Hello everyone,
In this article, I will share many developments and plans that are crucial to the Internet Computer ecosystem.
In the past few months, I have been using the term 'Internet Computer 2.0', which is increasingly related to the importance of hosting self-written applications (which users create through non-technical interactions with artificial intelligence) and providing mainstream cloud capabilities.
But there are other changes, including DFINITY 2.0, ongoing work to create for-profit enterprises in the ecosystem, and plans to promote network token economics deflation optimization.
To truly understand some of our guiding reasoning, we must start from the beginning.
Our infrastructure
Back in October 2016, I founded the DFINITY Foundation in Switzerland, an independent organization dedicated to advancing internet computing and fostering its ecosystem. At that time, the project had been running for a year and a half, but real progress had only begun to accelerate in the last six months.
I believe that if there is a primary force driving the development and support of the public internet, it should be a neutral, non-profit organization, such as an independent Swiss foundation. My view remains that if the internet were owned and developed by Microsoft, it wouldn't be the internet we know today. It's far better to be led by a more neutral organization like ICANN!
Due to the involvement of venture capital, the industry has adopted different models over time, with most decentralized networks primarily guided by for-profit companies, such as Solana (Solana Labs), Avalanche (Ava Labs), Polkadot (Parity Technologies), Cardano (IOHK), Sui (Mysten Labs), Aptos (Aptos Labs), FileCoin (Protocol Labs), and so on.
(Once these companies have established their own networks, they later create foundations for their own ecosystems, but these companies usually retain de facto control, which creates a different dynamic.)
Foundation Model vs. Corporate Model
The people at DFINITY, like those for-profit organizations that support blockchain, want the Internet of Things to succeed, but their framework is different.
For-profit companies have a fiduciary duty to maximize shareholder returns and seek the best return on investment (ROI) in all their activities. Foundations, on the other hand, have a fiduciary duty to pursue predetermined goals—in the case of DFINITY, the goal is to develop the technology that powers internet-connected computers and help drive the development of the related ecosystem.
Over the years, DFINITY has increasingly operated more like a research institution than a technology company. The Ethereum Foundation exhibits a similar pattern, although DFINITY's technological advancements are much faster.
How influential for-profit ecosystem companies can provide assistance
The Ethereum ecosystem has benefited from Consensys, a large, independent for-profit company led by one of its main founders, Joe Lubin. For years, Consensys has been a driving force in its ecosystem, working to increase institutional acceptance of its DeFi business and creating products and spin-offs such as MetaMask and Infura, which have directly driven the adoption of DeFi.
While I believe decentralized networks should be led by independent nonprofit foundations, I also believe that ecosystems need strong, risk-taking for-profit companies to drive growth—and ICPs need to benefit more from this.
In light of the above, DFINITY adopted two approaches to proceed.
Making DFINITY more like a technology company
DFINITY has been working to restructure and evolve into a "DFINITY 2.0" model. ICP technology has become extremely mature and stable, and it's time to push for its large-scale market adoption.
We have developed an ambitious roadmap, which I outline in this article. We have been restructuring to become more agile, reduce management layers to deliver products faster, and focus more on user experience and productization to connect internet-connected computing with market opportunities.
This involves internal restructuring, rebalancing resource allocation across functional departments, and cultural change, as we strive to make the foundation operate more like a technology company than a purely research institution, while maintaining our deep-rooted science and technology mission.
DFINITY is entering an exciting era. We are returning to our roots, driven by our mission as innovators, and taking a proactive approach to seek new opportunities with enormous potential. This article will explore this topic.
Building impactful, profitable technology companies for the ecosystem.
Given the unique advantages of for-profit technology companies, we are also seeking to create and support for-profit projects that can independently drive the large-scale use of the Internet, computer networks, and cloud.
Caffeine was the first. It's a platform that enables users to build applications, websites, and enterprise systems on internet-connected computers by interacting with artificial intelligence using natural language and providing non-technical documentation.
Crucially, Caffeine not only possesses the appropriate architecture to meet the challenges of building entirely new mass-market platforms, but it also independently raises funds to expand its market share in the AI platform market. This is critical—DFINITY itself cannot indefinitely fund such a large-scale project.
The project nearly failed after our first attempt to create the product ended in failure, reflecting the challenges of building such a platform within a foundation focused on underlying network protocols.
In the months leading up to the official release on June 4th, I took over and created the entirely new user experience you see now, thanks to the hard work of the 996+ team. Once the user experience had the necessary potential, I assembled a completely new team like a startup and abandoned the old codebase.
Caffeine currently operates as a hardcore startup, focusing on optimizing user conversion channels, rapidly iterating to improve the product, releasing updates early and frequently, and gradually activating monetization features to sustainably expand its market share.
I believe that Caffeine will ultimately drive the creation of millions of new applications on the internet and completely transform the entire ecosystem.
In fact, thanks to Caffeine, the number of developers working on Internet computers will soon far exceed the total number of developers in all other areas of the Web3 ecosystem, and this is just the beginning.
But that's not all. I'm also planning several other projects. One of them is UTOPIA, which will help users who can't access public cloud networks create their own private cloud networks. This will ultimately help expand the demand for internet-connected computing by establishing ICP as a universal computing standard and providing a pathway for users to enter the ecosystem.
Another project, called Convo, is currently in the secret preparation stage.
All these projects have one thing in common—they all drive the widespread use of internet-connected computers and consume enormous amounts of computing resources. We need more!
Blockchain is much more than just tokenization—for us, it's cloud computing.
By the end of 2013, I began researching how to make blockchain faster, more powerful, and more scalable. In 2014, I pioneered research on how to apply classic distributed computing technologies to the blockchain field, and may have become a leading global expert in this field in 2015 (see, for example, https://www.youtube.com/watch?v=dfGDhDR_3Gc).
My first project was called Pebble, which was originally intended to be the world's first horizontally scalable ledger, but as I began tentatively raising funds and building a team in the fall of 2014, I started to realize that some more interesting things were possible because of my involvement in early Ethereum projects.
While Ethereum plans to launch a proof-of-work network to host financial "smart contracts" for creating new tokens and supporting features such as simple decentralized token exchanges, I realize that: a) smart contracts can be redesigned as general-purpose backend software with groundbreaking new features; b) a true "world computer" can be created that is high-performance, efficient, and horizontally scalable, serving as a public cloud computing platform where hosted applications and services can truly enjoy the unique advantages of running on the blockchain.
In 2015, we realized that this required a dedicated project and a large, highly qualified team to conduct years of research and development, thus beginning our long journey. DFINITY's mission is to build blockchain into a decentralized cloud platform and an open alternative technology stack—we believe that by 2030, the cloud market is expected to reach $1.6 trillion to $2.4 trillion, and DFINITY is poised to capture a significant share of that market.
Subscribe to Dominic Williams' articles, delivered directly to your email address.
Join Medium for free to get the latest updates from this author.
subscription
In this respect, we differ from traditional Web3 projects: we hope that the Internet Computer will succeed because the network provides the world with hardcore technical utility, rather than the speculative incentives provided by the token itself.
Automatic writing application: ICP rocket has arrived and is accelerating forward.
Convincing developers outside of Web 3 to try building on new decentralized cloud platforms and open stacks like Internet Computer is a challenge, despite the fact that it offers significantly improved cloud capabilities in terms of security and resilience, reduced software development costs, and freedom from vendor lock-in.
Network effects have built a strong moat around traditional technologies. Without substantial resources, it's difficult to attract Web2 developers: they are unwilling to invest time learning new skills, unwilling to support applications on new cloud platforms while maintaining old ones, and they prefer to learn mainstream skills because more jobs require them.
However, artificial intelligence is rapidly paving the way forward.
Press Enter or click to view the full-size image.
In the future, artificial intelligence (rather than human developers) will increasingly create most new applications, websites, online services, and enterprise systems.
In the "self-written" application paradigm promoted by platforms such as the Internet Computer and Caffeine, application owners directly interact with artificial intelligence (AI), which acts as an automation technology team. In this model, application owners become new developers, as they can guide the AI through natural language interaction and share non-technical documentation.
When artificial intelligence (AI) is able to create the applications described above, development speed will increase thousands of times, and costs will decrease thousands of times. As AI becomes more intelligent and the platforms supporting AI-driven application development continue to improve, the proportion of AI applications developed under the guidance of application owners will also increase in the global application market. We are at the beginning of a transition to self-written cloud.
In the history of technology, it will no longer matter whether developers choose their technology stack based on their own needs (such as combining platform products like Amazon Web Services, Linux, Kubernetes, Postgres, and Node.js). Now, application owners will have the choice to choose the custom technology stack that provides the best performance. We are already seeing this trend.
The internet-connected computer possesses unique advantages due to its numerous benefits. Here are some of the valuable advantages it offers for independent writing:
Application complexity. After years of working to reinvent backend software in a more abstract, "serverless" form, we can now directly multiply AI's ability to meet complex backend needs, while simultaneously making it faster and cheaper to code. Over a year ago, we decided to repurpose the Motoko programming language for writing backend code, making it the world's first language designed specifically for AI. Motoko leverages unique characteristics of internet computer software, such as "orthogonal persistence," blurring the lines between logic and data. With the rapid development of this new language, platforms like Caffeine can simply fine-tune their AI clusters to use the latest features—and AI has never complained!
Application security and resilience. With the rapid development of artificial intelligence globally and the surge in the number of applications, human security and system management teams alone are simply insufficient to protect all applications. We believe the only solution is for applications to be tamper-proof, resistant to traditional cybersecurity attacks, and to ensure their logic continues to function, thereby guaranteeing data availability at all times. The Internet of Things (IoT) has provided this protection since its inception.
Application data security. In the new self-written application paradigm, AI must update applications at a chat-like pace, allowing application owners to rapidly iterate on their designs. However, for applications with backend data, major functional updates often involve migrating data to a new format, and if AI makes a mistake, some data may be lost (potentially undetectable during the update and therefore unrecoverable through rollback). This year, we launched the Internet Computer, and Motoko introduces a security mechanism to ensure that if an update would result in unexpected data loss, the update will be rejected, and AI will retry using different code.
Applications need to handle tokens. E-commerce and other applications will increasingly require the secure handling and custody of tokens. On the one hand, this enables their integration with Web 3; on the other hand, it also supports more traditional tokenization methods, such as the use of "stablecoins" (tokens pegged to fiat currencies like the US dollar). For example, AI agents will want to use stablecoins because they cannot have bank accounts; and new blockchains launched by companies like Stripe and Google reflect expectations for mainstream tokenization. Applications on internet-connected computers can securely and natively custody and handle tokens and interact with other blockchain platforms without trust—while applications on Web 2 are always at risk of being hacked.
Applications are not vendor-locked. Some virtual coding platforms, like Replit, aim to maximize AI automation, allowing semi-technical users to build applications using traditional technology stacks and export them to external cloud platforms—but technical personnel always need to be involved at some stage due to security, resilience, and reliability concerns. Platforms like Lovable, on the other hand, pursue full automation (i.e., automatic coding), developing SaaS services that host generated applications, requiring users to delegate security and other needs to them; their main drawback is vendor lock-in. In contrast, non-technical users of Caffeine can export code and run it independently on internet-connected computers (and exciting new features are under development).
To learn more about how Caffeine is built using internet computing, please visit (https://caffeine.ai/about).
Internet-connected computers will become the mainstream cloud solution.
DFINITY is refocusing its efforts to allow the market to experience internet computing in the form of non-subjective cloud products, although its underlying infrastructure will remain a public, decentralized network.
Press Enter or click to view the full-size image.
A key step is the new concept of “cloud engines.” These are essentially private subnets that users (such as entrepreneurs running large applications or companies building applications using Caffeine) can launch themselves by combining available nodes in the network through an engine configurator hosted by NNS (Network Nervous System, the governance framework for internet computer networks).
Of course, the Internet Computer will impose restrictions on which nodes can be combined to create engines, such as requiring them to be run by known independent entities from different data centers, operated by different carriers, and located in different geographical locations, depending on the specific security and resilience requirements of the engine being created (i.e., based on the existing “deterministic decentralization” approach used by the Internet Computer).
In the future, significant changes in new node standards and technologies will allow nodes to be created on cloud instances, not just dedicated bare metal. This means that virtualization technologies can be used, leveraging cloud instances from large tech companies like Amazon Web Services (AWS) and Google Cloud to create nodes. Of course, because cloud engines run on multiple cloud nodes from large tech companies, they can continue to operate normally even in the event of catastrophic cloud failures (such as the recent Amazon outage).
Therefore, the Internet computer will not adhere to preconceived notions but will allow the market to determine how engines are created and built, which can run on various nodes. (However, security-sensitive subnets, such as custodial token ledgers, chain-key tokens like Bitcoin, and subnets of the network's nervous system, will still run on dedicated sovereign hardware, which also helps ensure the network's independence.)
In the future, Internet computing engines' "customers" will experience them as revolutionary new cloud products. They will create them using a user-friendly configurator and top them up with credit cards—because the mechanism for purchasing ICP tokens and converting them into cycles can be hidden.
Traditional cloud service providers are even motivated to promote internet-connected computers and have engines run on their infrastructure.
Meanwhile, in the cloud engine space, node providers compete by promoting the use of their nodes, conducting marketing, and offering added value (such as obtaining security certifications and arranging the most reliable data center hosting for their nodes). Sometimes, node provider associations provide buttons on their websites that pre-configure engine configurators for customers, ensuring the inclusion of suitable nodes for a seamless experience.
This trend builds on Caffeine: in the future, internet-connected computers will become a new type of mainstream serverless cloud product, still driven by decentralized infrastructure, enabling the delivery of tamper-proof, unstoppable, optional autonomous, token-custodial applications that can interact with external blockchains without trust.
In this new ICP cloud engine model, engine owners can scale their applications without changing the application code—adding new nodes to the engine will scale query calls, while replacing existing nodes with more powerful nodes will scale update call capacity and the memory available to the application. If no more powerful nodes are available, the engine can be split into two new engines through generation (a process that divides the software and data units hosted on the original engine between two new engines) to continue scaling update calls and memory.
The ICP revolution is finally about to sweep through the mainstream cloud market!
Optimize the network token economic model to drive growth
Finally, we'll discuss the economics of internet computing tokens. DFINITY will strongly advocate for significantly reducing inflationary pressures, as extensive research suggests this will bring substantial advantages to the entire ecosystem. Inflation stems from two sources, and we believe both should be reduced.
The primary source of inflation is governance staking, where users stake their ICP tokens within the network's neural network framework, creating "neurons" to gain voting power. Currently, a large number of users have locked their tokens for up to 8 years, maximizing their voting power and available rewards.
A large portion of ICP token holders, including myself, have locked up most of their tokens for up to 8 years, and there's even a#8YearGangcommunity on social media. One way to reduce governance inflation is to shorten the maximum lock-up period for tokens, for example, to one or two years, and correspondingly reduce the rewards (this can also be combined with shortening the minimum lock-up period from 6 months to 1 month, thereby increasing the number of people participating in governance staking).
This idea has been proposed before and has sparked controversy, but now is a good time to re-examine its feasibility. Regarding the#8YearGangcommunity, we note that when Porsche launched its first all-electric vehicle, the Taycan, even though electric cars don't have turbochargers, its top-of-the-line models still retained the "Turbo" and "Turbo S" suffixes.#8YearGangcan continue forever.
The second source of inflation is node providers, who own and operate dedicated node machines across the globe that host the network. Currently, node providers receive a fixed monthly reward from the network (in fiat currency), which is paid out in ICP tokens (therefore, they receive more rewards when the price falls, although concerns about an inflationary spiral are unfounded, as the protocol takes the higher of the current ICP price (determined by the oracle) and 1 XDR).
While a secure value transfer mechanism makes it easier for our community of excellent node providers to operate nodes profitably and manage data center hosting contracts, it also diminishes their incentive to promote Internet Computing itself or optimize services. We hope to change this situation.
While traditional internet computer subnets will continue to operate as they do now, a different node provisioning model is possible as cloud engines play a larger role. In this model, node providers will not receive guaranteed payments, or will receive less, and will have to compete to get their nodes selected by end users and incorporated into their engines.
In this model, 80% of the ICP revenue generated by the cloud engine (calculated in terms of the number of ICP tokens required to power the engine) will be returned to the node provider, while 20% will be immediately burned. Therefore, the engine always maintains a deflationary rather than an inflationary system, and the node provider's operating model will be closer to that of a traditional enterprise, maximizing profits.
Please note that when internet-connected computer nodes are running on cloud and virtualization infrastructure, they can be started on demand, which explains why no upfront fees are required.
Of course, the path I'm describing here is just the beginning. DFINITY needs to conduct many analyses and consult with many people and groups before it can draw conclusions and propose what it believes is the best direction for the neural network.
But I've started talking about#Mission70on social media, reflecting my personal desire to keep inflation at 70% by 2026. The Internet Computer is one of the few networks that has achieved strong deflation multiple times (due to the proof-of-work network running on the Internet Computer). Our goal is to ensure it remains in strong deflation forever.
That's all for today. Go back to surfing the internet! Hello everyone,
In this article, I will share many developments and plans that are crucial to the Internet computing ecosystem.
For the past few months, I have been using the term “Internet Computer 2.0,” which relates to the growing importance of hosting self-written applications (which users create through non-technical interaction with artificial intelligence) and providing mainstream cloud functionality.
But there are other changes, including DFINITY 2.0, ongoing work to create for-profit businesses within the ecosystem, and initiatives to promote deflationary optimization of the network's token economy.
To truly understand some of our guiding reasoning, we must start from the beginning.
Our infrastructure
Back in October 2016, I founded the DFINITY Foundation in Switzerland, an independent organization dedicated to advancing internet computing and fostering its ecosystem. At that time, the project had been running for a year and a half, but real progress had only begun to accelerate in the last six months.
I believe that if there is a primary force driving the development and support of the public internet, it should be a neutral, non-profit organization, such as an independent Swiss foundation. My view remains that if the internet were owned and developed by Microsoft, it wouldn't be the internet we know today. It's far better to be led by a more neutral organization like ICANN!
Due to the involvement of venture capital, the industry has adopted different models over time, with most decentralized networks primarily guided by for-profit companies, such as Solana (Solana Labs), Avalanche (Ava Labs), Polkadot (Parity Technologies), Cardano (IOHK), Sui (Mysten Labs), Aptos (Aptos Labs), FileCoin (Protocol Labs), and so on.
(Once these companies have established their own networks, they later create foundations for their own ecosystems, but these companies usually retain de facto control, which creates a different dynamic.)
Foundation Model vs. Corporate Model
The people at DFINITY, like those for-profit organizations that support blockchain, want the Internet of Things to succeed, but their framework is different.
For-profit companies have a fiduciary duty to maximize shareholder returns and seek the best return on investment (ROI) in all their activities. Foundations, on the other hand, have a fiduciary duty to pursue predetermined goals—in the case of DFINITY, the goal is to develop the technology that powers internet-connected computers and help drive the development of the related ecosystem.
Over the years, DFINITY has increasingly operated more like a research institution than a technology company. The Ethereum Foundation exhibits a similar pattern, although DFINITY's technological advancements are much faster.
How influential for-profit ecosystem companies can provide assistance
The Ethereum ecosystem has benefited from Consensys, a large, independent for-profit company led by one of its main founders, Joe Lubin. For years, Consensys has been a driving force in its ecosystem, working to increase institutional acceptance of its DeFi business and creating products and spin-offs such as MetaMask and Infura, which have directly driven the adoption of DeFi.
While I believe decentralized networks should be led by independent nonprofit foundations, I also believe that ecosystems need strong, risk-taking for-profit companies to drive growth—and ICPs need to benefit more from this.
In light of the above, DFINITY adopted two approaches to proceed.
Making DFINITY more like a technology company
DFINITY has been working to restructure and evolve into a "DFINITY 2.0" model. ICP technology has become extremely mature and stable, and it's time to push for its large-scale market adoption.
We have developed an ambitious roadmap, which I outline in this article. We have been restructuring to become more agile, reduce management layers to deliver products faster, and focus more on user experience and productization to connect internet-connected computing with market opportunities.
This involves internal restructuring, rebalancing resource allocation across functional departments, and cultural change, as we strive to make the foundation operate more like a technology company than a purely research institution, while maintaining our deep-rooted science and technology mission.
DFINITY is entering an exciting era. We are returning to our roots, driven by our mission as innovators, and taking a proactive approach to seek new opportunities with enormous potential. This article will explore this topic.
Building impactful, profitable technology companies for the ecosystem.
Given the unique advantages of for-profit technology companies, we are also seeking to create and support for-profit projects that can independently drive the large-scale use of the Internet, computer networks, and cloud.
Caffeine was the first. It's a platform that enables users to build applications, websites, and enterprise systems on internet-connected computers by interacting with artificial intelligence using natural language and providing non-technical documentation.
Crucially, Caffeine not only possesses the appropriate architecture to meet the challenges of building entirely new mass-market platforms, but it also independently raises funds to expand its market share in the AI platform market. This is critical—DFINITY itself cannot indefinitely fund such a large-scale project.
The project nearly failed after our first attempt to create the product ended in failure, reflecting the challenges of building such a platform within a foundation focused on underlying network protocols.
In the months leading up to the official release on June 4th, I took over and created the entirely new user experience you see now, thanks to the hard work of the 996+ team. Once the user experience had the necessary potential, I assembled a completely new team like a startup and abandoned the old codebase.
Caffeine currently operates as a hardcore startup, focusing on optimizing user conversion channels, rapidly iterating to improve the product, releasing updates early and frequently, and gradually activating monetization features to sustainably expand its market share.
I believe that Caffeine will ultimately drive the creation of millions of new applications on the internet and completely transform the entire ecosystem.
In fact, thanks to Caffeine, the number of developers working on Internet computers will soon far exceed the total number of developers in all other areas of the Web3 ecosystem, and this is just the beginning.
But that's not all. I'm also planning several other projects. One of them is UTOPIA, which will help users who can't access public cloud networks create their own private cloud networks. This will ultimately help expand the demand for internet-connected computing by establishing ICP as a universal computing standard and providing a pathway for users to enter the ecosystem.
Another project, called Convo, is currently in the secret preparation stage.
All these projects have one thing in common—they all drive the widespread use of internet-connected computers and consume enormous amounts of computing resources. We need more!
Blockchain is much more than just tokenization—for us, it's cloud computing.
By the end of 2013, I began researching how to make blockchain faster, more powerful, and more scalable. In 2014, I pioneered research on how to apply classic distributed computing technologies to the blockchain field, and may have become a leading global expert in this field in 2015 (see, for example, https://www.youtube.com/watch?v=dfGDhDR_3Gc).
My first project was called Pebble, which was originally intended to be the world's first horizontally scalable ledger, but as I began tentatively raising funds and building a team in the fall of 2014, I started to realize that some more interesting things were possible because of my involvement in early Ethereum projects.
While Ethereum plans to launch a proof-of-work network to host financial "smart contracts" for creating new tokens and supporting features such as simple decentralized token exchanges, I realize that: a) smart contracts can be redesigned as general-purpose backend software with groundbreaking new features; b) a true "world computer" can be created that is high-performance, efficient, and horizontally scalable, serving as a public cloud computing platform where hosted applications and services can truly enjoy the unique advantages of running on the blockchain.
In 2015, we realized that this required a dedicated project and a large, highly qualified team to conduct years of research and development, thus beginning our long journey. DFINITY's mission is to build blockchain into a decentralized cloud platform and an open alternative technology stack—we believe that by 2030, the cloud market is expected to reach $1.6 trillion to $2.4 trillion, and DFINITY is poised to capture a significant share of that market.
Subscribe to Dominic Williams' articles, delivered directly to your email address.
Join Medium for free to get the latest updates from this author.
subscription
In this respect, we differ from traditional Web3 projects: we hope that the Internet Computer will succeed because the network provides the world with hardcore technical utility, rather than the speculative incentives provided by the token itself.
Automatic writing application: ICP rocket has arrived and is accelerating forward.
Convincing developers outside of Web 3 to try building on new decentralized cloud platforms and open stacks like Internet Computer is a challenge, despite the fact that it offers significantly improved cloud capabilities in terms of security and resilience, reduced software development costs, and freedom from vendor lock-in.
Network effects have built a strong moat around traditional technologies. Without substantial resources, it's difficult to attract Web2 developers: they are unwilling to invest time learning new skills, unwilling to support applications on new cloud platforms while maintaining old ones, and they prefer to learn mainstream skills because more jobs require them.
However, artificial intelligence is rapidly paving the way forward.
Press Enter or click to view the full-size image.
In the future, artificial intelligence (rather than human developers) will increasingly create most new applications, websites, online services, and enterprise systems.
In the "self-written" application paradigm promoted by platforms such as the Internet Computer and Caffeine, application owners directly interact with artificial intelligence (AI), which acts as an automation technology team. In this model, application owners become new developers, as they can guide the AI through natural language interaction and share non-technical documentation.
When artificial intelligence (AI) is able to create the applications described above, development speed will increase thousands of times, and costs will decrease thousands of times. As AI becomes more intelligent and the platforms supporting AI-driven application development continue to improve, the proportion of AI applications developed under the guidance of application owners will also increase in the global application market. We are at the beginning of a transition to self-written cloud.
In the history of technology, it will no longer matter whether developers choose their technology stack based on their own needs (such as combining platform products like Amazon Web Services, Linux, Kubernetes, Postgres, and Node.js). Now, application owners will have the choice to choose the custom technology stack that provides the best performance. We are already seeing this trend.
The internet-connected computer possesses unique advantages due to its numerous benefits. Here are some of the valuable advantages it offers for independent writing:
Application complexity. After years of working to reinvent backend software in a more abstract, "serverless" form, we can now directly multiply AI's ability to meet complex backend needs, while simultaneously making it faster and cheaper to code. Over a year ago, we decided to repurpose the Motoko programming language for writing backend code, making it the world's first language designed specifically for AI. Motoko leverages unique characteristics of internet computer software, such as "orthogonal persistence," blurring the lines between logic and data. With the rapid development of this new language, platforms like Caffeine can simply fine-tune their AI clusters to use the latest features—and AI has never complained!
Application security and resilience. With the rapid development of artificial intelligence globally and the surge in the number of applications, human security and system management teams alone are simply insufficient to protect all applications. We believe the only solution is for applications to be tamper-proof, resistant to traditional cybersecurity attacks, and to ensure their logic continues to function, thereby guaranteeing data availability at all times. The Internet of Things (IoT) has provided this protection since its inception.
Application data security. In the new self-written application paradigm, AI must update applications at a chat-like pace, allowing application owners to rapidly iterate on their designs. However, for applications with backend data, major functional updates often involve migrating data to a new format, and if AI makes a mistake, some data may be lost (potentially undetectable during the update and therefore unrecoverable through rollback). This year, we launched the Internet Computer, and Motoko introduces a security mechanism to ensure that if an update would result in unexpected data loss, the update will be rejected, and AI will retry using different code.
Applications need to handle tokens. E-commerce and other applications will increasingly require the secure handling and custody of tokens. On the one hand, this enables their integration with Web 3; on the other hand, it also supports more traditional tokenization methods, such as the use of "stablecoins" (tokens pegged to fiat currencies like the US dollar). For example, AI agents will want to use stablecoins because they cannot have bank accounts; and new blockchains launched by companies like Stripe and Google reflect expectations for mainstream tokenization. Applications on internet-connected computers can securely and natively custody and handle tokens and interact with other blockchain platforms without trust—while applications on Web 2 are always at risk of being hacked.
Applications are not vendor-locked. Some virtual coding platforms, like Replit, aim to maximize AI automation, allowing semi-technical users to build applications using traditional technology stacks and export them to external cloud platforms—but technical personnel always need to be involved at some stage due to security, resilience, and reliability concerns. Platforms like Lovable, on the other hand, pursue full automation (i.e., automatic coding), developing SaaS services that host generated applications, requiring users to delegate security and other needs to them; their main drawback is vendor lock-in. In contrast, non-technical users of Caffeine can export code and run it independently on internet-connected computers (and exciting new features are under development).
To learn more about how Caffeine is built using internet computing, please visit (https://caffeine.ai/about).
Internet-connected computers will become the mainstream cloud solution.
DFINITY is refocusing its efforts to allow the market to experience internet computing in the form of non-subjective cloud products, although its underlying infrastructure will remain a public, decentralized network.
Press Enter or click to view the full-size image.
A key step is the new concept of “cloud engines.” These are essentially private subnets that users (such as entrepreneurs running large applications or companies building applications using Caffeine) can launch themselves by combining available nodes in the network through an engine configurator hosted by NNS (Network Nervous System, the governance framework for internet computer networks).
Of course, the Internet Computer will impose restrictions on which nodes can be combined to create engines, such as requiring them to be run by known independent entities from different data centers, operated by different carriers, and located in different geographical locations, depending on the specific security and resilience requirements of the engine being created (i.e., based on the existing “deterministic decentralization” approach used by the Internet Computer).
In the future, significant changes in new node standards and technologies will allow nodes to be created on cloud instances, not just dedicated bare metal. This means that virtualization technologies can be used, leveraging cloud instances from large tech companies like Amazon Web Services (AWS) and Google Cloud to create nodes. Of course, because cloud engines run on multiple cloud nodes from large tech companies, they can continue to operate normally even in the event of catastrophic cloud failures (such as the recent Amazon outage).
Therefore, the Internet computer will not adhere to preconceived notions but will allow the market to determine how engines are created and built, which can run on various nodes. (However, security-sensitive subnets, such as custodial token ledgers, chain-key tokens like Bitcoin, and subnets of the network's nervous system, will still run on dedicated sovereign hardware, which also helps ensure the network's independence.)
In the future, Internet computing engines' "customers" will experience them as revolutionary new cloud products. They will create them using a user-friendly configurator and top them up with credit cards—because the mechanism for purchasing ICP tokens and converting them into cycles can be hidden.
Traditional cloud service providers are even motivated to promote internet-connected computers and have engines run on their infrastructure.
Meanwhile, in the cloud engine space, node providers compete by promoting the use of their nodes, conducting marketing, and offering added value (such as obtaining security certifications and arranging the most reliable data center hosting for their nodes). Sometimes, node provider associations provide buttons on their websites that pre-configure engine configurators for customers, ensuring the inclusion of suitable nodes for a seamless experience.
This trend builds on Caffeine: in the future, internet-connected computers will become a new type of mainstream serverless cloud product, still driven by decentralized infrastructure, enabling the delivery of tamper-proof, unstoppable, optional autonomous, token-custodial applications that can interact with external blockchains without trust.
In this new ICP cloud engine model, engine owners can scale their applications without changing the application code—adding new nodes to the engine will scale query calls, while replacing existing nodes with more powerful nodes will scale update call capacity and the memory available to the application. If no more powerful nodes are available, the engine can be split into two new engines through generation (a process that divides the software and data units hosted on the original engine between two new engines) to continue scaling update calls and memory.
The ICP revolution is finally about to sweep through the mainstream cloud market!
Optimize the network token economic model to drive growth
Finally, we'll discuss the economics of internet computing tokens. DFINITY will strongly advocate for significantly reducing inflationary pressures, as extensive research suggests this will bring substantial advantages to the entire ecosystem. Inflation stems from two sources, and we believe both should be reduced.
The primary source of inflation is governance staking, where users stake their ICP tokens within the network's neural network framework, creating "neurons" to gain voting power. Currently, a large number of users have locked their tokens for up to 8 years, maximizing their voting power and available rewards.
A large portion of ICP token holders, including myself, have locked up most of their tokens for up to 8 years, and there's even a#8YearGangcommunity on social media. One way to reduce governance inflation is to shorten the maximum lock-up period for tokens, for example, to one or two years, and correspondingly reduce the rewards (this can also be combined with shortening the minimum lock-up period from 6 months to 1 month, thereby increasing the number of people participating in governance staking).
This idea has been proposed before and has sparked controversy, but now is a good time to re-examine its feasibility. Regarding the#8YearGangcommunity, we note that when Porsche launched its first all-electric vehicle, the Taycan, even though electric cars don't have turbochargers, its top-of-the-line models still retained the "Turbo" and "Turbo S" suffixes.#8YearGangcan continue forever.
The second source of inflation is node providers, who own and operate dedicated node machines across the globe that host the network. Currently, node providers receive a fixed monthly reward from the network (in fiat currency), which is paid out in ICP tokens (therefore, they receive more rewards when the price falls, although concerns about an inflationary spiral are unfounded, as the protocol takes the higher of the current ICP price (determined by the oracle) and 1 XDR).
While a secure value transfer mechanism makes it easier for our community of excellent node providers to operate nodes profitably and manage data center hosting contracts, it also diminishes their incentive to promote Internet Computing itself or optimize services. We hope to change this situation.
While traditional internet computer subnets will continue to operate as they do now, a different node provisioning model is possible as cloud engines play a larger role. In this model, node providers will not receive guaranteed payments, or will receive less, and will have to compete to get their nodes selected by end users and incorporated into their engines.
In this model, 80% of the ICP revenue generated by the cloud engine (calculated in terms of the number of ICP tokens required to power the engine) will be returned to the node provider, while 20% will be immediately burned. Therefore, the engine always maintains a deflationary rather than an inflationary system, and the node provider's operating model will be closer to that of a traditional enterprise, maximizing profits.
Please note that when internet-connected computer nodes are running on cloud and virtualization infrastructure, they can be started on demand, which explains why no upfront fees are required.
Of course, the path I'm describing here is just the beginning. DFINITY needs to conduct many analyses and consult with many people and groups before it can draw conclusions and propose what it believes is the best direction for the neural network.
But I've started talking about#Mission70on social media, reflecting my personal desire to keep inflation at 70% by 2026. The Internet Computer is one of the few networks that has achieved strong deflation multiple times (due to the proof-of-work network running on the Internet Computer). Our goal is to ensure it remains in strong deflation forever.
That's all for today. Go back to surfing the internet!