Source: a16z
Original title: AI x crypto crossovers
Compiled and edited by: BitpushNews
The economics of the internet have begun to change . As the open web slowly merges into a reminder bar, we can’t help but ask: Will AI lead to an open internet, or a maze of new paywalls? Who will control it — large centralized companies, or a broad community of users?
This is where cryptocurrencies come in. We’ve discussed the intersection of AI and cryptocurrencies many times; in short, blockchains are a completely new way to build internet services and networks that are decentralized , trusted, neutral , and user-owned . They help enable a more open, robust internet by renegotiating the economic models that underpin today’s systems, providing checks and balances on many of the centralized forces we’ve seen in AI systems.
The idea that cryptocurrency can help build better AI systems, and vice versa, is not new, but it often lacks clear definition. Some intersections—for example, verifying “proof of humanness” given the proliferation of low-cost AI systems—are already attracting builders and users. But other use cases seem years, if not decades, away. So in this post, we share 11 use cases at the intersection of cryptocurrency and AI to inspire a conversation about possibilities, challenges to be solved, and more. They’re all rooted in technologies being built today , from processing large volumes of micropayments to ensuring humans have relational control over future AI.
1. Persistent Data and Context in AI Interactions
Generative AI relies on data, but for many applications, context—the state and background information relevant to the interaction—is equally or more important.
Ideally, an AI system — whether an agent, LLM interface, or other application — should remember details like the type of project you’re working on, your communication style, and your preferred programming language. But in practice, users often need to re-establish this context across different interactions within a single application — such as when launching a new ChatGPT or Claude Shell — not to mention switching between different systems.
Currently, context from one generative AI application is rarely, if ever, portable to other applications.
With blockchain, AI systems can enable key contextual elements to exist as persistent digital assets that can be loaded at the start of a session and transferred seamlessly between AI platforms. Furthermore, given that interoperability is a defining property of blockchain-based protocols, blockchain may be the only solution to this problem that is both forward-compatible and builds on the promise of interoperability .
A natural application of this is AI-mediated gaming and media, where preferences (from difficulty levels to key bindings) can persist across different games and environments. But the real value lies in knowledge applications, where the AI needs to understand what the user knows and how they learn; and in more specialized AI use cases, such as coding. Of course, individual businesses have developed their own custom bots that have a global context specific to a given business - but in this case, the context is often not portable, even between different AI systems used within an organization.
Organizations are only beginning to understand this problem, and the closest general solution we see today is custom bots with fixed, persistent context. But portability of context between users within a platform is starting to emerge off-chain; for example, Poe allows users to rent out their custom bots to others.
Bringing such activities on-chain will allow us to interact with AI systems that can share a contextual layer that encompasses key elements of all our digital activities. They will instantly understand our preferences and be able to better fine-tune and optimize our experiences. In turn, enabling AI to reference persistent on-chain context, like an on-chain intellectual property registry , creates the possibility of new and better market interactions around tips and information modules - for example, users can directly license or monetize their expertise while retaining custody of their data. Of course, shared context will enable many things we have not yet conceived.
2. General identity of the agent
Identity, the canonical record of what a thing is, is the invisible plumbing of today’s digital discovery, aggregation, and payment systems. Because platforms hide this plumbing behind walls, we experience identity in the form of finished products: Amazon assigns an identifier (ASIN or FNSKU) to a product, lists it in one place, and helps users discover and pay for it. Facebook is similar: a user’s identity is the foundation of their News Feed and discovery across apps, including Facebook Marketplace listings, organic posts, and paid ads.
As AI agents advance, all that will change. As more companies use agents for customer service, logistics, payments, and other use cases, their platforms will look less and less like single-interface applications. Instead, they will exist across multiple canvases and platforms, accumulating deep context and performing more tasks for users. However, tying an agent’s identity to just one marketplace makes it unusable in other important places—in email threads, Slack channels, and inside other products.
That’s 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 capabilities, understand who the agent works on behalf of, or track its reputation across applications and platforms. The agent’s identity needs to act as a wallet , an API registry, a change log, and social proof — so any interface (email, Slack, another agent) can resolve and communicate with it the same way. Without this shared primitive, every integration would need to rebuild this pipeline from scratch, discovery would remain ad hoc, and users would lose context every time they switched channels or platforms.
We have an opportunity to design proxy infrastructure from first principles. So how do we build a trusted neutral identity layer that is richer than a DNS record? Proxies should be able to accept payments, list functionality, and exist in multiple ecosystems without worrying about being locked into any specific platform. This is where the intersection of cryptocurrency and AI is particularly useful, as blockchain networks offer permissionless composability , which can help developers create more useful agents and better user experiences.
In general, vertically integrated solutions (like Facebook or Amazon) currently have a better user experience - one of the inherent complexities of building a great product is making sure the pieces work together from top to bottom. But this convenience comes at a high price, especially as the software costs of building aggregation, marketing, monetization, and distribution agents decrease and the surface area of agent applications expands. It will take effort to match the user experience of vertically integrated providers, but providing a trusted, neutral identity layer for agents will allow entrepreneurs to have their own passports - and encourage experimentation in both distribution and design.
3. Forward-compatible human proofs
As AI becomes more pervasive—powering bots and agents in all kinds of online interactions, including deepfakes and social media manipulation—it’s becoming increasingly difficult to tell whether the people you’re interacting with online are real humans. This erosion of trust isn’t a future worry; it’s already here. From armies of X status comments to bots on dating apps, reality is starting to blur. In this environment, human proof becomes essential infrastructure.
One way to prove you’re human is through a digital ID (including the centralized ones used by the TSA). A digital ID contains everything a person can use to verify their identity — username, PIN, password, and third-party proof (e.g., citizenship or credit worthiness), among other credentials. The value of decentralization is obvious here: when this data lives in a centralized system, the issuer can revoke access, impose fees, or facilitate surveillance. Decentralization flips this on its head: users, not platform gatekeepers, control their own identities, making it more secure and censorship-resistant.
Unlike traditional identity systems, decentralized proof-of-human mechanisms such as Worldcoin’s Proof of Human allow users to control and host their own identities and verify their human identity in a privacy-preserving and trust-neutral manner. Just like a driver’s license, which can be used anywhere regardless of when and where it was issued, decentralized proof-of-humanity (PoP) can serve as a reusable base layer for any platform, including those that don’t exist yet. In other words, blockchain-based PoP is forward-compatible because it provides:
Portability: The protocol is a public standard that any platform can integrate. Decentralized PoPs can be managed through a 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: Platforms can independently choose to identify PoP IDs without going through a gatekeeper API that may discriminate against different use cases.
The challenge in this space is adoption: while we haven’t seen many real-world human proof use cases at meaningful scale yet, we expect adoption to accelerate with a critical mass of users, a handful of early partners, and killer apps. Each application that leverages a given digital ID standard makes that ID type more valuable to users; this attracts more users to get IDs; which, in turn, makes IDs more attractive to applications as a way to authenticate human identities. (And because on-chain IDs are inherently interoperable, these network effects can grow rapidly .)
We’ve seen mainstream consumer apps and services in the gaming , dating , and social media sectors announce partnerships with World ID to help humans know they’re playing, chatting, and transacting with real humans — indeed, the specific humans they expect. We’ve also seen new identity protocols emerge this year, including the Solana Attestation Service (SAS). While SAS is not the issuer of human proofs, it allows users to privately link off-chain data (like KYC checks for compliance or accreditation status for investing) with their Solana wallet to build a user’s decentralized identity. All of this suggests that the inflection point for decentralized PoPs may not be far away.
Human proof is not just about banning bots, it’s about creating clear boundaries between AI agents and human networks. It enables users and applications to differentiate between human and machine interactions, creating space for better, safer, and more authentic digital experiences.
4. Decentralized Physical Infrastructure (DePIN) for AI
AI may be a digital service, but its progress is increasingly limited by bottlenecks in physical infrastructure. The Decentralized Physical Infrastructure Network, or DePIN — which provides a new model for building and operating real-world systems — can help democratize access to the computing infrastructure underlying AI innovation, making it cheaper, more resilient, and more censorship-resistant.
How is it done? The two biggest barriers to AI progress 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 join together to form a permissionless computing marketplace , creating a level playing field for building new AI products.
Other use cases include distributed training and fine-tuning of LLMs, and distributed networks for model inference. Decentralized training and inference have the potential to significantly reduce costs because they use 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 provide massively scalable computing infrastructure).
The concentration of AI models in the hands of a few companies is an ongoing concern ; decentralized networks can help create more cost-effective, censorship-resistant, and scalable AI.
5. Infrastructure and protection for interactions between AI agents, end service providers, and users
As AI tools become more adept at solving complex tasks and executing multi-layered chains of interactions, AI will increasingly interact with other AI without human controllers.
For example, an AI agent may need to request specific data related to a computation, or recruit a specialized AI agent to perform a specific task—such as assigning a statistics robot to develop and run model simulations, or engaging an image generation robot to create marketing materials. AI agents will also create great value in completing entire transaction flows or any other activities on behalf of users—such as finding and booking flights based on someone’s preferences, or discovering and ordering new books in their favorite genres.
There is currently no universal inter-agent marketplace in place - such cross-querying is primarily done through explicit API connections, or within AI agent ecosystems that maintain inter-agent calls as an internal function.
More broadly, most AI agents today operate in isolated ecosystems with relatively closed APIs and a general lack of architectural standardization. But blockchain technology can help establish open standards for protocols, which is important for adoption in the short term. In the long term, this also supports forward compatibility : as new types of AI agents evolve and are created, they can expect to be able to plug into the same underlying network. Given blockchains’ interoperable, open source, decentralized, and generally more easily upgradeable architectures, they can more easily adapt to novel AI innovations.
As the market develops, many companies are already building blockchain rails for inter-agent interactions: Halliday, for example, recently launched its protocol, which provides a standardized, cross-chain architecture for AI workflows and interactions — and provides protocol-level protections to ensure that AI does not exceed the user’s intent. Meanwhile, Catena, Skyfire, and Nevermind use blockchain to enable payments from one AI agent to another without human intervention. More such systems are in development, and Coinbase has even begun providing infrastructure support for these efforts.
6. Keep AI/Vibe encoding applications in sync
The recent revolution in generative AI has made it easier than ever to build software. Coding is orders of magnitude faster, and — perhaps most importantly — it can be done in natural language, allowing even inexperienced programmers to derive existing programs and build new ones from scratch.
However, while AI-assisted coding creates these new opportunities, it also introduces a lot of entropy into and across programs. “Vibe coding” abstracts away the complex web of dependencies underlying the software — but this can also make the program susceptible to functional and security flaws when source repositories and other inputs change. At the same time, as people use AI to create their own custom applications and workflows, it becomes more difficult for them to interact with other people’s systems. In fact, even two vibe-coded programs that actually perform the same task may have very different operations and output structures.
Historically, standardization has been provided by file formats and operating systems, and more recently by shared software and API integrations, to ensure consistency and compatibility. But in a world where software evolves, morphs, and branches in real time, the standardization layer needs to be widely accessible and continually upgradable—all while maintaining user trust. Moreover, AI alone cannot solve the problem of incentivizing people to build and maintain these connections.
Blockchain solves both problems at once: protocolized synchronization layers that are wrapped into people’s 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, an engineer can create a custom interface to view sales information in a weekend, but as the amount of custom software grows, developers will need help keeping these applications in sync and running.
This is similar to how open source software libraries are developed today, but with continuous updates rather than periodic releases — and with incentive packaging. Both of these are made easier to achieve with cryptocurrency. Just like other blockchain-based protocols, shared ownership of the sync layer incentivizes active investment in its improvements. Developers, users (and/or their AI agents), and other consumers can be rewarded for introducing, using, and evolving new features and integrations.
In turn, shared ownership gives all users a stake in the overall success of the protocol, which can act as a buffer against malicious behavior. Just as Microsoft would not break the .docx file standard because of the impact it has on its users and brand, co-owners of the sync layer will not introduce clumsy or malicious code into the protocol.
As with all the software standardization architectures we’ve seen before, there’s huge potential for network effects here. As the Cambrian explosion of AI-encoded software continues, 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. Support micropayments for revenue sharing
AI agents and tools like ChatGPT, Claude, and Copilot promise a convenient new way to navigate the digital world. But for better or worse, they are also undermining the economics of the open internet . We’re already seeing this happen — for example, educational platforms have seen significant drops in traffic as students increasingly use AI tools, and several US newspapers are suing OpenAI for copyright infringement. If we don’t realign incentives, we may see an increasingly closed internet with more paywalls and fewer content creators.
There will always be policy solutions, of course, but as those work their way through the courts, some technical solutions are emerging. Perhaps the most promising (and technically complex) solution is to build a revenue-sharing system into the fabric of the network. When an AI-driven action leads to a sale, the source of the content that facilitated that decision should receive a share. The affiliate marketing ecosystem already does attribution tracking and revenue sharing like this; a more sophisticated version could automatically track and reward all contributors in the chain of information. Blockchain could obviously play a role in tracking that chain of provenance.
But such a system will also require new infrastructure for other functions—in particular, micropayment systems that can handle tiny transactions across multiple sources, attribution protocols that fairly value 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 institution Catena Labs , and financial infrastructure protocol 0xSplits —show potential in this regard, enabling near-zero-cost transactions and more granular payment splitting.
Blockchain will implement complex proxy payment systems through the following mechanisms:
Nanopayments can be distributed to multiple data providers, allowing a single user interaction to trigger micropayments to all contributing sources through automated smart contracts.
Smart contracts allow enforceable retroactive payments to be triggered once a transaction is completed, compensating the source of the information that led to the purchase decision in a fully transparent and traceable manner.
Additionally, blockchain enables complex and programmable payment splits, ensuring that revenue is distributed fairly through code-enforced rules rather than centralized decisions, creating trustless financial relationships between autonomous agents.
As these emerging technologies mature, they can create a new economic model for media that captures the full value creation chain from creators to platforms to users.
8. Blockchain as a registry for intellectual property and traceability
Generative AI brings with it a pressing need for efficient and programmable mechanisms for registering and tracking intellectual property—both to determine provenance and to support business models around IP access, sharing, and remixing. Existing IP frameworks—which rely on expensive middlemen and ex post enforcement—are ill-suited to a world where AI consumes content instantly and generates new variations at the click of a button.
What is needed are open public registries that provide clear proof of ownership, that IP creators can interact with easily and efficiently — and that AI and other web applications can interact with directly. Blockchains are ideal for this because they register IP without relying on middlemen and provide immutable proof of provenance; they also enable third-party applications to directly identify, license, and interact with that IP.
Understandably, there’s a lot of skepticism about the whole idea that technology can somehow protect IP, as the first two eras of the web — and the ongoing AI revolution — are often associated with a decline in intellectual property protection. One problem is that many of today’s IP-based business models focus on excluding derivative works, rather than trying to incentivize and monetize them. But programmable IP infrastructure not only enables creators, franchises, and brands to clearly establish ownership of their IP in the digital space — it also opens the door to business models that are explicitly centered around sharing IP for use in generative AI and other digital applications. In effect, this turns one of the main threats of generative AI to creative works into an opportunity.
We’ve seen creators experiment with newer models early on in the NFT space, with companies leveraging NFT assets on Ethereum to support network effects and value accrual under CC0 brand building . More recently, we’ve seen infrastructure providers building protocols and 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 remixes through protocols such as Alias, Neura, and Titles. Incention ’s Emergence series of works, through a blockchain registry built on the Story Protocol to track who created what, thereby engaging fans to co-create a sci-fi universe and its characters.
9. Web crawlers that help compensate content creators
Today, the AI agents with the best product-market fit are neither used for coding nor for fun. They are web crawlers — autonomously navigating the web, collecting data, and deciding which links to follow.
It’s estimated that nearly half of all internet traffic currently originates from nonhumans. Bots routinely flout the conventions of robots.txt — a file that’s supposed to tell automated web crawlers whether something is welcome but is rarely authoritative in practice — and use the data they extract to bolster the defenses of some of the world’s largest tech companies. Worse, websites end up footing the bill for these unwanted guests, paying for bandwidth and serving CPU resources to what feels like a never-ending stream of faceless scrapers. In response, companies like Cloudflare and other CDNs (content delivery networks) offer blocking services. It’s a patchwork of services that shouldn’t exist.
We ’ve argued before that the original protocol of the internet — the economic contract between content creators and distribution platforms — is likely to break down. This is already starting to show in the data: over the past twelve months, website owners have begun blocking AI-facing bots en masse. In July 2024, about 9% of the top 10,000 websites banned AI bots, and now that number is 37% . This number will only increase as more website operators become sophisticated, and as users continue to get frustrated.
So, instead of paying CDNs to completely block anything that looks like a bot, what if we took a middle ground? Instead of hogging a system designed to bring human traffic to a website, AI bots could pay for the right to collect data. That’s where blockchain comes in: in this scenario, each web crawler agent would own some cryptocurrency and negotiate on-chain with each website’s “gatekeeper” agent via x402 or a paywall agreement. (The challenge, of course, is that the robots.txt system, also known as the Robots Exclusion Standard, has been ingrained into the business models of internet companies since the 1990s. Overcoming this would require large-scale group coordination, or the involvement of a CDN like Cloudflare).
But humans can prove themselves as humans in a separate channel through World ID (see above) and obtain content for free. In this way, content creators and website owners can be compensated for their contributions to large AI datasets at the point of data collection, while humans can continue to enjoy an Internet with free information.
10. Personalized and private advertising
AI is already starting to impact the way we shop online, but what if the ads we see every day were… helpful? There are a lot of obvious reasons why people don’t like ads. Irrelevant ads are pure noise. Meanwhile, not all personalization is created equal. AI-driven ads can feel intrusive if they’re too precise — drawn from reams of consumer data. Other apps have tried to monetize by limiting content, such as streaming services or game levels, which are blocked by no-skip ads.
Cryptocurrencies can help solve some of these problems, offering an opportunity to reimagine how advertising works. Combined with blockchain, personalized AI agents can bridge the gap between irrelevant and weird, serving ads based on user-defined preferences. But importantly, they can do this without exposing user data globally , while directly compensating users for sharing their data or interacting with ads.
Some of the technical requirements here include:
Low-fee digital payments: To compensate users for engaging with ads (views, clicks, conversions), companies need to send small, frequent payments. To operate at scale, we need a fast, high-throughput system with negligible fees.
Privacy-preserving data verification: AI agents need to be able to prove that consumers meet certain demographic attributes. Zero-knowledge proofs can verify demographic attributes while preserving privacy.
Incentive Model: If the internet embraced micropayment-based monetization (e.g., < $0.05 per interaction, as described above ), users would be able to opt-in to ads in exchange for small payments, inverting the current model from extraction to engagement.
People have been trying to make online ads relevant for decades — and offline ads for centuries, too. But rethinking ads through the lens of cryptocurrency and AI could finally make them more useful. Personalized without being creepy, and in a way that benefits everyone: For developers and advertisers, it unlocks new incentive structures that are more sustainable and aligned. For users, it provides more ways to discover and navigate the digital world.
All of this would make ad space more valuable, not less. It could also replace today’s entrenched, extractive ad economy with a more human-centric system: one that treats users as participants rather than products.
11. AI companions owned and controlled by humans
Many people spend more time on their devices than in face-to-face interactions, and this time is increasingly spent interacting with AI models and AI-curated content (especially). All of these models already provide a form of companionship, whether it’s entertainment, information, satisfying a niche interest, or educating children . It’s not hard to imagine a near future where AI companions for education, healthcare, legal advice, and friendship become popular modes of human interaction.
The AI companions of the future will be infinitely patient and customized to specific individuals and their specific use cases. They won’t just be helpers or robot servants, they could become highly valuable relationships. As a result, who will own and control these relationships — whether it’s the user or the companies and other middlemen — becomes equally important. If you’ve already worried about curation and censorship on social media over the past decade, this issue will become exponentially more complex and more personal in the future.
It’s not a new argument that censorship-resistant, hosted platforms like blockchains offer the most compelling path to censorship-resistant, user-controllable AI . Granted, individuals can run device models and buy their own GPUs, but most people either can’t afford them or simply don’t know how.
While we’re still a long way from widespread AI companions, all of these technologies are improving rapidly: Human-looking text-based companions are already excellent. Visual imagery has also improved dramatically. Blockchains are also getting more performant. To ensure censorship-resistant companions are easy to use, we need to rely on better user experiences for crypto applications. Thankfully, wallets like Phantom are making it simpler to interact with blockchains, and embedded wallets , keys , and account abstractions enable users to hold self-custodial wallets without the complexity of storing seed phrases themselves. Technologies like high-throughput trustless computers using optimistic and ZK coprocessors will also make it possible to build meaningful and lasting relationships with digital companions.
In the near future, the discussion will shift from when we’ll see lifelike digital companions and avatars to who and what will be able to control them.
Twitter: https://twitter.com/BitpushNewsCN
BitPush TG communication group: https://t.me/BitPushCommunity
Bitpush TG subscription: https://t.me/bitpush
