James Allan Left ICP for Kaspa — And I Believe He's Got the Technology Backwards
James Allan left ICP for Kaspa, and in my view he's got the technology completely backwards. Now, to be clear, I love James Allan. I had a great one-on-one call with him a year ago and we really had a great conversation. I'm inviting James to do his own research further, because from everything I've studied, almost everyone promoting Kaspa has no idea what the actual tech is and offers no relevant comparison to better tech like ICP. The truth as I see it is that Kaspa solves nothing. And you don't have to take my word for it. If you have ChatGPT Pro, just ask it to do some deep research through the actual Kaspa documentation and then compare that to ICP's infrastructure, and that is the conclusion the AI will come to. How do I know? Because I already did exactly that. Kaspa solves nothing.
I pulled a ton of data for this. I used three different AIs: Google's Gemini, Grok to pull a bunch of material from X, and then ChatGPT Pro to put it all together. And this is the conclusion. In my opinion, every single person telling you glowing things about Kaspa either has no idea what they're talking about in the best-case scenario — they're keeping it simple, they're not diving deep, they're talking in big generalities while the technical reality of what they're saying makes no sense — or something worse is going on. I'm going to take apart the Kaspa promoter arguments with evidence that is very clearly organized, and that you can replicate yourself with AI.
This is a sign of what's coming. AI is able to cut through all kinds of shallow opinions and ineffective arguments, because AI can take in so much more data, put it all in context, and get down to the truth. The AI actually read every piece of documentation and put it all together, and this is what you get. So James Allan, I invite you: think things through. We would all love for you to change back to ICP when you see that, in my analysis, everything being said about Kaspa makes no technical sense. Every other Kaspa video I've reviewed from Kaspa creators is, in my honest opinion, completely ungrounded in any technical reality. That's a problem, and it's why you need AI to help you break down what other people are saying.
The Kaspa Argument on the Surface
James Allan, known as Cityscape, is, to be fair, saying essentially the same thing the other Kaspa promoters are saying. On the surface — and I watched multiple videos of his to get this information — his argument sounds technical: that replicated computation is inefficient, that ICP uses subnets which he says fragment security, that off-chain computation loses trust, and that ZK proofs restore trust without destroying computation. He was recently criticizing ICP and talking about how superior Kaspa is, which in my analysis is provably false — the AI will show it to you in a few minutes of research, and it takes apart the entire Kaspa thesis. In my view, the Kaspa promoter talking points about vProgs, the Kaspa blockchain, and the narratives are blatantly false information — it's all junk. This isn't financial advice; it is YouTube advice, though. I believe they do not know what they're talking about, and I do, and the AI will agree with me and not them. The problem is that their conclusions do not survive contact with the actual architecture as the AI lays it out. Welcome to crypto, where what people say so often doesn't match what the documentation on the website says.
James compares the visible cost of ICP's working production system against the promised benefits of an unfinished Kaspa framework — a framework which, in my opinion, in 99% of scenarios will go nowhere and offer nothing valuable, because even what it's aiming to do is not useful next to what ICP has already been doing for five years. And I say anyone presenting Kaspa in a positive light is, in my belief, doing poor research most of the time — shallow research. If you can't be bothered to get a ChatGPT subscription for $100 a month and fact-check the things you say in your own videos like I do, that's poor research; and if it isn't poor research, I struggle to see an innocent explanation. The fact is, though, that in my view audiences across crypto, especially around Kaspa, are being grossly misled.
My Core Argument: On-Chain Applications vs. Ordering Receipts
My argument is simple. ICP runs applications on chain, and this is the most breakthrough technology in all of crypto — AI doing deep research and reading the documentation from all the other coins will absolutely agree and reaffirm this. Meanwhile, all Kaspa does is order application inputs, move the real computation off chain, and verify the receipt later. In my opinion this is worthless, because a whole bunch of coins already do this. Kaspa has almost no chance on marketing, it has not solved the computation trilemma, and it has simply made tradeoffs that ordinary people don't understand.
Now, I will acknowledge the small part James gets right. Yes, running computation once is cheaper than having independent machines execute it. And you know what? If I want the cheapest computation on the biggest model, I'll run it with ChatGPT or Google or Anthropic. ICP is not designed to compete with centralized models, and Kaspa is, in my view, utterly useless by comparison. The replicated blockchain state deliberately performs redundant work. Yes, ICP does have some greater expenses. But the idea is that it's providing infrastructure that is hard to mess with, which Kaspa does nothing for. Any value Kaspa is trying to produce is better produced by huge companies that are already doing it much better with bigger resources.
James describes ICP incorrectly — and so do these creators in masks who come on YouTube. I can't imagine a real person would watch somebody in a ridiculous-looking mask talk about a coin that AI will take apart for you in two minutes. Just forgive yourself if you've been tricked into watching something like that, and take it as a good opportunity to change your life and listen to somebody who might know a thing or two and actually cares about you. Now, James puts his face out there, which I'm proud of him for, and I like James. The guys with the masks — I've already described how I feel about that.
In my analysis, James's argument about ICP is incredibly shallow, and it does not stand up under any level of research. People watching videos on YouTube are vulnerable because, one, you don't know this stuff for yourself, and two, you don't take time to look it up with tools like AI that could explain all of it and present it in a simple way. So you need to do one of three things: learn it yourself, check it with AI, or pick somebody you can trust who actually cares about you and presents information that's technically accurate — so that when you check the information with AI, what they say matches what you find. In my experience, almost nobody else in crypto says things that match the technical reality of what they talk about. There are exceptions, like Bobby O and the people in the ICP community.
The Horse and the Car
James's argument describes ICP completely inaccurately. ICP scales horizontally instead of forcing one global machine to replay every application. His is an incredibly simple argument based on attacking the blockchain with the most advanced technology in crypto and comparing it to something like Kaspa, which in my view is basically a copy of a few other blockchains — extremely simple, with almost no technical capabilities compared to ICP. It's like comparing how a horse performs to how a car drives, and then saying, well, I don't have to put gas in this horse, so a horse is better than a car. It depends. If you want an animal that's a friend, then yes, a horse is better. But if you're trying to drive somewhere, a car is going to be better. What does the Kaspa blockchain do? If you just want to send transactions, Kaspa does that. But technically it's like a unicycle against ICP being a car. In my opinion, it's useless for what matters.
And ICP does not need to replicate every single request. This gets pretty technical, and I've already gone on about it for a while, but James says ICP's design means everything gets executed repeatedly — some of the other Kaspa promoters say the same thing. That is inaccurate, and on ICP, the replication is the security product. He calls it inefficient, as if some of the most brilliant cryptographers and researchers in the world failed to notice that several machines use more resources than one. Of course they noticed. The replication provides fault tolerance, independent verification, resistance to a single dishonest operator, and application continuity when machines fail. There's almost none of that in most of the rest of crypto — especially, in my view, in Kaspa with the setup it has, where there's hardly any resistance to one dishonest operator.
Saying replicated compute is wasteful is like saying database replication is wasteful. Sure, if you have data you don't care about, why replicate it? But ICP is designed for data you really care about. It's like saying backups are wasteful, RAID storage is wasteful, redundant flight control systems are wasteful. Do you need a co-pilot driving your car down the street? Probably not — though some of you driving out there, maybe you should have a co-pilot in case you don't drive so well. In a car, the stakes are not that high, although car accidents cause a lot more deaths than plane flights do. But in a plane, you can't afford to have one pilot go down, because then the whole plane crashes. ICP operates the same way: it's built for applications you can't afford to have go down. If you can afford downtime, there are lots of other cheap places to put your application where it doesn't matter. ICP is providing a value the rest of the world only offers in cheap, weaker forms if you don't care about your data. So the argument just makes no sense in the bigger context. One server running a program is cheaper, sure — but it means the server owner has basically total control.
In my view, James's argument absolutely crumbles under a little bit of research. Everybody in ICP sees that. But people watching Kaspa videos are taking this in and believing it. It's like wondering why your body doesn't perform well when you eat junk food — if you eat junk food, it's going to cause results. I'm not a medical doctor, as well as not a financial advisor, but I am very careful with what I put in my body, and my body looks beautiful and functions insanely well. Look at my shopping cart and see what I buy — it's no surprise that I'm insanely healthy. I have no issues, no pain, my body works amazingly, and it has tons of energy all day. I can go do hot yoga, go out on a date, play tennis, create a whole bunch of videos, and play with my kids.
Junk Food, Junk Information
I have practically infinite energy until it's bedtime, at which point I fall asleep easily and sleep well. If you look at what I eat, it makes sense. This is why you need to pay attention to what information you're taking in. If you take in junk information from Kaspa promoters, it's no surprise that crypto is hard for you. I have empathy for you, and I offer you a better way: my videos are much better information than almost anyone else in crypto puts out.
Do ICP Subnets Fragment Trust? I Say No
We went heavy into ICP already, and I'm eleven minutes in, so let's keep going. ICP subnets do not create a random collection of disconnected blockchains. Again, this is too basic of an argument, and it doesn't make sense against a more advanced infrastructure. It's like judging a car by horse standards and complaining that it doesn't have a tail. Sure — a car doesn't need a tail, James. It operates differently than a horse. ICP is operating at a different level than Kaspa, because hundreds of the smartest crypto researchers in the world worked together to build something that I believe is vastly superior to anything else. There are people desperately trying to copy ICP. I promise you, in my opinion, there's no one trying to copy Kaspa right now. Just saying.
The claim that ICP subnets fragment trust is, in my view, ridiculous. It all operates under one single protocol, with messages authenticated by the origin subnet and the receiving subnet. This system is so advanced that it doesn't work very well to try to explain all of it quickly. ICP does partition execution — security comes from the hosting subnet — but it does not fragment its entire cryptographic trust model.
And here's another basic argument that so many people get wrong in crypto. James implies that smaller subnets automatically mean weaker decentralization. Deterministic decentralization is more sophisticated than counting anonymous validators — because anonymous validators, especially in crypto where almost nobody would want to run a validator on some random coin like Kaspa, are often all insiders running all of them. In my belief, there's often no real decentralization in almost any of these blockchains, because almost no real person wants the thankless task of managing a validator. Implying that smaller subnets mean weaker decentralization based on validator count alone tells you nothing, because a network can advertise thousands of validators — look at Solana, where 40% of those anonymous validators went offline when one single cloud provider shut them off. A network can say it has however many validators, but if they're all set up by the same people and there's no transparency, it's basically useless. Meanwhile, ICP's system is very transparent. And man, we've only gotten a third of the way through this, so let's speed it up.
Chain-key cryptography is totally missing from James's criticism. A stable public master key that can be verified across everything is so much more advanced than anything I believe Kaspa is ever going to be capable of. It enables things like having AI agents send Bitcoin, and so many other capabilities. James reduces the argument to "small shards," and so do the other Kaspa promoters — while talking about how delusional I am and going on about ICP's price chart and all this junk, which I don't watch their videos for anyway. Why would I waste my time watching somebody in a mask talk about Kaspa? Based on all this research, I'm confident I know more about Kaspa than they do, because in my view they're just parroting hype and getting views. I'm doing all this research because I do this full time, I've done it for a long time, and I'm pretty smart with all this tech stuff.
What Kaspa Actually Does, In My Analysis
Meanwhile, if you look at what Kaspa actually does, it does fast proof-of-work ordering, which in my opinion is basically useless in today's environment. Maybe that would have been useful a decade ago. Its BlockDAG architecture is way overhyped. All Kaspa is is a new way to send payments, order transactions, handle settlement, and put some transactions through. It's legitimate technology — it's not absolute zero — but fast ordering is not the same as running a decentralized application. All it can handle is basic transactions. It can't execute a social network, an AI agent, a database, a web application, a messaging platform, or general cloud logic. All it does is move the application off chain, which is what almost everything else in crypto does.
James's argument that off-chain computation is better and that on-chain is too expensive and not needed is, in my view, absolutely wrong for the use case of crypto, as the AI research lays out. On-chain computation really matters if you care about having your application secure and running properly, which in crypto and in other important infrastructure is very important. For Kaspa's so-called "based apps" and vProg-style architecture, the chain orders the user's operation and then an off-chain runtime interprets the operation. This is where the whole thing falls apart. As soon as you put things off the blockchain, the trust is gone. You might as well just run all of it in some centralized service. There's no reason to deal with the Kaspa blockchain and then put everything off chain, where you lose all the security and verifiability of the blockchain.
This is where I believe everyone promoting this is either completely unaware, incentivized not to look, or worse. Most of the time, I think people simply are not that understanding — they haven't done that level of research, so they don't grasp how big a deal this is. As soon as you move things off the blockchain, you lose security and you lose verifiability, and without those you might as well make a centralized app, because it would be cheaper and faster than dealing with the Kaspa blockchain plus off-chain infrastructure.
The separate proof infrastructure generates evidence of computation. It's evidence — it's not a guarantee. You have an executor performing a computation, and there's a little bit of evidence. The proofs can be batched and aggregated. The actual computation therefore depends on things outside the consensus: executors, provers, state databases, witness servers, proof aggregators, settlement operators, and reorganization handling. I might as well just put everything on Amazon if I'm going to deal with all of that from an unknown group of people working around the Kaspa blockchain. The technical reality, as I see it, is that Kaspa is ordering transactions and then later verifying a proof produced by someone else — utterly useless in 99% of scenarios. Might as well run it on centralized tech, or use ICP to build out what you really need.
Here's the takeaway: ICP runs the application. Kaspa verifies a receipt after an off-chain system runs it. In my opinion that's useless — all these other blockchains are already doing that with the same structure as Kaspa, and they have more money and more marketing. There's no reason for somebody from another blockchain like Solana, which already runs like this, to move over to Kaspa, and no reason for a new developer to come over either, especially when Kaspa doesn't have the money to pay them like these other chains do. ICP is the only setup where I see genuine organic adoption. Kaspa, without money to just pay everyone to promote it like other chains, is in my view useless and doomed outside of manipulated or paid pumps, or just literal insanity.
The Off-Chain Trust Problem and Composability
Kaspa has not solved anything, because of the off-chain trust problem. ZK proofs can prove that a computation was executed correctly, and that has some uses. But you need more than mathematical correctness. All the proof can show is that the transaction was valid. It doesn't guarantee that someone will execute the next transaction, that anyone will continue storing the application, that witness data stays available, that proof generation will remain economically viable, or that the front end will remain available. Those are critical failures that, in my analysis, ruin the entire value proposition of vProgs and everything anyone is saying about them.
On top of that, there's a composability problem. James admits ZK computation loses composability, and then claims vProgs solved that problem. In my analysis, no. Proof-based applications have separate state, execution, and settlement boundaries, which means applications can't always synchronously access one another's state in an unsettled condition. The project's own architecture material acknowledges that based-app interaction with the L1 may be delayed or exit-style. That means you can't really use everything together, and other applications can't treat that state as the final L1 state. Kaspa has not solved any key problem here. In my view, it is desperately building and trying not to go to zero based on technology that is fundamentally worthless to humanity. Have I put that clearly enough for everyone to understand?
Meanwhile, there's still a cost. The cost may be less than putting everything on ICP, but there still is a cost, and you have to make all of it economically viable — because even if you're using one single machine, somebody still has to pay for it, and therefore you still need money coming from somewhere to pay for all this. Otherwise everything goes to zero and none of it works. General-purpose ZK proving has enormous overhead. ZK proofs are extraordinary for the right applications, but proving every piece of general cloud computation is a completely different economic proposition. Kaspa's current reference approach is not identical to every other zkVM system, so I'm not going to pretend that one project's cost multiplier automatically applies to Kaspa. But the execution costs resources, then it costs additional resources to prove it, to aggregate it, and to witness it, and then there are on-chain computation costs on top. So again, why would anybody deal with all of this when you can have a better experience with more liquidity on other blockchains — or have the best of both worlds with the best technology in crypto, which eliminates almost all of this? Yes, the cost is a little higher, maybe significantly higher, but it eliminates huge amounts of uncertainty, which is very important if you care about your data security and about having an app that runs continuously and operates smoothly.
James also selectively counts ICP's complexity, and in my view this is what all the Kaspa promoters do: they ignore Kaspa's complexity, probably because they're unaware of it, while treating the Kaspa stack like it's one elegant solution. Yes, ICP is extremely complicated — but it does miraculous things that nothing else in crypto can do. Here's everything a Kaspa based app may need: proof-of-work consensus, transaction lanes, an application runtime, an executor, a state database, a witness system, a zkVM, proof workers, proof aggregation, settlement submission, covenant logic, and reorganization handling. Compare that to my experience on ICP: all I had to do was talk to the AI, feed it data from my videos, and it cranked the website out directly onto the ICP blockchain.
ICP's Complexity Is Now Easy to Access Through AI
So even though ICP has a huge amount of complexity in the backend and in how it all works, the execution for a developer who wants to take something and put it directly on the blockchain — and immediately have access to a website that serves straight from the blockchain, secure, with all the properties — is remarkably simple. My website is not going to go down because one server crashes, because it has redundancy built in. ICP's complexity is extremely easy to access through AI now, where you can go directly into the command line interface with Claude and it will put the whole thing on the blockchain. I literally took my friend's website: she gave me the zip file, I dumped it into Claude, and Claude rebuilt the website and deployed it onto the ICP blockchain. In my view, that is so far advanced past anything Kaspa can do that it's blatantly ridiculous anyone could think one positive thought about Kaspa without realizing it's like they've been in a trance — like they got put in a trance and didn't critically think about any of this.
And there's real value in realizing that. I hope somebody has a spiritual awakening during this, wondering, how did I ever think anything positive about Kaspa — and thank God somebody cares enough about me to tell me this stuff instead of just telling me what I want to hear. Do I think that's going to be the normal reaction? Even if one person has that awakening — whoa, what have I been doing with my life? — it's worth it.
Proof of Work Is So 2009
Meanwhile, in my opinion, proof of work absolutely sucks. That's why almost none of the new blockchains use proof of work — long story short, it sucks. Proof of work is so 2009. You have to run this intensive mining network for consensus, then a separate off-chain infrastructure on top. That's a nightmare, whereas ICP has everything in one environment. Calling ICP's replicated execution wasteful while treating proof-of-work hashing as economically invisible is, in my view, an extraordinary and ridiculous double standard.
Kaspa's framework is also still early. James and the other Kaspa promoters present vProgs as though the trilemma has been solved. In reality, the framework is early stage and evolving, and there are tons of production concerns right now. Researching a possible architecture is not the same as proving it works at global scale. ICP already works at global scale — my website is proof of that. It's on the blockchain right now, serving thousands of pages, instantly, nearly for free, directly in your web browser, with no third party in the middle.
James, Come Back to ICP
And James, come back from Kaspa. I feel like I'm trying to talk to somebody who's left their religion, saying, come back to God, my son. In my honest opinion, Kaspa is so junky that anyone talking positively about it makes me wonder — how do you not know this stuff? This is what you should know. Compare the value delivered: Kaspa provides fast proof-of-work transaction ordering, and ICP provides everything I've described. To me it's astonishing that there could be any positive thought about Kaspa in the mind of someone who also knows about ICP. Bitterness, resentment, and a closed heart can unfortunately color our perceptions and stop us from seeing clearly, and I hope this encourages some clear seeing. A fast ledger like Kaspa is not a world computer. A world computer is the future, and in my belief, Kaspa has nothing to offer the future of humanity. Now, I understand ICP has some weaknesses, and yet compared to Kaspa's, in my view they're not even worth mentioning. Yes, the network itself is much more complex on ICP, but it can do so much more, and it's easier to access.
I've ranted for twenty-some minutes. I love James Allan. I hope he will leave the Kaspa promoter community behind, and I hope viewers will stop watching Kaspa videos when they see that, in my judgment, the people making them don't know the things they should know about — things you would want to know, which I do know, and which other people in the ICP community know too. I've gone after the Kaspa videos over and over again because, to me, it's so blatantly obvious how weak the technology story is that I'm shocked anyone cannot see it.
And I invite James: you changed your mind and left all those other cryptos to come into ICP once before. My impression from the outside was that you were frustrated about DFINITY not paying you to do anything with ICP, and then you went over to Kaspa. As far as I've seen, you haven't said whether or not you've been paid anything to make Kaspa videos, and I would genuinely love to know. For my part, I haven't been paid anything to do ICP videos since 2024 — I've done ICP videos for free for two years, because I believe it's the best tech in crypto. I wouldn't accept money from anyone offering to pay me to promote some other blockchain, because ICP is the best tech in crypto and I'm not going to talk about anything else. You can see all of that work for yourself on my ICP Crypto playlist.
So I invite you, James: please come back to ICP. We'd love to have you back. I invited you back once before, and here's my real truth: to me, someone who loves you is going to tell you the truth straight. In my honest opinion, the Kaspa videos don't hold up, and I believe most people watching can see that. I know people say the same about my ICP videos. I claim the future is going to prove me very right — like Leonardo da Vinci drawing helicopters hundreds of years ago, I'm just seeing things ahead of time. You could call me a technology prophet if you feel like it. There are a lot of areas of life I'm pretty clueless in; technology I know pretty well. I care about each of you as my viewers and I want you to have the best information. The best information, in my opinion, is that Kaspa is not worth your time at all. None of this is financial advice, and all I hold is ICP. My position is transparent, honest, and clear. And that's enough.