ICP should go back to cloud providing

Developers
1

I have spent enough time in this ecosystem to know that it is time to cut our losses and move back to a simple cloud providing model governed by blockchain.

I appreciate what Dfinity has done but let us be honest with ourselves. We are working on “alien tech” - tech that has never and will never really be used by real human beings. Why? Lack of focus. As the crypto ecosystem has matured, it is apparent that the concept of a blockchain singularity is delusional.

Dfinity seems to think the next “thing”; be it Caffeine or Cloud Engines will save us. It won’t. The pattern has been very clear and even the deepest of believers need no convincing on this. I have written software and I think I can speak on this. By no means am I the best in the world, but I have done my time around software that has been used by actual people at scale, including the use of cloud and AI tools in the building and deployment of software.

In my opinion, Caffeine should be halted. It is very clear that we will not win against big labs since there is no IP we have. Caffeine fully relies on Claude models. Having used Claude Code I can say that it is sort of futile to compete against these tools. Trying to build websites without code is a pipe dream. You might get 95% right, but the last 5% will make things impossible. If you do not understand what the software should do, you can probably get to an MVP phase before the 5% starts to bite. That is why WordPress and Shopify popped up. Also most of the apps we consume are mobile native apps, which Caffeine cannot build for. Please do not mention PWAs since we know those have never really been used.

This is just a microcosm of what is wrong with this project. It is not the lack of talent. It is the lack of direction and flexibility. We have pigeonholed ourselves into this vision that no longer makes sense. We have created Motoko, canister architecture and more software that has made mass adoption harder. We have sunk ourselves into a hole and it seems instead of trying to dig ourselves out of it, we keep digging deeper, thinking that it will resolve itself. We literally make 5k a day. A billion dollar project. We must ask ourselves where things went wrong. While I appreciate everything Dfinity has done, including the building of said software, maintaining it and working on governance initiatives like Mission 70, I think what we need is positive momentum, which we have not really had since 2021. Price no longer matters since most investors are really deep in the red, but it is unfair that the problems we have are so obvious, and they will keep getting worse since it seems the leadership at Dfinity is eliciting signs of a gambler’s fallacy.

Here is my suggestion. Dfinity should decommission all the software initiatives it has now. Decommission Motoko, canisters, Caffeine and everything to do with software for now. It should focus on getting the architecture right, and that is a cloud model similar to AWS, GCP, Hetzner…
We already have the tools and software I believe that such a change would not be too drastic. We should be able to support similar software as AWS does. Databases, Redis, Queues, Load Balancers, Object storage etc. At the very basic, one should be able to run the same software they do in Dfinity nodes as they do in AWS, and have the same deployment workflows as they would there. It is very hard to change software when we already have mature ecosystems and stacks used in enterprises. The money is no longer in software but in compute, that is why companies like SanDisk have grown massively in the past year.

Dfinity should simplify the things they have built. While I acknowledge the tools they have built are sophisticated and even in some areas superior, Dfinity should repurpose them to handle modern software stacks. Here is where blockchain comes in. We have NNS and such initiatives and those can be used in governance of the cloud environment. ICP should be used as a proxy for the cost of compute and should be decentralized. We can still have grants for people to build on it, but they should be able to use Python, Java, Rust, Go etc. We should be agnostic to the software used in the platform. It might seem like cutting back, but in that way we actually increase the market we serve. My proposal is essentially follow a Vercel or Cloudflare strategy (which are massively successful by the way).

This is the only way we stay alive. I am sure of it. Wasting time on canisters and Caffeine is a bet we will surely lose. It is so clear I am not sure why Dfinity has not really done this. It is fine to admit defeat. Sinking heads in the sand is a losing strategy. Differentiation through complexity and alien tech is simply wasting our hard earned investment.

Anyway, those are my two cents.

2

You’ve had your say, which is fair. My say is that this is the silliest thing I’ve ever read - sorry.

3

I’m not smart enough to speak on all the other stuff you said.

But this is a minimum requirement. That’s just common sense.

Because:

need it.

There’s already a canister page in the (old NNS dapp) why can’t someone drag drop deploy a basic project or a static site to an asset canister like you can on every other cloud hosting provider?

Someone should be able to deploy an asset canister in a few clicks.

When this WordPress project is mature, that should also be able to be deployed in a few clicks. Straight from the NNS dapp.

image

5

What is your actual rebuttal, if any?

6
  1. You think that they can’t compete with the frontier labs but should instead compete against Amazon?
  2. No one wants a second job. People use AWS because they want their queue to just work…they don’t want to “govern” their queue any more than they want to support it. They want it to just work.

A clear direction on actor-based architecture and a decentralized/agnostic stack that “just works” is the one thing they’ve actually done that is different than anyone else. If you want generic “infrastructure” + consensus, there is always the manifest network, which was founded by an ex-Dfinity person who didn’t like the canister-based setup and wanted to do what you are describing. I don’t know that they are doing any better.

The governance is what should go away. What works should drive the network and hardfork if you disagree, should rule the day. The governance theater is getting way beyond old. It doesn’t work. Vitalik told everyone in 2021 it wasn’t going to work. It still doesn’t work.

7

I do not disagree what we are doing is different. On a technical level it is quite good. But how do you expect hordes of developers to move to just Motoko and a new canister architecture. Abandon Spring Boot, Quarkus, Express, Redis, Postgres. They have been used for years, well documented and debugged.

It is really about pragmatism. If you feel any contention, we can argue on why we make 5k over 49 subnets. There is clearly a blocking factor and my mind says it is the restrictive paradigm. We can still design actor achitectures in other languages

What Dfinity has that manifest does not is presence and an already working network.

Or we might as well say we were funding an experiment.

Competing with aws is easier and a much better proposal than saying we need to be complex in order to compete. We need to streamline the architecture to be agnostic first, and then we can introduce software on top like cloudflare with workers for example. Allow people to build cheap and use things like cloud engines as enhancements. That way we serve a much larger market. No one cares about actor architecture unless they are deeply technical. I am not saying it will work 100%, but it is a much much better strategy

8

Unfortunately, these things are not the priority here.

Some people prefer to sell dreams to dreamers, rather than see things materialize.

9

I agree that governance should be streamlined. When i was talking about governance I meant like on the business side not technical, sort of like investing in projects, not every code change

10

The resiliency and trust drawbacks of conventional IT infrastructure are increasingly being demonstrated to the world, with data centres being targets in wars, outages in critical infrastructure due to central points of failure, even crypto related infrastructure that should be resilient, but isn’t at all → (2) Coinbase Support on X: “On May 7th Coinbase experienced service disruptions. Here’s a quick summary of what happened: → Around 8PM ET, Coinbase systems flagged high error rates across multiple services. → We traced these errors to amazon failures in Availability Zone (use1-az4) in the AWS US-EAST-1” / X

And while the world is just beginning to wake up, you suggest the IC should turn away from the entire value proposition? The whole stack is built around supporting canister smart contracts that deliver resilient compute (and facilitating trust-minimised compute in an ideal future scenario where the network is more decentralised).

Regarding languages and adoption, AI will increasingly act as lubricant here, and indeed already is. There are no signs of this slowing down, with distillation from larger models making smaller models increasingly capable. Every year there seems to be another huge leap forward.

The IC is very well positioned, and you’re suggesting completely turning away from the USP.

Why is the price in the dumps? Because the world is asleep. It will wake up, and there are signs all around that it is starting to.

Now pull yourself together man.

12

Is the world asleep? or are we asleep?

What is real?

You think that’s air your breathing?

image
image275×183 16.6 KB

13

So your proposal against attacks and data leakage is to move to a network that has 49 subnets and makes 5k a day, is unproven at any major scale, uses an esoteric language stack and a heap as a database, is supported by a small team in Switzerland whose cutting edge researchers left so they could be lean and focus on a vibe coding platform​ :joy:

You get yourself together! You are very funny

14

Ahhh…now that is a good question. Typically, you need to build something that is 10x better than anything else out there. It doesn’t need to be 10x better than EVERYTHING at EVERYTHING, but it is does need to be 10x better at SOMETHING. At genesis, they easily had a 10x blockchain infrastructure platform; instead of releasing that, they tried to come up with a bunch of new buzzwords and to be the smartest people in the room with new-fangled architecture and buzzwords.

I think I can make a pretty solid case that the 2nd, 3rd or 4th worst decision DFINITY ever made was making RUST a first-class citizen on the IC and building all it’s system canisters with it. The IC is a shitty rust platform. It is a 1000x better than any distributed actor-based platform, which Motoko specializes in. If DFINITY had the discipline to lock into and sell out to the unique value prop, we’d have had all the software reoriented to actor and canister-based architecture by now, out of necessity. It would be tribal and have velocity. It, of course, shouldn’t have been just Motoko because motoko is bad at defi(but so is icp flavored rust). But that doesn’t seem to be a very hard answer to get to either. Byegones are Byegones, though. Here we are.

And I still insist that cycles are not and will never be the driving force of value. We could burn a cycle a day but if that cycle ran global commerce people would pay millions to control ICP. You have to put control on the chain and then it has value. If you think you’re going to burn your way to profitability, you’ve never read The Innovator’s Dilemma or observed the ingenuity of a script kiddie. The instant the burn gets extractive to ICP holders is the instant that computation finds a way off chain. Such is the way of the world.

15

what is the worst decision?

16

Could you explain what you mean by extractive? If you mean ‘more expensive than off-chain’ then this implies nobody would ever be willing to pay a premium to do things on chain, which is not true.

The IC is a first mover in the space of real on-chain compute. Everybody else is having to play catchup. It’s not a bad thing that other’s are trying to catchup. It’s a good thing. We need more competition in this space, and more eyes on the fact that there is meaningful utility to be exploited and marketed downstream.

18

As soon as it becomes expensive to do things on chain there will be a cryptographic solution that makes it no longer necessary to do it on chain. You still need the chain for the cryptographic link/evidence but that burns almost nothing.

This is the classic “the file isn’t on chain” “I don’t care the hash is and that is good enough” “nuhuh you can’t compute over a hash” “fine I’ll compute off chain and put a zkproof that I had the content on chain” conundrum.

There was never any cake in the cycle burn argument. It is a lie. The utility of the cryptographic material is what will establish the value. Maintaining its integrity when the value of its integrity grows. The cycle burn can be 0 but if the root hash secure $6B then the market cap is $6B. That was always the only value here and it seems odd why it is so impossible for everyone to see this.

19

But then if I extend your reasoning, why use the IC to secure a hash. Why not existing blockchains with thousands of nodes?

20

I think you’ll find what you’re saying doesn’t make much sense for the vast majority of use cases where consensus and liveness are important.

21

Because there is plenty of cheap commodity computation that is profitable to do and worth not offloading and jumping through hoops for on the IC. But people speak of the cycle opportunity as if it will explode and we’ll all get rich because someone will create something so great and so expensive computation wise that there will be a cycle bonfire to such an extent that ICP will go to the moon because no one can find any. That is extremeistan thinking while playing in mediocristan…and it isn’t going to happen. That process will find a way to offload the bulk of computation. The burn will be boring campfires and you are going to have to make it up in boring normal apps while running against Moores law that makes that bonfire from four years ago into a campfire today. Every 18 months your going to need to double your compute or nodes end up sitting around under utilized(this is on average, I get the actual gen 1-2-3 rollout is looking a little different, but I’m not sure why)

So it isn’t that we can’t burn a bunch of cycles or that the compute on chain can’t be “worth it,” it is just a long term bad bet to be the thing that drives the most value.

Demand from governance has extremistan qualities. People can attribute abstract value and build layer upon layer of abstract value. Governance gets the tailwind of metcalf’s(the value of a network grows with the square of participants added), so if I think one governance aspect so important and you think another aspect is important, our combine valuation grows to larger than our individual evaluations.

Cycle burn is like the baseline…or foundation…you want it and need it and it is a valuable component of the tokenomics mix, but you kind of get it for free and it is the building on top of the foundation that is actual interesting, but we all seem to be so autistic that we can’t stop measuring if the foundation is flat enough or not.

22

I think you will find that the vast number of instances where liveness and consensus are important require remarkably little compute that becomes half as expensive every 18 months.

You’re betting that we will find one instance where it won’t be cheap and in that instance everyone will just eat it and never innovate.

I don’t like your bet and suggest that everyone delete cycle burn from their brain and focus on the other 15 possible tokenomic vectors that ICP can draw on.