later frames um
15:24don't don't have approvals that they
15:26they weren't authorized. Um just
15:28something to think about when you're
15:28looking at EIP 8141.
15:32Cool. So, we spoke about uh different
15:34things in your uh attacker model, mainly
15:37application logic that you'd be looking
15:38at. This one is a very interesting paper
15:40I thought I'd mention uh around the guys
15:42from Lurbec and Google. They did uh a
15:45row hammer attack uh for those of you
15:46who aren't familiar. It's where you take
15:48um an attackers controlled memory region
15:50and a victim controlled memory region
15:51and you just do a ton of writes to it to
15:53try flip the bits at a hardware level.
15:55Um, effectively they proved that this
15:57would work in OpenSSL's implementation
15:59of SLHDSA. Um, they called it slashdsa
16:03where they would basically target the
16:05the watts key signing. So this part when
16:08you register another xss tree or falls
16:10uh in the in the hyper tree and
16:13basically that would mean that the watts
16:15key would sign a different tree. Um and
16:17then maybe you can if you know engineer
16:19this so that it deterministically sign
16:21some attacker control tree or or a
16:23subset of it which would allow uh an
16:25attacker forge proof. Uh the thing is is
16:28is really like is this in your threat
16:29model. So they reported this to OpenSSL
16:31and OpenSSL said this isn't in our
16:33threat model. So um as a [clears throat]
16:35security researcher you want to be
16:36checking what what your application you
16:38know cares about. Is it remote signing?
16:40Is there RC access? Are you running in
16:41in enclaves and whatnot?
16:44And the way this works is that basically
16:47if the Watts instant get compromised
16:49similar to the the one-time signing that
16:51I mentioned before, uh you can sign
16:53something a route of a tree that's
16:55untrusted and then an attacker can
16:56provide that tree um and then initiate
16:59some action similar to you know fund
17:01transfer or or delegation or whatnot. Um
17:03and that works for XSS all fours.
17:08Cool. So the you know moving on
17:13basically under this scenario you want
17:15to be checking that recovery always
17:17preserves the security policy. So one
17:20thing to talk about here is that we
17:22spoke about the compact signing and
17:24stateless signing. Basically you want to
17:27check that the stateless signing always
17:28works. And you know this this may be
17:30because you've lost the device. You
17:32don't know the the value of Q. um the
17:34comp the keys being compromised and you
17:36you want to rotate um or you have some
17:38ECDSA key uh and and basically you want
17:40to always be checking that the intended
17:42policy matches uh what your organization
17:44wants it to do.
17:46One of the reasons I mentioned this is
17:47because for example the SPX verification
17:50uh gas implementation today I believe is
17:522676,000
17:55um whereas the draft frame transaction
17:57uh is only going to allow 100,000 gas uh
18:00for validation. So as a security
18:02researcher you might be thinking well if
18:05it takes 276,000 gas how am I going to
18:07fit this into a frame transaction. Uh
18:09this is all up for discussion still so
18:11you can or contribute today. The
18:14question here is more about the
18:15integration challenge but you might be
18:16thinking okay how how can I exploit this
18:18you know as as a as a red team. So you
18:22want to be thinking about bridges you
18:23want to be thinking about uh external
18:24edge cases. Okay. And you know, assume
18:27that they've uh an institution finds out
18:29that they've signed three times with the
18:31same queue due to some faulty um
18:33operating system or hardware. You now
18:35immediately want to do an SPX stateless
18:37transaction to register a new route uh
18:39or send to a a new account. But you the
18:41question is can you fit it in the relay
18:43or or pay master system that you want?
18:45Um and the reason I'm bringing this up
18:48is you want to be forcing the hard path
18:50uh and the and the easy path. So you
18:52might be using a stateless state full
18:54signing uh for day-to-day, but you and
18:57and stateless signing may come up
18:59relatively rarely. Imagine like an ETF
19:01that doesn't move their funds uh at all.
19:03You want to be checking that the SPX
19:05stateless path does does always work.
19:06Maybe you want to run periodic SIMs and
19:08whatnot. Um and and checking that under
19:10all current relay conditions, maybe
19:12you're doing crosschain bridging or
19:13something that this is going to work.
19:17Cool. Um I wanted to mention here some
19:20briefly about assurance and some of the
19:22limitations around that as well as some
19:24of the art in the space. Uh the and and
19:27yeah just so that you guys are aware of
19:28what you can take a look at. There's a
19:31really cool implementation in Jasmine.
19:33Um those of you who aren't familiar with
19:34Jasmine lets you do uh low-level
19:36cryptographic implementations uh in
19:38assembly and it has constant time
19:40checkers so you can check that the
19:42implementation is constant time um and
19:45you can uh with formal verification. So
19:47the implementation here uh actually has
19:49an XMSS an XMSS MT so that's multi-ree
19:53um and you can prove that it links to
19:55the security theorem. So this is like an
19:57active area of research uh also in in
20:00the web 2 space uh just classical cryp
20:02cryptography um but these these models
20:05are available so you can start comparing
20:06them. So if you're implementing you know
20:08a new GU fork or something um with a
20:11custom XMSS implementation then you can
20:13be uh diff fuzzing and and you know
20:15checking against some of these
20:16implementations.
20:18Similarly um they actually used as well
20:20before but uh there's also a sphinx plus
20:22model that's type bound checked in easy.
20:24This is more for the model itself but uh
20:27you know if you are changing your uh
20:29tree structure and things like that like
20:31the number of layers and and whatnot um
20:33you can actually test that model in easy
20:34crypt and formally verify that as well.
20:38Cool. So, as a security practitioner,
20:41what what are you looking at? I think um
20:42formal verification always has a place.
20:44Um and so does some of the other tools.
20:46You want to be layering these wherever
20:47you can. The proof boundary definitely
20:49becomes the review boundary. So the you
20:52know, as much as this as it's great at
20:54verifying some of the implementation,
20:55you want to be checking the full stack,
20:57especially with this uh stateful
20:59stateless risk. So you want to be
21:00checking do the inputs where do the
21:02inputs come from? What are the edge case
21:04scenarios that you maybe never thought
21:05about with weird race conditions or
21:07hardware corruption? Um, and is that
21:09modeled by formal verification? If it's
21:11not, can you can you model it some other
21:13way through fuzzing or even hardware
21:15testing and things like that?
21:18Cool. So, I want to leave you guys with
21:19some key takeaways from this talk. Um,
21:22the most important one during the
21:24migration, I think, is that the just
21:26because we have postquantum crypto, it
21:28does not retire the old authority. So
21:30you need to be checking ECDSA keys and
21:32and what they control at the L1 layer.
21:34Uh um and yet you may have leaked
21:36signatures that you you are not aware
21:37of. Maybe you want to you know some
21:39teams here want to build some automation
21:41around tracing all these roots and that
21:43recovery always works. You can always
21:44upgrade your protocol and but you can
21:46never do it with the old keys.
21:49The compact proofs um we spoke about
21:51Jardin. There are also some other ones
21:53on the uh sphinx flam family um that
21:56also stateful. they do depend on durable
21:58signing history. So that that value Q
22:00that we spoke about, you need to be
22:02keeping a track of that and making sure
22:03that uh is very closely um you know
22:07durable and used. Um and you want to be
22:10testing all of the edge cases. I mean
22:11I've just mentioned a few here, but
22:13always be looking at whether you're
22:15leaking key material, which would be
22:16super important if you're regarding
22:17billions of TVL.
22:19And finally, you want to check that
22:21obviously everything uh agrees. This is
22:23more at the integration layer. Does the
22:25renewal and recovery work through real
22:27relay mechanisms which may be bridging,
22:29maybe gas relays, maybe pay masters? Um
22:32maybe as some EVM fork works differently
22:34to another one and that your spx uh
22:36stateless one, you know, the gas might
22:38be too expensive for you to actually
22:39include it in any real block. Um but
22:41we'll see where we end up. And to
22:44emphasize this, we're still super early.
22:45So these are just some things to be
22:47thinking about and some keywords you
22:48guys can go and Google afterwards. Um
22:50but yeah, a lot of this stuff is
22:52scheduled, not deployed. A lot of it's
22:53in prototypes. Um, and yeah, the
22:56engineering effort really remains here.
22:58So, uh, yeah, if you're doing wallets,
23:00L1's, L2s, um, application layer, you
23:02want to be starting to have a think
23:04about this, uh, today.
23:07So, thank you for listening. Um, yeah,
23:09you can check out asymmetric.com on the
23:11QR code. Um, and yeah, any questions?
23:16[applause]
23:30Cool. Thanks everyone.
Source and accuracy details
- Names, organizations, technical terms, code and numbers may be misrecognized.
- Source wording and disfluencies are preserved; no editorial corrections were applied.
- Export contains whole-second cue starts only; cue ends and subsecond timing are unavailable.
- A full caption-track export does not establish that every spoken word is represented.
- Automatic YouTube captions; not human-verified against the audio.
Source export SHA256: 393ab7ae51ccda5fe49e61862daa0a53c8008ce86cc3ae1d8aba5693953d9ec1