Internet Computer (ICP)

1/30) ICP has a terrible design: Insecure, low capacity & highly centralized Worst of all, they are dangerously misleading the public Despite outlandish claims, ICP can be taken down by attacking a handful of nodes in known data centers! ICP's modular design is the problem: 🧵

2/30) ICP is made up of independent subnets, as opposed to being a single chain (monolithic) However, unlike DOT & AVAX, there is no "shared security" Subnets are responsible for their own security! This inevitably results in extremely low validator counts, endangering users: https://t.co/ai2SMfezea

3/30) At most there are less than 40 nodes protecting the user at any one time! At worst, security depends upon a mere 13 nodes, this is not decentralized by any stretch of the imagination... As the security of using ICP is entirely dependent upon whatever subnet you are using!

4/30) In practice, this translates into extremely low-security guarantees for the user This also applies to all features! From HTTPS Outcalls (oracles), canisters (smart contracts), chain-key cryptography & BTC integration All are vulnerable to ICP's weak security guarantees!

5/30) ICP is not a direct PoS system & is even more delegated than DPoS Because stake is not required for block production Instead, stakeholders verify & vote on specific validators! Publicly specifying the data centers these nodes are hosted on; allowing for targeted attacks!

6/30) An attacker would only have to take over/regulate 2/3rds of a handful of known nodes in these data centers! This makes ICP far more vulnerable to attack compared to conventional PoS systems As it has, in part, effectively disconnected stake from the "Cost to Attack"!

7/30) PoS generally uses crypto-economic-incentives to prevent attack Literally putting billions of dollars at stake, in the case of ICP, there is nothing at stake! Except for reputation, ICP, in other words, relies more on trust Going against the very ethos of cryptocurrency!

8/30) ICP also makes the outlandish claim of infinite scalability (Dfinity's namesake) ICP's subnet design absolutely has a scaling limit; unlike what ICP claims; Individual subnets are limited & you cannot "combine" subnets due to "fragmentation" Proofing this to be a lie: https://t.co/9DMk7Y16z8

9/30) Even though DOT & AVAX take a "similar" approach to scaling, neither claim "infinite scalability" As that would be a deeply misleading form of marketing in a finite universe As without shared security; subnets are islands of consensus that fragment the rule & security set

11/30) The same is also true when increasing the number of subnets! Because subnets communicate directly with each other & Distributed Key Generation (DKG) is computationally expensive: That is what creates a hard scaling limit far removed from ICP's initial outlandish claim!

10/30) Furthermore, as subnet replication is increased, the block rate goes down! With higher replication factors, more replicas (nodes) are needed to send messages for the block to be agreed upon Which causes message latency to go up, slowing down the network! (not infinite!)

12/30) ICP also outlandishly claims to have solved several major problems in blockchain tech These are lies, or at least should be considered extreme forms of dishonest marketing; From solving the Oracle problem to bringing smart contracts to BTC & more misleading terminology:

13/30) Starting with "bringing native smart contracts to BTC" As these contracts are only as secure as the subnet & "canisters" (smart contracts) they are built on! As this is where the BTC private keys are actually stored! Allowing the "canister" to directly interact with BTC https://t.co/WsUEb0SR81

14/30) This is made possible due to the chain-key ECDSA signing built into ICP This is in no way unique or innovative, as it exists in multiple ETH contracts now While not being anywhere near as secure as native BTC! On the contrary, it is dangerous & is putting users at risk!

15/30) As if the subnet is compromised, so are the private keys As subnets only have a subset of ICP's security, some with different security trade-offs! While the "canister" could even have admin keys, which could steal the BTC! Calling this native BTC is a massive red flag!

16/30) This is where it starts to get even more ridiculous; ICP claims that it has solved the oracle problem through what they refer to as HTTPS outcalls Which allows smart contracts to get data directly from Web2 APIs This misses the point of what the oracle problem even is! https://t.co/ad6Mse48Es

17/30) Relying on centralized APIs is not a solution to the oracle problem at all It is, instead, the very definition of the oracle problem! As the oracle problem is defined by the inability of a blockchain to know facts outside of itself without relying on centralized oracles!

18/30) ICP is full of this sort of misdirection & technobabble Making it far more difficult to assess through comparative analysis For instance, their choice to call DAO's nervous systems & smart contracts canisters Might seem innovative & complex but to me screams obfuscation

19/30) Tricking new users into thinking that ICP has a revolutionary, innovative design When that could not be further from the truth, requiring us to "translate" the terminology: Cannisters = Smart Contracts Replicas = Nodes Service Nervous System = DAO Neurons = Staked ICP

20/30) Credit where credit is due; ICP has set up a robust system of on-chain governance Including a formalized proposal system with voting weight based on time-locked Both features I have extensively advocated for & are often not included in chains I do currently still support

21/30) It was the case that running a validator required permission This has changed as this vetting role has been moved over to the main DAO This is an excellent way to decentralize a formally centralized system Reminding me of the challenge ALGO faced with its relay nodes!

22/30) Another aspect of ICP's design that deserves praise is its token economics Adopting a similar design to ETH, where there is a small amount of inflation to pay for validator reward While all fees are burned in the form of "cycles", a native stablecoin that acts like gas

23/30) Most of my critique is based on how ICP presents itself I have a problem with any blockchain making misleading & false statements As these are investment assets available to retail This demands higher standards from us, ICP falls way below that bar from my perspective

24/30) Special thanks to the CTO of ICP; @JanCamenisch for reviewing the original critique Of course, he disagrees with my conclusions; that is understandable However, he was able to correct, deepen & nuance the critique more The community also asked for this; I respect that

25/30) ICP can fix these problems: I would advocate for a form of "shared security" which, to their credit, is being worked on However, I would take this a step further & make the leap to full sharding; By combining all validators into one set & randomizing subnet selection!

26/30) This would put ICP security in a range to equivalent L1's As the ICP's security would no longer be split across all of these subnets In other words; the subnets would become shards solving the "shared security" problem! ICP even has "cross-shard" communication built in!

27/30) The language should also be brought in line with industry standards Stop giving things new names! We already have established terminology for everything in ICP It sounds so "scammy"; as it sounds like they are claiming to have reinvented the wheel, a common deception...

28/30) Some time has passed since I first made this criticism Unfortunately, ICP has not cleaned up its communication, only doubled down! There is also the controversy of the initial supply, neither Arkham nor ICP have disclosed the addresses... A severe lack of transparency!

29/30) The marketing is dishonest, whether intentional or not People are being sold the moon, as ICP claims to have solved all of crypto's problems Yet, in reality, it is highly centralized & insecure! There are some good parts to ICP, but that does not redeem all that is bad

30/30) I wish the best for ICP & its supporters, as we are fighting for the same dream I could even become a supporter if ICP switched to sharding! The misleading marketing also has to be cleaned up As I think everyone would be better off with just a little bit more honesty ❤️

@Justin_Bons lol… 1 on 1 debate. Minimal 2 hours. Ideally 3. Equal time for each. ICP vs SUI. Or any L1 you want. Agreed upon debate format beforehand. Sometime in the next 2-3 weeks. What do you say? Repeating “tHeReS oNLy 13 NoDeS” won’t work just fyi https://t.co/t6IS9492qb