Author: Scott Duke Kominers, Sam Broner, Jay Drain, Guy Wuollet, Elizabeth Harkavy, Carra Wu, Matt Gleason
Compiled by: Saoirse, Foresight News
The economic logic of the Internet has quietly changed. As the open network gradually shrinks into a "command input bar", we have to think: Will artificial intelligence lead us to an open Internet, or will it fall into a maze of new payment barriers? Will control be in the hands of large centralized companies or a broad user base?
This is where Crypto comes in. We have discussed the intersection of AI and Crypto many times before - in short, blockchain is a new paradigm for reconstructing the Internet service architecture, which can build a decentralized, trusted, neutral and user-owned network system. By redefining the economic rules that support the existing system, blockchain provides an effective way to check and balance the centralized trend in the field of AI, so as to achieve a more open and resilient Internet ecosystem.
The idea of two-way empowerment of Crypto and AI systems has long existed, but the combination of the two has always lacked a clear definition. Some cross-cutting areas (such as "human identity" verification in the context of the proliferation of low-cost AI tools) have attracted the attention of developers and users, while other application scenarios may take years or even decades to land. Therefore, this article will share 11 cross-border application scenarios of AI and Crypto, aiming to promote relevant discussions: explore the potential possibilities and challenges of combining AI and Crypto, and look forward to more innovative directions. These scenarios are all based on the current level of technology, covering a variety of fields from massive micropayment processing to ensuring that humans have dominance in future AI interactions.
Identity and Data Management
1. Persistent Data and Context in AI Interactions
By Scott Duke Kominers
The development of generative AI is highly dependent on data, but in many application scenarios, context (i.e., interaction-related status and background information) is just as important as data, or even more critical.
Ideally, whether it is an agent, a large language model interface, or other application, the AI system can remember many details such as the user's work type, communication style, preferred programming language, etc. However, in reality, users often need to reset these contexts in different sessions of the same application (such as starting a new ChatGPT or Claude session), let alone switching between different systems. Currently, the context of a generative AI application is almost impossible to migrate to other applications.
With the help of blockchain technology, AI systems can transform key contextual elements into persistent digital assets, which can be loaded when a session starts and seamlessly transferred between different AI platforms. In addition, based on its characteristics, blockchain may be the only solution that meets the requirements of "forward compatibility" and "interoperability" at the same time.
This is particularly true in AI-mediated gaming and media, where user preferences (from difficulty settings to key bindings) can be kept consistent across games and scenarios. But the real value is in knowledge application scenarios (where AI needs to understand the user’s knowledge and learning patterns) and specialized AI applications (such as programming assistance). Of course, some companies have developed custom robots for specific business contexts, but in such scenarios, context is often not portable across systems, and is even difficult to share between different AI tools within the company.
Organizations are just beginning to realize this problem, and the current common solution is a custom robot with a fixed context. However, context transfer between users within the platform has begun to emerge off-chain: for example, on the Poe platform, users can rent out custom robots to others for use.
By putting such scenarios on the chain, the AI systems we interact with will be able to share a contextual layer that contains the key elements of all digital activities. They will immediately understand user preferences and optimize the user experience more accurately. Conversely, such as the on-chain intellectual property registration system, allowing AI to reference persistent on-chain context also creates the possibility of new market-based interactions around prompt words and information modules - users can directly authorize or commercialize their own expertise while retaining data control. Of course, shared context will also give rise to many possibilities that we have not yet foreseen.
2. Universal identity system for agents
By Sam Broner
Identity (i.e., “the authoritative record of something’s essential attributes”) is the underlying architecture that supports today’s digital discovery, aggregation, and payment systems. As platforms enclose this architecture within their ecosystem walls, identity in the eyes of users has become part of the product’s functionality: Amazon assigns unique identifiers (ASIN or FNSKU) to products, centrally displays products, and assists users in discovery and payment; the same is true for Facebook, where user identity is the core foundation of its information flow and the entire in-app discovery function, including product listings, native posts, and paid ads.
This landscape is about to change as AI agents evolve. As more companies adopt agents (for scenarios such as customer service, logistics management, payment processing, etc.), their platforms will no longer be limited to single-interface applications, but will span multi-platform ecosystems, accumulate deep context, and perform more diverse tasks for users. However, if the agent identity is tied to a single market, it will lose usability in other key scenarios (such as email threads, Slack channels, and other products).
Therefore, agents need a single, portable "digital passport". Without this passport, there will be no way to determine how to pay the agent, verify its version information, query its functional attributes, know who it serves, or trace its reputation record between different applications and platforms. The agent identity needs to have multiple functions such as wallet, API registry, update log and social proof to ensure that any interface (email, Slack or other agents) can be parsed and interacted with unified standards. Without the shared information of "identity", each system integration needs to build the underlying architecture from scratch, the discovery mechanism will always be in a temporary state, and the user will lose contextual information every time he switches channels or platforms.
We have the opportunity to design proxy infrastructure from first principles. So how do we build a trusted neutral identity layer that is better than DNS records? Proxies should not repeat the mistakes of monolithic platforms that “bind identity to discovery, aggregation, and payment functions”, but should be able to accept payments and display functions in multiple ecosystems without worrying about being locked into a specific platform. This is the value of the cross-border integration of Crypto and AI - the permissionless composability provided by blockchain networks can help developers build more practical proxies and better user experiences.
In general, vertically integrated solutions (such as Facebook or Amazon) currently have a better user experience. One of the inherent challenges of building great products is to ensure that the components work together from top to bottom. But this convenience comes at a high price, especially as the cost of building agent aggregation, marketing, monetization, and distribution software continues to decline, and the coverage of agent applications continues to expand. Although there is still work to do to achieve the user experience level of vertically integrated platforms, building a trusted and neutral identity layer for agents will enable entrepreneurs to take control of their "digital passports" and promote innovative exploration at the distribution and design levels.
3. Forward-compatible “human identity” proof mechanism
By Jay Drain Jr., Scott Duke Kominers
As AI technology penetrates into various online interaction scenarios (including deep fakes and social media manipulation), it becomes increasingly difficult to determine whether one is interacting with a real person online. The collapse of this trust system has already occurred - from the comment army on the X platform (formerly Twitter) to the robot accounts in dating apps, the boundary between virtual and reality is gradually blurring. In this context, proof of "human identity" has become the core infrastructure of the digital ecosystem.
One way to prove you’re human is with a digital ID (including the centralized ones used by the TSA). A digital ID includes everything from usernames, PINs, passwords, third-party proofs (like citizenship or credit ratings) that can be used to authenticate your identity. Decentralization is valuable here: when data is stored in a centralized system, issuers may revoke access, charge extra fees, or implement surveillance; a decentralized model reverses this logic — users (not platform managers) have control over their identities, making them more secure and censorship-resistant.
Unlike traditional identity systems, decentralized "human identity" proof mechanisms (such as World's Proof of Human) allow users to manage their identity information independently and verify "human attributes" in a privacy-preserving and trusted neutral manner. Just as a driver's license can be used in any region (regardless of when and where it is issued), decentralized "human identity" proof can be used as a universal underlying protocol across platforms, even for emerging platforms that have not yet been born. In other words, blockchain-based "human identity" proof has forward-looking compatibility because it has the following advantages:
- Portability: The relevant protocols are open standards and can be integrated into any platform. Decentralized "human identity" proof can be managed through public infrastructure and controlled by users. It is fully portable and compatible with any platform now or in the future;
- Permissionless accessibility: Platforms can choose to recognize “human identity” IDs on their own without going through APIs that may discriminate against different use cases.
The challenge in this area is implementation: although there are no real-scale “human identity” proof application scenarios, we expect adoption to accelerate as the number of users reaches critical mass, early partnerships are formed, and killer applications emerge. Each application that adopts a specific digital ID standard increases the value of that ID to users, which in turn attracts more users to obtain IDs, forming a positive cycle (and because on-chain IDs are interoperable by design, network effects can accumulate quickly).
We have seen mainstream consumer applications such as games, dating, and social media announce cooperation with World ID to help users confirm that they are playing games, chatting, and trading with real people (not programs); this year, new identity protocols such as Solana Attestation Service (SAS) have emerged. Although SAS is not the issuer of "human identity" proof, it allows users to privately associate off-chain data (such as KYC compliance or investment qualifications) with Solana wallets to help build a decentralized identity system. All these signs indicate that the turning point of decentralized "human identity" proof may not be far away.
The significance of "human identity" proof is not only to ban robots, but also to draw a clear boundary between AI agents and human networks. It enables users and applications to distinguish between human and machine interactions, thereby creating a better, safer and more authentic digital experience.
Decentralized Infrastructure
4. Decentralized Infrastructure in AI (DePIN)
By Guy Wuollet
Although AI is a digital service, its development is increasingly constrained by physical infrastructure. Decentralized Infrastructure Network (DePIN) provides a new model for building and operating real-world systems, helping to democratize the computing infrastructure behind AI innovation, making it more economical, resilient, and censorship-resistant.
How to achieve this goal? The two core challenges facing AI development are computing power supply and chip acquisition. Decentralized computing power networks can provide more computing power, and developers are also using DePIN to aggregate idle chip resources from gaming PCs, data centers, and other sources. These computing devices can form an unlicensed computing market, creating a fair competitive environment for the development of new AI products.
Other application scenarios include distributed training and fine-tuning of large language models, and distributed networks for model reasoning. Decentralized training and reasoning (due to the use of idle computing resources) can significantly reduce costs while providing censorship resistance to ensure that developers will not be terminated by hyperscale cloud service providers (such as centralized cloud service giants).
The problem of a few companies monopolizing AI models has long existed, and decentralized networks will help build a more economical, censorship-resistant, and scalable AI ecosystem.
5. Infrastructure and rules framework for interaction between AI agents, end service providers and users
By Scott Duke Kominers
As AI tools become more capable of solving complex tasks and executing multi-level chains of interactions, there will be an increasing need for AI systems to interact with other AI systems without human intervention.
For example, an AI agent may need to request specific data related to a computational task, or recruit specialized AI agents to complete a specific task (such as assigning a statistical robot to develop and run model simulations, or calling an image generation robot when creating marketing materials). AI agents will also create significant value in completing transactions on behalf of users or other activities, such as finding and booking flights based on user preferences, or discovering and ordering new books in their favorite genres.
Today, there is no mature general-purpose agent-to-agent interaction market, and such cross-system queries are mostly only possible through explicit API connections or in AI ecosystems that use agent-to-agent calls as internal functions.
In general, most AI agents today operate in isolated ecosystems with relatively closed API interfaces and a lack of architectural standardization. However, blockchain technology can help protocols establish open standards, which is crucial for short-term application implementation; in the long run, this also supports forward-looking compatibility - as new AI agents evolve and emerge, they can access the same underlying network. Blockchain can more flexibly adapt to the innovation needs in the field of AI due to its interoperable, open source, decentralized and generally easier to upgrade architectural characteristics.
As the market develops, many companies have begun to build blockchain infrastructure for agent-to-agent interactions: for example, Halliday recently launched a related protocol to provide a standardized cross-chain architecture for AI workflows and interactions, and provide protection mechanisms at the protocol layer to ensure that AI behavior does not exceed user intent; Catena, Skyfire, and Nevermind use blockchain technology to support automatic payments between AI agents without human intervention. More such systems are under development, and Coinbase has even begun to provide infrastructure support for these explorations.
6. Ensure synchronization of AI/custom programming applications
Written by: Sam Broner, Scott Duke Kominers
The innovation of generative AI has enabled a qualitative leap in software development efficiency: coding speed has increased by several orders of magnitude, and most importantly, it can be done through natural language - even inexperienced programmers can replicate existing programs or build new applications from scratch.
But while AI-assisted coding creates new opportunities, it also introduces a lot of uncertainty inside and outside the program. "Vibe coding" abstracts the complex network of dependencies underlying the software, but this also makes the program vulnerable to functional and security vulnerabilities when the source library and other inputs change. In addition, when people use AI to create personalized applications and workflows, it becomes more difficult to interact with other people's systems - in fact, even two "vibe coding" programs with the same function may have significant differences in their operating logic and output structure.
Historically, the standardization work of ensuring software consistency and compatibility was first undertaken by file formats and operating systems, and in recent years has relied on shared software and API integration. However, in the new era of real-time software evolution, iteration, and branching, the standardization layer needs to be widely accessible and continuously upgradable while maintaining user trust. In addition, AI alone cannot solve the problem of "motivating people to build and maintain these connections."
Blockchain technology solves both problems at the same time: the protocolized synchronization layer can be embedded in the user's custom software architecture and dynamically updated to ensure cross-system compatibility as the environment changes. Historically, large companies may pay millions of dollars to "system integrators" such as Deloitte to customize Salesforce instances. Today, engineers can create custom interfaces for viewing sales information on weekends, but as the number of customized software grows, developers need professional support to keep these applications running in sync. (Note: Salesforce is a customer relationship management (CRM) software service provider founded in the United States in March 1999)
This is similar to the development model of open source software libraries today, but with continuous updates (rather than periodic releases) and incentives - both of which are easier to achieve with Crypto technology. As with other blockchain-based protocols, the shared ownership mechanism of the sync layer incentivizes all parties to actively invest in improvements: developers, users (and their AI agents), and other consumers can be rewarded for introducing, using, and optimizing new features and integrations.
Conversely, shared ownership binds all users tightly to the overall success of the protocol, creating a buffer against malicious behavior — just as Microsoft would be reluctant to break the .docx file standard (due to the knock-on effects on users and brand), the co-owners of the sync layer will be less inclined to introduce inefficient or malicious code into the protocol.
As with all the software standardization architectures we have seen, there is huge potential for network effects here. As the "Cambrian Explosion" of AI coding software continues to evolve, the network of heterogeneous systems that need to maintain communication will expand exponentially. In short: "Custom programming" requires not only "coding style", but also Crypto technology to keep the system synchronized.
New Economy and Incentive Model
7. Micropayment system to support revenue sharing
By Liz Harkavy
AI agents and tools like ChatGPT, Claude, and Copilot offer new and convenient ways to navigate the digital world, but for better or worse, they are shaking the economic foundations of the open internet. We are already seeing concrete manifestations of this trend - for example, educational platforms have seen a sharp drop in traffic due to the heavy use of AI tools by students, and several US newspapers are suing OpenAI for alleged copyright infringement. Without a realignment of incentives, we may face an increasingly closed internet with more paywalls and fewer content creators.
Of course, policy solutions exist, but as they make their way through the judicial process, a range of technical solutions are emerging. Perhaps the most promising (and technically challenging) solution is to embed revenue sharing mechanisms into the network architecture: when AI-driven behavior leads to a transaction, the content source that provided information to support that decision should receive a corresponding share. The affiliate marketing ecosystem is already doing similar attribution tracking and revenue sharing, and more advanced versions can automatically track and reward all contributors in the information chain - blockchain technology can obviously play a key role in tracking this traceability chain.
But such systems will also require new infrastructure with other capabilities — in particular, micropayment systems that can handle microtransactions across multiple sources, attribution protocols that fairly assess the value of different contributions, and governance models that ensure transparency and fairness. Many existing blockchain tools (such as Rollups and Layer2, AI-native financial institutions Catena Labs, and financial infrastructure protocols 0xSplits) have shown potential for application, supporting near-zero-cost transactions and more granular payment splitting.
Blockchain can implement complex proxy payment systems through a variety of mechanisms:
- Nanopayments can be split across multiple data providers, allowing a single user interaction to trigger micropayments to all contributing sources via automated smart contracts.
- Smart contracts enable triggering of executable retroactive payments after a transaction is completed, compensating the sources that informed the purchasing decision in a fully transparent and traceable manner.
- In addition, blockchain supports complex and programmable payment splitting and allocation, ensuring that revenue is distributed fairly through code-enforced rules rather than relying on centralized decision-making, creating trustless financial relationships between autonomous agents.
As these emerging technologies mature, they will create new economic models for the media industry, capturing the entire value creation chain from creators to platforms to users.
8. Blockchain as an intellectual property and traceability registration system
By Scott Duke Kominers
The development of generative AI has created an urgent need for efficient and programmable intellectual property registration and tracking mechanisms - both to clarify ownership and to support business models around access, sharing and re-creation of intellectual property. The existing intellectual property protection framework relies on expensive intermediaries and post-action enforcement measures, which cannot adapt to the needs of the era where AI consumes content instantly and generates new variants at the click of a button.
What we need is an open public registration system that provides clear proof of ownership, enables efficient interaction between intellectual property creators, and allows direct connection between AI and other network applications. Blockchain technology is an ideal choice: it can complete intellectual property registration without intermediaries, provide tamper-proof traceability, and enable third-party applications to easily identify, authorize and use the intellectual property.
Some are skeptical of the idea that technology can protect IP — after all, the first two eras of the internet (and the ongoing AI revolution) are often associated with weakened IP protections. This is partly because many of today’s IP-based business models focus on prohibiting derivative works rather than incentivizing and commercializing derivative creations. But programmable IP infrastructure not only enables creators, brands, and IP owners to clearly establish ownership rights in the digital space, it also opens the door to business models around IP sharing (for generative AI and other digital applications) — effectively turning generative AI’s primary threat to creativity into an opportunity.
We have seen creators experiment with new models in the NFT space early on: businesses are leveraging NFT assets on Ethereum to support network effects and value accumulation under CC0 branding; recently, infrastructure providers are also building protocols (such as Story Protocol) and even dedicated blockchains for standardized and composable IP registration and licensing. Some artists have begun using these tools to license their styles and works for creative re-creation through protocols such as Alias, Neura, and Titles. Incention's Emergence series invites fans to co-create a sci-fi universe and its characters, and the blockchain registry built on Story Protocol can track the creators of each element.
9. Web crawler mechanism to compensate content creators
By Carra Wu
Today, the most marketable AI agents are not programming or entertainment tools, but web crawlers—they autonomously browse the web, collect data, and decide where to crawl.
It is estimated that nearly half of all web traffic today originates from non-human agents. Crawlers often ignore robots.txt (a file that is supposed to tell automated crawlers whether to access a website, but has little actual binding force) and use the data they scrape to strengthen the market barriers of tech giants. Worse, websites have to pay for these uninvited guests and bear the cost of providing bandwidth and CPU resources to a large number of unidentified crawlers. In response, CDNs (content distribution networks) such as Cloudflare provide blocking services, but this is just a patchwork solution that should not exist.
We have argued that the Internet’s native protocols (the economic agreements between content creators and distribution platforms) may be breaking down, and the data is showing that this is happening. Over the past 12 months, website owners have been blocking AI bots en masse: in July 2024, only 9% of the world’s top 10,000 websites banned AI bots, but now that figure has reached 37%, and will continue to rise as more website operators upgrade their technology and user dissatisfaction grows.
If we don’t rely on CDN to completely block the access of suspected crawlers, can we find a compromise? AI crawlers should not use the system designed for human traffic for free, but should pay for the right to crawl data. This is where blockchain comes in: each web crawler agent can hold Crypto and negotiate on-chain with the "access agent" or paywall protocol of each website through the x402 protocol (of course, the challenge is that the robots.txt protocol has been deeply rooted in the business logic of the Internet since the 1990s, and it requires large-scale collaboration or the participation of CDNs such as Cloudflare to break through).
At the same time, humans can access content for free by proving their identity through World ID (see Chapter 3). In this model, content creators and website owners can be compensated when AI data sets are collected, while human users can still enjoy the "information freedom" of the Internet.
The future of AI ownership
10. Privacy-preserving personalized advertising
By Matt Gleason
AI has begun to affect the online shopping experience, but what if the ads we see every day can be "really useful"? The reason why people dislike ads is obvious: ineffective ads are pure noise, and overly accurate AI ads based on massive consumer data appear to violate privacy. Other apps make profits by limiting content (such as streaming services or game levels) through "unskippable ads."
Crypto makes it possible to reconstruct the advertising model. Personalized AI agents combined with blockchain can find a balance between "irrelevant advertising" and "overly precise advertising" and deliver advertising based on user-defined preferences. More importantly, this model does not need to expose the user's global data, and can directly compensate users who actively share data or interact with advertisements.
To achieve this goal, the following technical requirements must be met:
- Low-fee digital payments: To compensate users for ad interactions (views, clicks, conversions), companies need to send small payments at high frequencies, which requires systems with high-speed processing capabilities and near-zero fees.
- Privacy-preserving data verification: AI agents need to prove that users meet certain demographic attributes, and zero-knowledge proofs can complete attribute verification while protecting privacy;
- Incentive model: If the Internet adopts a profit model based on micropayments (such as less than $0.05 per interaction, see Chapter 7), users can choose to "watch ads in exchange for small rewards", transforming the existing "exploitative" model into a "participatory" model.
For decades, online (and even offline for hundreds of years) advertising has always been in pursuit of "relevance". Reconstructing advertising from the perspective of Crypto and AI will eventually make it more practical - customized but not intrusive, benefiting all parties: for developers and advertisers, unlocking more sustainable and incentive-compatible business models; for users, gaining more ways to explore the digital world.
This will not only increase the value of advertising space, but is also likely to overturn today’s entrenched “exploitative” advertising economy and build a more people-oriented system - users are no longer commodities to be traded, but rather participants.
11. AI companions owned and controlled by humans
By Guy Wuollet
Today, people spend more time on their devices than offline interactions, and increasingly interact with AI models and AI-generated content. These models have begun to provide companionship value, whether it is entertainment, information acquisition, satisfying niche interests, or educating children. It is not difficult to imagine that in the near future, AI companions for education, medical care, legal advice, and emotional companionship will become the mainstream interaction method.
The AI companions of the future will be infinitely patient and customized to individual needs — they will no longer be just tools or robot servants, but will likely become highly valued relationships. Therefore, the question of who owns and controls these relationships is crucial (the user, or an intermediary such as a company). If you have been concerned about content filtering and censorship on social media for the past decade, the issue will become more complex and more personal in the future.
The idea that a censorship-resistant blockchain hosting platform is the most viable path to user-controlled AI has been discussed many times (as mentioned above). In theory, individuals can run on-device models or purchase their own GPUs, but most people either cannot afford it or lack the technical ability.
Although it will take time to popularize AI companions, related technologies are rapidly iterating: anthropomorphic text interaction companions are already quite mature, visual avatar technology has made significant progress, and blockchain performance is constantly improving. To ensure that censorship-resistant companions are easy to use, we need to rely on better user experience to implement encryption-driven applications. Fortunately, wallets such as Phantom have greatly simplified blockchain interactions, and embedded wallets, cryptographic keys, and account abstraction technologies allow users to hold self-hosted wallets without seed phrase. High-throughput trusted computing technology based on Optimistic and ZK coprocessors will also help build a deep and lasting relationship with a digital companion.
In the near future, the discussion will shift from when realistic digital companions will appear to who will control them.
