11 AI and Crypto Integration Scenarios Shaping the Future of the Internet

Written by: a16z crypto

Translated by AididiaoJP, Foresight News

The economic model of the internet is changing. As the open internet increasingly shrinks into a prompt box, we can't help but wonder: Will AI bring a more open internet or a new paid maze? Who will control it — large centralized companies or the broader user community?

This is precisely where cryptocurrency comes into play. We have discussed multiple times the intersection of AI and cryptocurrency. To summarize briefly, blockchain provides a completely new way to build internet services and networks. It is decentralized, trustlessly neutral, and can be owned by users. By reshaping the economic foundation of current systems, it can counterbalance the increasingly centralized trends we are seeing in AI, helping to build a more open and robust internet.

The idea that cryptocurrencies can help build better AI systems—and vice versa—is not new, but it is often vaguely defined. Some overlapping areas, such as verifying "human identity" in an era of low-cost AI, have already attracted developers and users. However, other use cases appear to be years or even decades away from realization. Therefore, this article lists 11 use cases where AI and cryptocurrencies intersect, aiming to spark discussions on their feasibility and unresolved challenges. These concepts are all based on technologies currently under development, ranging from handling massive volumes of microtransactions to ensuring that humans maintain control over their relationships with future AI.

Generative AI relies on data, but for many applications, context (i.e., the state and background information relevant to interactions) is equally important, if not more so.

Ideally, an AI system (whether it's an agent, a large language model interface, or another application) should remember details about the projects you're working on, your communication style, preferred programming languages, and more. In reality, however, users often have to repeatedly reconstruct this context across different sessions within the same application (e.g., every time they open a new ChatGPT or Claude window) or even between different systems.

Currently, the context of a generative AI application is almost impossible to transfer to another application.

With the help of blockchain, AI systems can convert key contextual elements into persistent digital assets. These assets can be loaded at the beginning of a conversation and seamlessly transferred across different AI platforms. Moreover, given that interoperability and forward compatibility are fundamental characteristics of blockchain protocols, it may be the only solution to address this issue and establish a commitment to interoperability.

A natural application scenario is AI-mediated games and media, where user preferences—from difficulty levels to key configurations—can be consistently maintained across different games and environments. However, the true value lies in educational applications, where AI needs to understand what users know and how they learn. It also lies in more specialized use cases, such as programming. Of course, individual companies have already customized bots with global context for their businesses, but such context is typically non-transferable, even between different AI systems within the same organization.

The closest current general solution is a customized bot with a fixed, persistent context. However, context portability between users within platforms has started to emerge in an off-chain form. For example, the Poe platform allows users to rent out their customized bots to others.

Bringing such activities on-chain will enable the AI systems we interact with to share a "context layer" that aggregates key information from all our digital activities. AI will instantly understand our preferences, thereby better optimizing our experience. In return, just as intellectual property registration is made possible on-chain, allowing AI to reference persistent on-chain context also opens the door to innovative market interactions around prompts and information modules. For example, users can directly monetize or license their expertise while retaining control over their data. Of course, shared context will also enable many future applications we have yet to imagine.

Identity is standardized information that records "who or what," serving as the "invisible pipeline" that supports today's digital discovery, aggregation, and payment systems. Because platforms enclose these pipelines within their walls, the identities we experience are part of the finished product: Amazon assigns identifiers (ASIN or FNSKU) to products, lists them in one place, and helps users discover and pay for them. Similarly, Facebook uses user identity as the foundation for all its discovery features, including the Facebook Marketplace, organic posts, and paid advertisements.

With the development of AI agents, all of this is about to change. The more companies use agents in scenarios such as customer service, logistics, and payments, the less their platforms will resemble single-interface applications. Instead, agents will operate across multiple interfaces and platforms, accumulating deep contextual information and performing more tasks for users. However, if an agent's identity is bound to a single market, it will not be usable in other important scenarios.

Therefore, an agent needs a single, portable "passport." Without it, there would be no way to know how to pay the agent, verify its version, query its capabilities, understand who it works for, or track its reputation across different applications and platforms. An agent's identity must serve as a wallet, API registry, changelog, and social proof, enabling any interface (email, Slack, other agents) to identify and interact with it consistently. Without this shared foundational component of "identity," every integration would require rebuilding these pipelines from scratch, leading to ad-hoc and fragmented functionality, and users would lose context every time they switch channels or platforms.

We have the opportunity to redesign the agent infrastructure from first principles. So, how can we build an identity layer that is richer and more trustlessly neutral than DNS records? Agents should be able to accept payments, showcase their capabilities, and exist across multiple ecosystems without being locked into any specific platform—rather than rebuilding a "monolithic" platform that bundles identity with discovery, aggregation, and payments. This is precisely where the combination of cryptocurrency and AI becomes especially powerful, as blockchain networks provide permissionless composability, enabling developers to create more useful agents and better user experiences.

In general, vertically integrated solutions like Facebook or Amazon currently offer a better user experience. Part of the complexity in building excellent products lies in ensuring that all components are seamlessly aligned from top to bottom. However, the cost of this convenience is high—especially as the software costs for building aggregators, marketing, monetizing, and distributing agents decrease, and as agent applications expand in scope. Matching the user experience of vertically integrated providers requires effort, but a trusted, neutral agent identity layer would give entrepreneurs their own "passport," encouraging more experimentation in distribution and design.

As AI permeates various online interactions—including deepfakes and social media manipulation—it is becoming increasingly difficult to determine whether we are interacting with real humans online. This erosion of trust is not a future concern; it is already here: from comment brigades on the X platform to bots on dating apps, reality is becoming blurred. In such an environment, "human verification" becomes critical infrastructure.

One way to prove human identity is through digital IDs (including centralized IDs used by the U.S. Transportation Security Administration). Digital IDs contain all information that can be used to verify identity, such as usernames, PINs, passwords, and third-party attestations (like nationality or credit status). The value of decentralization is evident here: when this data resides in a centralized system, the issuer can revoke access at any time, charge fees, or facilitate surveillance. Decentralization reverses this dynamic: users, rather than the platform, control their own identities, making them more secure and resistant to censorship.

Unlike traditional identity systems, decentralized human verification mechanisms (such as Worldcoin's Proof of Personhood) allow users to control and manage their own identities, while verifying their human attributes in a privacy-preserving and trustlessly neutral manner. Just as a driver's license is universally valid regardless of when or where it is issued, decentralized human verification can serve as a reusable foundational layer applicable to any platform—including platforms that have not yet been created. In other words, blockchain-based human verification is forward-compatible, as it provides:

• Portability: The protocol is a public standard and can be integrated on any platform. Decentralized human proof can be managed through public infrastructure and controlled by users. This makes it fully portable, ensuring compatibility with any current or future platform.
• Accessibility without permission: The platform can independently choose to recognize this human proof ID without needing to go through a "gatekeeper" API that might discriminate against different use cases.

The challenge in this field lies in adoption. Although we have not yet seen many real-world applications at scale, we anticipate that reaching a critical mass of users, a few early partners, and a killer application will accelerate its adoption. Each application that adopts a specific digital identity standard makes that identity more valuable to users, thereby attracting more users to obtain that identity, which in turn makes the identity more attractive to additional applications (as a means of verifying humans). Since on-chain IDs are designed to be interoperable, this network effect can grow rapidly.

We have already seen mainstream consumer applications and services in the gaming, dating, and social media sectors announce partnerships with World ID to help users verify that they are interacting with real people (and the specific real person they expect) while playing games, chatting, and transacting. New identity protocols have also emerged this year, such as the Solana Attestation Service (SAS). While SAS does not issue proofs of humanity, it allows users to privately associate off-chain data—such as KYC checks or investment eligibility verification required for compliance—with their Solana wallets, enabling the construction of users' decentralized identities. All of this suggests that a turning point for decentralized proofs of humanity may be just around the corner.

Human verification is not only for blocking bots, but also for clearly distinguishing between AI agents and human networks. It enables users and applications to differentiate between human and machine interactions, paving the way for better, safer, and more authentic digital experiences.

Although AI is a digital service, its development is increasingly constrained by physical infrastructure. Decentralized Physical Infrastructure Networks (DePIN) offer a new model for building and operating physical systems, helping to acquire the computational infrastructure that supports AI innovation, making it more affordable, resilient, and resistant to censorship.

How can this be achieved? Two major obstacles in AI development are energy and access to chips. Decentralized energy can help provide more electricity, while developers are leveraging DePIN to aggregate idle chips from sources like gaming computers and data centers. These computers can collectively form a permissionless computing market, creating a fair competitive environment for building new AI products.

Other use cases include distributed training and fine-tuning of large language models, as well as distributed networks for model inference. Decentralized training and inference could significantly reduce costs by utilizing otherwise idle computational resources. They also offer censorship resistance, ensuring that developers are not deprived of platform access by hyperscale cloud providers (large centralized cloud providers offering highly scalable computational infrastructure).

The concentration of AI models in the hands of a few companies remains a persistent concern; decentralized networks help create AI that is more cost-effective, censorship-resistant, and scalable.

As AI tools become more capable of solving complex tasks and executing multi-step interaction chains, AI will increasingly need to interact with other AIs without human intervention.

For example, an AI agent may need to request specific data related to computations or recruit specialized AI agents for particular tasks (such as assigning a statistics bot to develop and run model simulations, or enlisting an image generation bot to participate in creating marketing materials). AI agents will also create significant value by completing entire transaction processes or other activities on behalf of users, such as searching for and booking flights based on preferences, or discovering and ordering new books from preferred genres.

Currently, there is no mature general market for interactions between intelligent agents. Most of these cross-platform queries are conducted either through explicit API connections or within closed ecosystems that support inter-agent invocation.

More broadly, most AI agents today operate within isolated ecosystems, with relatively closed APIs and a general lack of architectural standardization. However, blockchain technology can help establish open protocol standards, which are crucial for short-term adoption. In the long run, it also supports forward compatibility: as new types of AI agents evolve and emerge, they can be expected to connect to the same underlying network. Given its interoperable, open-source, decentralized, and often easily upgradable architecture, blockchain can more flexibly adapt to AI innovation.

As the market develops, many companies are already building blockchain "tracks" for interactions between intelligent agents. For example, Halliday recently launched its protocol, offering a standardized cross-chain architecture for AI workflows and interactions, with protocol-level safeguards to prevent AI from deviating from user intent. Meanwhile, companies like Catena, Skyfire, and Nevermind are leveraging blockchain to enable direct payments between agents without human intervention. More such systems are under development, and even Coinbase has begun providing infrastructure support for these efforts.

The recent generative AI revolution has made software development easier than ever before. Coding speed has increased by orders of magnitude, and most importantly, it can now be done using natural language, enabling even inexperienced programmers to fork existing programs or create new ones from scratch.

However, while AI-assisted coding brings new opportunities, it also introduces a significant amount of "entropy increase" (disorder) both within and between programs. The vibe framework abstracts away the complex dependency networks in software foundations, but this can also make programs prone to functional and security flaws when inputs such as source code repositories change. At the same time, when people use AI to create personalized applications and workflows, integration with others' systems becomes challenging. In fact, even two vibe applications with identical functionalities may differ greatly in their internal operations and output structures.

Historically, the work of ensuring consistency and compatibility was first undertaken by file formats and operating systems, and more recently by shared software and API integrations. However, in a world where software continuously evolves, transforms, and branches in real time, the standardization layer must be widely accessible, continuously upgradable, and still maintain user trust. Moreover, AI alone cannot address the issue of motivating people to build and maintain these connections.

Blockchain simultaneously provides solutions to both of these problems: a protocol-based synchronization layer. It can be embedded into people's custom software and dynamically updated to ensure cross-platform compatibility as changes occur. In the past, a large company might have spent millions hiring a "systems integrator" like Deloitte to customize a Salesforce instance. Today, an engineer can spend a weekend creating a custom interface to view sales information. However, as the number of custom software applications grows, developers will need assistance to keep these applications synchronized and running smoothly.

This is similar to the development model of current open-source software libraries, with the difference being the need for continuous updates rather than periodic releases, and the inclusion of an incentive layer. Both of these are more easily achieved with the help of cryptocurrencies. Like other blockchain-based protocols, shared ownership of the synchronization layer incentivizes people to actively invest in improving it. Developers, users (or their AI agents), and other consumers can be rewarded for introducing, using, and developing new features and integrations.

Conversely, shared ownership ties all users to the overall success or failure of the protocol, effectively deterring undesirable behavior. Just as Microsoft is unwilling to damage the .docx file standard, which would harm users and its brand, co-owners of the sync layer are similarly reluctant to introduce clumsy or malicious code into the protocol.

Just like all the software standardization architectures we've seen before, there is tremendous potential for network effects here. As the "Cambrian explosion" of AI coding software continues, a rapidly expanding network of heterogeneous, diverse systems that need to communicate with each other will emerge. In short, "atmospheric coding" cannot stay in sync just by relying on "atmosphere." Cryptocurrency is the solution.

AI agents and tools like ChatGPT, Claude, and Copilot promise a new, convenient way to navigate the digital world. But, for better or worse, they are also shaking up the economic model of the open internet. We are already seeing the effects: for example, as students increasingly use AI tools, traffic to educational platforms has significantly declined; several U.S. newspapers are suing OpenAI for copyright infringement. Without realigning incentives, we may face an increasingly closed internet: more paywalls and fewer content creators.

Of course, there are always policy solutions, but as legal processes move forward, many technical solutions are also emerging. Perhaps the most promising (and technically complex) solution is to embed a revenue-sharing system into the architecture of the web. When AI-driven actions lead to sales, the content sources that provided the information for those decisions should receive a share of the revenue. The affiliate marketing ecosystem already performs similar attribution tracking and revenue sharing; a more sophisticated version could automatically track and reward all contributors in the information chain. Blockchain technology is clearly well-suited to play a role in tracking these chains of sources.

However, such a system would also require new infrastructure with additional features: specifically, a micropayment system capable of handling small transactions from multiple sources, attribution protocols that fairly evaluate different contributions, and a governance model that ensures transparency and fairness. Many existing blockchain-based tools—such as rollups and L2s, AI-native financial institutions like Catena Labs, and financial infrastructure protocols like 0xSplits—show promise in enabling near-zero-cost transactions and more granular payment splitting.

Blockchain will achieve complex agent payment systems through several mechanisms:

• Nano-payments can be distributed to multiple data providers: Through automated smart contracts, a single user interaction can trigger tiny payments flowing to all contributing sources.
• Smart contracts support traceable payments: After a transaction is completed, smart contracts can enforce retroactive payments, transparently and traceably compensating the information sources that contributed to the purchasing decision.
• Implement complex, programmable payment splitting: Ensure income is fairly distributed through code-enforced rules rather than centralized decisions, establishing trustless financial relationships among autonomous agents.

As these emerging technologies mature, they can create a new economic model for the media, capturing the complete value chain from creators to platforms and ultimately to users.

Generative AI urgently needs efficient and programmable mechanisms to register and track intellectual property—mechanisms that not only verify sources but also pave the way for business models involving IP access, sharing, and remixing. Existing IP frameworks rely on costly intermediaries and post-hoc enforcement, which are no longer suitable for an era where AI instantly consumes content and generates new variations with a single click.

What we need is an open and public registry that can provide clear proof of ownership, allowing IP creators to easily and efficiently interact with it, while also enabling AI and other web applications to directly integrate with it. Blockchain is the ideal choice because it enables IP registration without intermediaries, provides tamper-proof provenance records, and allows third-party applications to easily identify, license, and interact with the IP.

It is understandable that there is considerable skepticism regarding the overall idea that "technology can somehow protect intellectual property (IP)," given that the first two eras of the internet and the ongoing AI revolution have often been accompanied by weakened IP protection. One issue is that many IP-based business models today focus on preventing derivative works rather than encouraging them and monetizing them. However, programmable IP infrastructure not only enables creators, brands, and IP holders to clearly establish ownership in the digital space, but also opens the door to new business models around sharing IP in digital applications such as generative AI.

We have already seen creators experimenting with new models in the early days of the NFT space, with companies leveraging NFT assets on Ethereum to support network effects and value accumulation under the CC0 branding strategy. Recently, we have observed infrastructure providers building protocols, and even dedicated blockchains (such as Story Protocol), to standardize and enable composable intellectual property (IP) registration and licensing. Some artists have already begun using protocols like Alias, Neura, and Titles to license their styles and works for creative remixing. Meanwhile, Incention's Emergence series has enabled its fans to co-create a science fiction universe and characters, using a blockchain registry built on Story to record authorship and contributions.

Today, the AI agent with the highest product-market fit is not used for coding or entertainment, but for web crawling—it autonomously browses the web, collects data, and decides which links to follow.

It is estimated that nearly half of all internet traffic now originates from non-human sources. Bots often ignore the robots.txt protocol (a file intended to inform automated crawlers whether they are welcome, but which in practice has very low authority), and use the scraped data to strengthen the moats of some of the world's largest technology companies. Worse still, websites ultimately end up paying for these uninvited guests, providing bandwidth and computing resources for seemingly endless anonymous scrapers. In response, companies like Cloudflare and other content delivery networks offer blocking services—patchwork solutions that should not even be necessary.

We have pointed out that the original internet contract—the economic agreement between content creators and distribution platforms—is likely to break down. Data is beginning to reflect this: over the past year, website owners have started to block AI crawlers on a large scale. In July 2024, only about 9% of the top 10,000 websites blocked AI crawlers; today, that figure has risen to 37%. As more website operators upgrade their technology and users continue to express frustration, this percentage is likely to keep increasing.

What if we don't pay the CDN to directly block all suspected robot traffic, but instead seek a middle ground? AI robots could pay for data collection rights rather than freely using systems designed to attract human traffic. This is where blockchain comes into play: in this scenario, each web crawler agent holds some cryptocurrency and engages in on-chain negotiations with a "gatekeeper" agent or paywall protocol at each website via x402.

Meanwhile, humans can prove their identity through a World ID on another channel to access content for free. This allows content creators and website owners to be compensated for their contributions to large AI datasets at the data collection point, while humans can continue to enjoy an internet where information is meant to be free.

AI has already begun to influence the way we shop online. But what if the advertisements we see every day were actually useful? People dislike ads for many obvious reasons. Irrelevant ads are simply noise, and not all personalization is satisfactory. AI-driven advertisements that are overly precise, based on massive amounts of consumer data, can feel intrusive and privacy-violating. Other applications attempt to monetize content through non-skippable ads.

Cryptocurrencies can help address some issues and provide opportunities to reshape the advertising model. Combined with blockchain technology, personalized AI agents can bridge the gap between irrelevant ads and overly precise targeting that feels intrusive. These agents can deliver advertisements based on user-defined preferences. The key is that they can do this without globally exposing user data, and they can directly compensate users who share their data or interact with advertisements.

This requires some technical requirements:

• Low-cost digital payments: To compensate users for ad interactions (views, clicks, conversions), companies need to make small, frequent payments. To scale operations, we require a fast, high-throughput system with extremely low costs.
• Privacy-preserving data verification: AI agents need the ability to verify that consumers meet certain demographic attributes. Zero-knowledge proofs can validate such attributes while preserving privacy.
• Incentive Mechanism: If the internet adopts a monetization model based on micro-payments (e.g., less than $0.05 per interaction), users could choose to watch ads in exchange for small payments, shifting the current model from "exploitation" to "participation."

People have been trying to make online advertising relevant for decades, but rethinking advertising through the lens of cryptocurrency and AI could ultimately make it genuinely useful. It can be personalized without being annoying, and in a way that benefits all parties: for builders and advertisers, it unlocks new, more sustainable and consistent incentive structures; for users, it provides more ways to discover and explore the digital world.

All of this will make advertising space more valuable rather than devaluing it. It could also replace today's deeply entrenched, extractive advertising economy with a more human-centered system—one that treats users as participants rather than products.

Many people now spend more time on devices than in face-to-face interactions, and an increasing portion of this time is spent interacting with AI models and AI-curated content. These models already offer a form of companionship, whether through entertainment, information, catering to niche interests, or educating children. It is not hard to imagine that, in the near future, AI-based education, healthcare, legal advice, and even companionship will become popular ways for humans to interact.

Future AI companions will possess unlimited patience and be customized for specific individuals and their use cases. They will not only be assistants or robotic servants but may become cherished relationships. Therefore, who will own and control these relationships—the users or the companies and other intermediaries—becomes equally important. If you have been concerned about content moderation and censorship on social media over the past decade, this issue will become exponentially more complex and personal in the future.

Existing arguments suggest that censorship-resistant hosting platforms like blockchain represent the clearest path toward user-controlled, uncensorable AI. Indeed, individuals can run on-device models and purchase their own GPUs, but most people either cannot afford them or lack the technical know-how.

Although AI companions are not yet widely adopted, the relevant technologies are advancing rapidly: text-based, human-like companions are already quite impressive; visual representations have improved significantly; and blockchain performance continues to enhance. To ensure censorship-resistant companions are user-friendly, we need cryptocurrency applications that offer better user experiences. Fortunately, wallets like Phantom have already made interacting with blockchains much simpler, and embedded wallets, passkeys, and account abstraction allow users to hold self-custody wallets without managing complex mnemonics themselves. Leveraging high-throughput, trustless technologies such as optimistic and zero-knowledge co-processors will also make it possible to form meaningful and lasting relationships with digital companions.

In the near future, the focus of discussion will shift from "When will we see lifelike digital companions and avatars?" to "Who and what will be able to control them?"