AI x Crypto, a16z's latest report reveals 11 disruptive fusion scenarios

Source: a16z

Original Title: AI x crypto crossovers

Compiled and organized by: BitpushNews

The economic model of the internet has begun to change. As the open web gradually integrates into a prompt bar, we can't help but ask: Will AI bring an open internet or a maze filled with new paywalls? Who will control it - large centralized companies or the vast user community?

This is where cryptocurrency comes into play. We have discussed the intersection of AI and cryptocurrency many times; simply put, blockchain is a completely new way to build internet services and networks that are decentralized, trustworthy and neutral, and user-owned. They provide checks and balances to many centralized powers in the AI systems we have already seen by renegotiating the economic models that underpin today's systems, helping to achieve a more open and robust internet.

Cryptocurrency can help build better AI systems, and vice versa—this idea is not new, but it often lacks clear definition. Some cross-disciplinary areas—such as verifying "human proof" in light of the proliferation of low-cost AI systems—have already attracted builders and users. However, other use cases seem to require years, or even decades, to realize. Therefore, in this article, we share 11 use cases at the intersection of cryptocurrency and AI, hoping to inspire dialogue around possibilities, challenges to be addressed, and more. They are all rooted in the technology being built today, from handling large volumes of micropayments to ensuring that humans retain control over their relationship with future AI.

1. Persistent Data and Context in AI Interaction

Generative AI relies on data, but for many applications, context – the state and background information related to interactions – is equally important, if not more so.

In an ideal scenario, AI systems—whether they are agents, LLM interfaces, or other applications—should remember many details such as the type of projects you are working on, your communication style, and your preferred programming language. However, in practice, users often need to re-establish this context across different interactions within a single application—such as when starting a new ChatGPT or Claude Shell—not to mention switching between different systems.

Currently, there is little context from a generative AI application that can be ported to other applications (if any ever existed).

By leveraging blockchain, AI systems can enable key contextual elements to exist as persistent digital assets that can be loaded at the beginning of a conversation and seamlessly transferred between AI platforms. Furthermore, given that interoperability is a defining attribute based on blockchain protocols, blockchain may be the only solution that is both forward-compatible and can establish interoperability commitments to address this issue.

A natural application in this area is AI-mediated games and media, where preferences (from difficulty levels to key bindings) can persist across different games and environments. But the real value lies in knowledge application, where AI needs to understand what the user knows and how they learn; as well as more specialized AI use cases such as coding. Of course, various businesses have developed their own custom robots that have a global context specific to the given business—but in this case, the context is often not transferable, even between different AI systems used within the organization.

Organizations have just begun to understand this issue, and the closest universal solution we currently see is custom bots with fixed, persistent context. However, the portability of context between users within the platform is beginning to emerge off-chain; for example, Poe allows users to rent out their custom bots to others.

Bringing such activities on-chain will enable us to interact with AI systems that can share a contextual layer containing key elements from all our digital activities. They will instantly understand our preferences and be able to better fine-tune and optimize our experiences. In turn, just like the on-chain intellectual property registry, this allows AI to reference persistent on-chain context, creating the possibility for new and better market interactions around prompts and information modules— for example, users can directly license or monetize their expertise while retaining custody of their data. Of course, sharing context will make many things we have yet to conceive possible.

2. General Identity of the Agent

Identity, which is a normative record of what things are, is the invisible pipeline of today's digital discovery, aggregation, and payment systems. Since platforms hide this pipeline behind walls, we experience identity in a finished form: Amazon assigns identifiers (ASIN or FNSKU) to products, lists them in one place, and helps users discover and pay. Facebook is similar: a user's identity is the basis for their dynamic messages and the discovery across applications (including Facebook Marketplace listings, organic posts, and paid ads).

With the advancement of AI agents, everything will change. As more and more companies use agents for customer service, logistics, payments, and other use cases, their platforms will increasingly resemble not just single-interface applications. Instead, they will exist across multiple canvases and platforms, accumulating deep context and performing more tasks for users. However, binding the identity of the agent to just one market will render it unusable in other important places (email threads, Slack channels, and within other products).

This is why agents need a single, portable "passport". Without it, there's no way to know how to pay the agent, verify its version, query its functions, understand who the agent works for, or track its reputation across applications and platforms. The identity of the agent needs to serve as a wallet, API registry, changelog, and social proof—thus any interface (email, Slack, another agent) can parse and communicate with it in the same way. Without this shared primitive of "identity", every integration would need to rebuild this pipeline from scratch, discovering that it remains temporary, and users would lose context every time they switch channels or platforms.

We have the opportunity to design agent infrastructure from first principles. So, how do we build a richer, trust-neutral identity layer than DNS records? Agents should be able to accept payments, list functionalities, and exist across multiple ecosystems without the worry of being locked into any specific platform. This is precisely where the intersection of cryptocurrency and AI is particularly useful, as blockchain networks provide permissionless composability, which can help developers create more useful agents and better user experiences.

Generally speaking, vertically integrated solutions (such as Facebook or Amazon) currently offer a better user experience—one of the inherent complexities of building an excellent product is ensuring that all parts work together seamlessly from top to bottom. However, this convenience comes at a high cost, especially as the costs of building aggregation, marketing, monetization, and distribution agent software decrease, and the surface area of agent applications expands. Matching the user experience of vertically integrated providers requires effort, but providing agents with a trustworthy neutral identity layer would allow entrepreneurs to have their own passport—and encourage experimentation in distribution and design.

3. Forward-Compatible Human Proof

As AI becomes increasingly prevalent—driving robots and agents in various online interactions, including deepfakes and social media manipulation—it is becoming more difficult to determine whether the entities you interact with online are real human beings. This erosion of trust is not a concern for the future; it already exists. From armies of dynamic comments on X to bots on dating apps, reality is starting to blur. In this environment, human proofing has become an essential infrastructure.

One way to prove you are human is through a digital ID (including centralized IDs used by TSA). A digital ID contains everything a person can use to verify their identity—username, PIN, password, and third-party proof (such as citizenship or creditworthiness) along with other credentials. The value of decentralization is evident here: when this data exists in a centralized system, the issuer can revoke access, impose fees, or foster surveillance. Decentralization reverses this situation: users, rather than platform gatekeepers, control their own identities, making it more secure and resistant to censorship.

Unlike traditional identity systems, decentralized proof of humanity mechanisms (such as Worldcoin's Proof of Human) allow users to control and manage their identities in a privacy-preserving and trust-neutral manner, while verifying their human identity. Just like a driver's license, which can be used anywhere regardless of where it was issued, decentralized proof of humanity (PoP) can serve as a reusable foundational layer applicable to any platform, including those that do not yet exist. In other words, blockchain-based PoP is forward-compatible because it provides:

• Portability: The protocol is a public standard that can be integrated by any platform. Decentralized PoP can be managed through public infrastructure and controlled by users. This makes it fully portable, and any platform can be compatible with it now or in the future.
• Permissionless Access: The platform can independently choose to recognize PoP IDs without relying on gatekeeper APIs that may discriminate against different use cases.

The challenge in this field lies in adoption: while we have yet to see many real-world human proof use cases of meaningful scale, we anticipate that a large number of users, a few early partners, and killer applications will accelerate adoption. Every application that utilizes a given digital ID standard will make that ID type more valuable to users; this will attract more users to acquire the ID; in turn, this makes the ID more attractive to applications as a means of verifying human identity. (And because on-chain IDs are inherently interoperable, these network effects can grow rapidly.)

We have seen mainstream consumer applications and services in gaming, dating, and social media announce partnerships with World ID to help humans know they are interacting with real people—specifically, the particular individuals they expect—while gaming, chatting, and trading. We have also seen new identity protocols emerge this year, including Solana Proof Service (SAS). Although SAS is not an issuer of human proofs, it allows users to privately associate off-chain data (such as compliance-related KYC checks or certification status for investment) with their Solana wallet to build a decentralized identity for the user. All of this suggests that the inflection point for decentralized PoP may not be far off.

Human proof is not just to prohibit robots, but to establish clear boundaries between AI agents and human networks. It allows users and applications to distinguish between human and machine interactions, creating space for a better, safer, and more authentic digital experience.

4. Decentralized Physical Infrastructure for AI (DePIN)

AI may be a digital service, but its advancement is increasingly constrained by the bottlenecks of physical infrastructure. Decentralized Physical Infrastructure Networks, or DePIN—which provides a new model for building and operating real-world systems—can help democratize access to the underlying computing infrastructure for AI innovations, making it cheaper, more resilient, and more censorship-resistant.

How to achieve this? The two biggest barriers to AI advancement are energy and chip access. Decentralized energy can help provide more power, but developers are also using DePIN to aggregate unused chips from gaming PCs, data centers, and other sources. These computers can come together to 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 LLMs, as well as distributed networks for model inference. Decentralized training and inference have the potential to significantly lower costs as they utilize otherwise idle computing resources. They can also provide censorship resistance, ensuring that developers are not deplatformed by hyperscale cloud service providers (centralized cloud service providers that offer large-scale scalable computing infrastructure).

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

5. Infrastructure and protection for interaction between AI agents, terminal service providers, and users

As AI tools become increasingly adept at solving complex tasks and executing multi-layer interaction chains, AI will interact more and more with other AIs without human controllers.

For example, AI agents may need to request specific data related to calculations or recruit specialized AI agents to perform specific tasks—such as assigning a statistical bot to develop and run model simulations, or having an image generation bot participate in creating marketing materials. AI agents will also create significant value in completing the entire transaction process or any other activity on behalf of users—for example, finding and booking flights according to someone's preferences, or discovering and ordering new books in their favorite genre.

Currently, there is no established universal broker market—such cross-querying mainly occurs through explicit API connections, or within an AI agent ecosystem where broker calls are maintained as internal functions.

More broadly, most AI agents today operate in isolated ecosystems, with APIs being relatively closed and generally lacking architectural standardization. However, blockchain technology can help protocols establish open standards, which is crucial for adoption in the short term. In the long term, this also supports backward compatibility: as new types of AI agents evolve and are created, they can expect to connect to the same underlying network. Given the interoperable, open-source, decentralized, and often more easily upgradeable architecture of blockchain, they can more readily adapt to novel AI innovations.

As the market develops, many companies have been building blockchain tracks for agent-to-agent interactions: for example, Halliday recently launched its protocol, which provides a standardized, cross-chain architecture for AI workflows and interactions—offering protocol-level protections to ensure that AI does not exceed the user's intentions. Meanwhile, Catena, Skyfire, and Nevermind use blockchain to enable one AI agent to pay another AI agent without human intervention. More such systems are under development, and Coinbase has even begun to provide infrastructure support for these efforts.

6. Keep AI/Vibe coding applications in sync

The recent revolution in generative AI has made building software easier than ever before. Coding speed has increased by several orders of magnitude, and—perhaps most importantly—can be done in natural language, allowing even inexperienced programmers to derive existing programs and build new ones from scratch.

However, despite the new opportunities created by AI-assisted coding, it has also introduced a significant amount of entropy within and across programs. "Vibe coding" abstracts the complex web of dependencies underlying software — but this also makes programs susceptible to functional and security flaws when the source libraries and other inputs change. Meanwhile, as individuals use AI to create their own custom applications and workflows, it becomes increasingly difficult for them to interact with others' systems. In fact, even two vibe coding programs that perform the same task may have very different operating and output structures.

Historically, standardization was initially provided by file formats and operating systems to ensure consistency and compatibility, and more recently by shared software and API integration. However, in a world where software evolves, morphs, and branches in real-time, the standardization layer needs to be widely accessible and continuously upgradable - while also maintaining user trust. Furthermore, relying solely on AI cannot solve the problem of incentivizing people to build and maintain these connections.

Blockchain simultaneously addresses these two issues: a protocolized synchronization layer, which is encapsulated in custom software builds and dynamically updated to ensure cross-compatibility as things change. Historically, a large enterprise might spend millions of dollars hiring a "system integrator" like Deloitte to customize a Salesforce instance. Today, engineers can create custom interfaces to view sales information in a weekend, but as the number of custom software increases, developers will need help keeping these applications synchronized and running.

This is similar to the development approach of today’s open-source software libraries, but with continuous updates instead of periodic releases—and with incentive packaging. Both become easier to achieve through cryptocurrency. Like other blockchain-based protocols, the shared ownership of the synchronization layer incentivizes active investment in its improvement. Developers, users (and/or their AI agents), and other consumers can earn rewards by introducing, using, and developing new features and integrations.

Conversely, shared ownership gives all users a stake in the overall success of the protocol, which can buffer against malicious behavior. Just as Microsoft would not undermine the .docx file standard due to its impact on users and the brand, the co-owners of the sync layer would not introduce clumsy or malicious code into the protocol.

As with all standardized software architectures we've seen before, there is also huge potential for network effects here. With the Cambrian explosion of AI coding software continuing, the network of heterogeneous and diverse systems that need to communicate with each other will expand dramatically. In short: vibe coding requires more than just vibe to stay in sync. Cryptocurrency is the answer.

7. Micro-payments with Revenue Sharing Support

AI agents and tools like ChatGPT, Claude, and Copilot promise a convenient new way to navigate the digital world. But whether good or bad, they are also disrupting the economics of the open internet. We are already seeing this happen— for example, educational platforms are experiencing a significant drop in traffic as students increasingly use AI tools, and several U.S. newspapers are suing OpenAI for copyright infringement. If we do not realign incentives, we may see an increasingly closed internet, more paywalls, and fewer content creators.

Of course, there will always be policy solutions, but as these solutions go through the courts, some technical solutions are emerging. Perhaps the most promising (and technically complex) solution is to build an income-sharing system into the architecture of the network. When AI-driven actions lead to sales, the content source that facilitated that decision should receive a share. The affiliate marketing ecosystem has already done attribution tracking and revenue sharing like this; a more complex version could automatically track and reward all contributors in the information chain. Blockchain can clearly play a role in tracking that provenance chain.

However, such a system still requires a new infrastructure for other functions—specifically, micropayment systems capable of handling small transactions across multiple sources, attribution protocols that fairly assess different types of contributions, and governance models that ensure 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 potential in this regard, enabling near-zero-cost transactions and more granular payment splits.

Blockchain will achieve a complex proxy payment system through the following mechanisms:

• Nano Payments can be distributed to multiple data providers, allowing a single user to interact and trigger small payments to all contributing sources through automated smart contracts.
• Smart contracts allow for enforceable retroactive payments to be triggered after a transaction is completed, compensating the information sources that facilitated the purchasing decision in a fully transparent and traceable manner.
• In addition, blockchain supports complex and programmable payment splitting, ensuring that income is fairly distributed through rules enforced by code rather than centralized decision-making, thereby creating trustless financial relationships among autonomous agents.

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

8. Blockchain as a Registry for Intellectual Property and Traceability

Generative AI has created an urgent demand for efficient and programmable mechanisms for registering and tracking intellectual property—both to establish provenance and to support business models around IP access, sharing, and remixing. Existing IP frameworks—which rely on costly intermediaries and post facto enforcement—are unable to adapt to a world where AI instantly consumes content and generates new variants at the push of a button.

What we need is an open public registry that provides clear proof of ownership, allowing IP creators to interact with it easily and efficiently—while AI and other web applications can interact with it directly. Blockchains are an ideal choice in this regard, as they allow for the registration of IP without relying on intermediaries and provide immutable proof of provenance; they also enable third-party applications to identify, license, and interact with the IP directly.

It is understandable that there are many doubts about the whole idea of technology being able to protect IP in some way, as the first two eras of the internet—and the ongoing AI revolution—are often associated with a decline in intellectual property protection. One issue is that many of today's IP-based business models focus on excluding derivative works rather than trying to incentivize and monetize them. However, programmable IP infrastructure not only allows creators, franchises, and brands to clearly establish ownership of their IP in the digital space—but it also opens the door for business models centered around the explicit use of shared IP for generative AI and other digital applications. In fact, this transforms one of the main threats that generative AI poses to creative works into an opportunity.

We have seen creators early on in the NFT space experiment with newer models, with companies leveraging NFT assets on Ethereum to support network effects and value accumulation under CC0 brand building. Recently, we have seen infrastructure providers building protocols, even dedicated blockchains (such as Story Protocol), for standardized and composable IP registration and licensing. Some artists have begun using these tools to license their styles and works for creative remixing through protocols like Alias, Neura, and Titles. Incention's Emergence series works to track who created what through a blockchain registry built on the Story protocol, thereby attracting fans to co-create a sci-fi universe and its characters.

9. Help compensate content creators' web crawlers

Nowadays, the AI agent that best fits the product market is neither for coding nor for entertainment. It is a web crawler—autonomously navigating the web, collecting data, and deciding which links to follow.

It is estimated that currently nearly half of internet traffic does not originate from humans. Bots often disregard the conventions of robots.txt—a file that is supposed to inform automated web crawlers whether they are welcome, but in practice has little authority—and use the data they extract to bolster the defenses of some of the largest tech companies in the world. Worse still, websites ultimately have to foot the bill for these uninvited guests, paying for bandwidth and service CPU resources to what feels like an endless faceless scraping. In response, companies like Cloudflare and other CDNs (Content Delivery Networks) offer blocking services. This is a patchwork of a service that should not exist.

We have previously argued that the original protocol of the internet—the economic contract between content creators and distribution platforms—is likely to collapse. This is already reflected in the data: over the past twelve months, website owners have begun to block AI crawlers on a large scale. As of July 2024, about 9% of the top 10,000 websites banned AI crawlers, and that number is now 37%. As more website operators mature and users continue to feel frustrated, this number will only increase.

So, what if we take a middle ground instead of paying the CDN to completely block anything that looks like a bot? AI bots could pay for the right to collect data instead of occupying systems designed to bring human traffic to websites for free. This is where blockchain comes in: in this scenario, each web crawler agent would hold some cryptocurrency and negotiate on-chain with each website's "gatekeeper" agent or paywall protocol via x402. (Of course, the challenge lies in the robots.txt system, also known as the robot exclusion standard, which has been deeply embedded in the business models of internet companies since the 1990s. Overcoming this requires large-scale collective coordination or the involvement of CDNs like Cloudflare.)

However, humans can prove their identity as humans through a separate channel using World ID (as mentioned above) and access content for free. This way, content creators and website owners can receive compensation for their contributions to large AI datasets at data collection points, while humans can continue to enjoy a free and open internet.

10. Personalized and Privacy-Protecting Advertising

AI has already begun to influence the way we shop online, but what if the ads we see every day are... really helpful? There are many obvious reasons why people dislike ads. Irrelevant ads are purely noise. At the same time, not all personalization is created equal. AI-driven ads, if too targeted—extracted from a vast amount of consumer data—can feel intrusive. Other applications try to monetize by limiting content (like streaming services or game levels), which is blocked by non-skippable ads.

Cryptocurrencies can help address some of these issues and provide an opportunity to reimagine how advertising works. Combined with blockchain, personalized AI agents can bridge the gap between irrelevant and bizarre by serving ads based on user-defined preferences. Importantly, they can achieve this while not globally exposing user data, and directly compensating users who share data or engage with ads.

Some technical requirements here include:

• Low-cost digital payments: To compensate users for interacting with ads (viewing, clicking, converting), the company needs to send small, frequent payments. To achieve large-scale operations, we need a fast, high-throughput system with negligible costs.
• Privacy-Preserving Data Verification: AI agents need to be able to demonstrate that consumers meet certain demographic attributes. Zero-knowledge proofs can verify demographic attributes while preserving privacy.
• Incentive Model: If the internet embraces monetization based on micropayments (for example, less than $0.05 per interaction, as mentioned above), users will be able to opt into advertisements in exchange for small payments, reversing the current model from extraction to participation.

For decades, people have been striving to make online advertising relevant—and offline advertising has been doing the same for centuries. But rethinking advertising through the lens of cryptocurrency and AI could ultimately make advertising more useful. Personalized yet not creepy, and in a way that benefits everyone: it unlocks a more sustainable and consistent new incentive structure for developers and advertisers. For users, it offers more ways to discover and navigate the digital world.

All of this will make advertising space more valuable, not less. It can also replace today's entrenched, extractive advertising economy with a more human-centered system: one that views users as participants rather than products.

11. AI companions owned and controlled by humans

Many people spend more time on devices than on face-to-face interactions, and this time is increasingly spent interacting with AI models and AI-curated content (especially). All of these models have provided a form of companionship, whether for entertainment, information, satisfying niche interests, or educating children. It is not hard to imagine that in the near future, AI companions used for education, healthcare, legal consultation, and friendship will become a popular mode of interaction among humans.

Future AI companions will have infinite patience and will be customized according to specific individuals and their particular use cases. They are not just helpers or robotic servants; they may become very valuable relationships. Therefore, who will own and control these relationships—whether users, companies, or other intermediaries—becomes equally important. If you have been concerned about the curation and censorship of social media over the past decade, this issue will become exponentially more complex and more personalized in the future.

A censorship-resistant hosting platform like blockchain provides the most compelling path for achieving censorship-free, user-controlled artificial intelligence, and this is not a new argument. Indeed, individuals can run device models and purchase their own GPUs, but most either cannot afford it or simply do not know how to operate it.

Although we have a long way to go before the widespread adoption of AI companions, all these technologies are rapidly improving: text-based companions that seem human-like are already quite impressive. Visual representations have also seen significant improvements. The performance of blockchain is also increasing. To ensure that uncensored companions are easy to use, we need to rely on better user experiences to achieve crypto applications. Fortunately, wallets like Phantom make interacting with the blockchain much simpler, and embedded wallets, keys, and account abstraction allow users to hold self-custodial wallets without the complexities of storing seed phrases themselves. Technologies like high-throughput optimistic and ZK rollups will also make it possible to establish meaningful and lasting relationships with digital companions.

In the near future, the focus of people's discussions will shift from when we can see lifelike digital companions and avatars to who and what can control them.