a16z: 11 Applications Scenarios of the Fusion of Crypto and AI

Compiled by | Saoirse, Foresight News

The economic logic of the internet has quietly changed. As the open network gradually shrinks into a “command input bar”, we must ponder: will artificial intelligence lead us to an open internet, or will we fall into a maze constructed by new types of paywalls? Will control be in the hands of large centralized companies or in the hands of a broad user base?

This is precisely where Crypto comes into play. We have discussed the intersection of AI and Crypto multiple times — in short, blockchain is a new paradigm for reconstructing the architecture of internet services, capable of building a decentralized, trust-neutral, and user-owned network system. By redefining the economic rules that underpin existing systems, blockchain provides an effective means to counterbalance the centralization trend in the field of AI, aiming to create a more open and resilient internet ecosystem.

The concept of mutual empowerment between Crypto and AI systems has long existed, but the ways in which they combine have lacked clear definition. Some crossover areas (such as “human identity” verification in the context of the proliferation of low-cost AI tools) have attracted the attention of developers and users, while other application scenarios may take years or even decades to materialize. Therefore, this article will share 11 cross-border application scenarios of AI and Crypto, aiming to promote relevant discussions: exploring the potential possibilities and challenges of the combination of AI and Crypto, and looking forward to more innovative directions. These scenarios are all based on the current level of technology and cover a diverse range of fields from massive micro-payment processing to ensuring that humans maintain dominance in future AI interactions.

1. Persistent data and contextual environment in AI interaction

Written by: Scott Duke Kominers

The development of generative AI highly depends on data, but in many application scenarios, the importance of context (i.e., the interactive related states and background information) is no less than that of data, and may even be more critical.

Ideally, whether it’s an agent, a large language model interface, or other applications, AI systems should be able to remember numerous details about users, such as their work type, communication style, preferred programming languages, and more. However, in reality, users often need to reset these contexts across different sessions in the same application (for example, starting a new ChatGPT or Claude session), not to mention switching between different systems. Currently, the context of a particular generative AI application is almost impossible to transfer to other applications.

With the help of blockchain technology, AI systems can transform key contextual elements into persistent digital assets, allowing them to be loaded when a conversation starts and seamlessly transferred between different AI platforms. Furthermore, based on its characteristics, blockchain may be the only solution that simultaneously meets the requirements of “forward compatibility” and “interoperability.”

This application is particularly suitable in AI-mediated gaming and media fields—user preferences (from difficulty settings to key bindings) can remain consistent across different games and scenarios. However, the real value is reflected in knowledge application scenarios (such as when AI needs to understand the user’s knowledge base and learning patterns) and specialized AI applications (such as programming assistance). Of course, some companies have developed custom robots tailored to specific business contexts, but in such scenarios, the context is often not transferable across systems, and even within different AI tools in the same enterprise, sharing is difficult.

Institutions have just begun to realize this issue, and the current universal solution is custom robots with fixed backgrounds. However, the context migration among users within the platform has begun to emerge off-chain: for example, on the Poe platform, users can rent out their custom robots for others to use.

After putting such scenarios on the chain, our interactive AI system will be able to share a contextual layer that includes all key elements of digital activities. They will immediately understand user preferences and optimize the user experience more accurately. Conversely, the on-chain intellectual property registration system allows AI to reference persistent on-chain context, which also creates possibilities for new market-oriented interactions around prompts and information modules — users can directly authorize or commercialize their own expertise while retaining data control. Of course, sharing context will also give rise to many possibilities that we have yet to foresee.

2. General Identity System for Agents

Written by: Sam Broner

Identity (i.e., “the authoritative record of the essential attributes of something”) is the underlying architecture that supports today’s digital discovery, aggregation, and payment systems. As platforms enclose this architecture within ecological walls, identity in the eyes of users has become part of the product functionality: Amazon assigns a unique identifier (ASIN or FNSKU) to products, centrally displays them, and assists users in discovery and payment; Facebook operates similarly, where user identity is the core foundation of its information flow and the entire in-app discovery function, including product listings, native posts, and paid advertisements.

With the evolution of AI agents, this landscape is about to change. As more businesses adopt agents (for customer service, logistics management, payment processing, and other scenarios), their platforms will no longer be limited to single interface applications, but will span multi-platform ecosystems, accumulate deep context, and perform more diversified tasks for users. However, if the agent identity is only tied to a single market, it will lose its usability in other key scenarios (such as email threads, Slack channels, and other products).

Therefore, agents need a single, portable “digital passport.” Without this passport, it will be impossible to determine how to pay the agent, verify its version information, query its functional attributes, know its service targets, or trace its credibility records across different applications and platforms. The agent’s identity needs to incorporate multiple functions such as a wallet, API registry, update log, and social proof, to ensure that any interface (email, Slack, or other agents) can interpret and interact according to a unified standard. Lacking this shared information of “identity,” every system integration must reconstruct the underlying architecture from scratch, and the discovery mechanism will always be in a temporary state, causing users to lose contextual information each time they switch channels or platforms.

We have the opportunity to design proxy infrastructure starting from the fundamental principles. So, how can we build a more robust and trustworthy neutral identity layer than DNS records? Proxies should not repeat the mistakes of “identity and discovery, aggregation, and payment functions being tied to a monolithic platform,” but should be able to accept payments and showcase functionalities across multiple ecosystems without the concern of being locked into a specific platform. This is precisely where the value of the crossover between Crypto and AI lies — the permissionless composability offered by blockchain networks can help developers create more practical proxies and provide a better user experience.

Overall, vertically integrated solutions (such as Facebook or Amazon) currently offer a superior user experience. One of the inherent challenges in creating excellent products is ensuring that all components work together seamlessly from top to bottom. However, the cost of this convenience is high, especially as the costs of building agent aggregation, marketing, commercialization, and distribution software continue to decline, and the coverage of agent applications continues to expand. While achieving the user experience level of a vertically integrated platform still requires effort, building a trustworthy and neutral identity layer for agents will enable entrepreneurs to take control of their “digital passports” and promote innovative explorations in distribution and design.

3. Forward-compatible “Human Identity” proof mechanism

Written by: Jay Drain Jr., Scott Duke Kominers

As AI technology permeates various online interaction scenarios (including deep fakes and social media manipulation), determining “whether one is interacting with a real human online” is becoming increasingly difficult. The collapse of this trust system has already occurred — from comment bots on X platform (formerly Twitter) to bot accounts in dating apps, the boundary between the virtual and the real is becoming increasingly blurred. In this context, proving “human identity” has become the core infrastructure of the digital ecosystem.

One way to prove “you are human” is by using a digital ID (including the centralized ID used by TSA). Digital IDs encompass all elements that can be used for authentication, such as usernames, PINs, passwords, and third-party verifications (like citizenship or credit ratings). Decentralization demonstrates significant value here: when data is stored in centralized systems, the issuer may revoke access, charge additional fees, or implement surveillance; whereas the decentralized model reverses this logic—users (rather than platform administrators) maintain control over their own identity, making it safer and more resistant to censorship.

Unlike traditional identity systems, decentralized “human identity” proof mechanisms (such as World’s Proof of Human) allow users to manage their identity information independently and verify “human attributes” in a privacy-preserving and trust-neutral manner. Just as a driver’s license can be used in any region (regardless of when and where it was issued), decentralized “human identity” proof can serve as a universal underlying protocol across platforms, even suitable for emerging platforms that have yet to be born. In other words, blockchain-based “human identity” proof possesses forward compatibility features due to its following advantages:

· Portability: The relevant protocols belong to open standards and can be integrated by any platform. The decentralized “human identity” proof can be managed through public infrastructure, controlled by users themselves, offering complete portability, and can be compatible with any platform now or in the future;

· Permissionless Accessibility: The platform can autonomously choose to recognize “human identity” IDs without needing to go through an API that may discriminate against different use cases.

The challenge faced in this field is the practical application: although there are currently no real-scale applications for “human identity” proof scenarios, we anticipate that as the number of users reaches a critical mass, early partnerships are formed, and killer applications emerge, the adoption rate will accelerate. Each application that adopts a specific digital ID standard will enhance the value of that ID to users, thereby attracting more users to acquire the ID, creating a positive feedback loop (and since on-chain IDs are designed to be interoperable, network effects can accumulate rapidly).

We have seen mainstream consumer applications such as gaming, dating, and social media announce partnerships with World ID to help users confirm that they are interacting with real humans (rather than bots) during gaming, chatting, and trading; this year has also seen the emergence of new identity protocols like the Solana Attestation Service (SAS) — although SAS is not an issuer of “human identity” proof, it allows users to privately associate off-chain data (such as compliant KYC or investment qualifications) with their Solana wallets, helping to build a decentralized identity system. All these signs indicate that the turning point for decentralized “human identity” proof may not be far off.

The significance of the “human identity” proof lies not only in banning robots but also in delineating a clear boundary between AI agents and human networks. It enables users and applications to distinguish between human and machine interactions, thus creating a higher quality, safer, and more authentic digital experience.

4. Decentralized Infrastructure in the AI Field (DePIN)

Written by: Guy Wuollet

Although AI belongs to digital services, its development is increasingly constrained by physical infrastructure. Decentralized Physical Infrastructure Networks (DePIN) provide a new model for building and operating real-world systems, contributing to the democratization of the computational infrastructure behind AI innovation, making it more cost-effective, resilient, and censorship-resistant.

How to achieve this goal? The two core challenges facing AI development are computing power supply and chip acquisition. Decentralized computing networks can provide more computing power, while developers are also leveraging DePIN to aggregate idle chip resources from sources such as gaming PCs and data centers. These computing devices can form a permissionless computing market, creating a fair competitive environment for developing new AI products.

Other application scenarios include distributed training and fine-tuning of large language models, as well as distributed networks for model inference. Decentralized training and inference (due to the utilization of idle computing resources) can significantly reduce costs while providing censorship resistance, ensuring that developers are not cut off by oversized cloud service providers (such as centralized cloud service giants).

The issue of a few companies monopolizing AI models has long existed, and decentralized networks help to build a more economical, censorship-resistant, and scalable AI ecosystem.

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

Written by: Scott Duke Kominers

As the capabilities of AI tools in solving complex tasks and executing multi-level interaction chains continue to improve, AI systems will increasingly need to interact with other AI systems without human intervention.

For example, a certain AI agent may need to request specific data related to computational tasks, or recruit specialized AI agents to complete specific tasks (such as assigning statistical robots to develop and run model simulations, or calling image generation robots when creating marketing materials). AI agents will also create significant value by completing the entire transaction process or other activities on behalf of users, such as finding and booking flights according to user preferences, or discovering and ordering new books of their favorite types.

Currently, there is no mature universal agent-to-agent interaction market, and such cross-system queries can mostly only be achieved through explicit API connections or by implementing agent calls as internal functionalities within AI ecosystems.

Overall, most AI agents today operate in isolated ecosystems, where API interfaces are relatively closed and lack architectural standardization. However, blockchain technology can help protocols establish open standards, which is crucial for short-term application implementation; in the long term, this also supports forward compatibility — as new AI agents evolve and emerge, they can connect to the same underlying network. Due to its interoperability, open-source nature, decentralization, and typically easier upgradeability, blockchain can adapt more flexibly to the innovative demands of the AI field.

With the development of the market, several companies have begun to build blockchain infrastructure for agent-to-agent interactions: for example, Halliday recently launched a relevant protocol to provide a standardized cross-chain architecture for AI workflows and interactions, and offers a protection mechanism at the protocol layer to ensure that AI behavior does not exceed user intent; Catena, Skyfire, and Nevermind utilize blockchain technology to support automatic payments between AI agents without human intervention. More such systems are being developed, and Coinbase has even started to provide infrastructural support for these explorations.

6. Ensure the synchronization of AI/custom programming applications

Written by: Sam Broner, Scott Duke Kominers

The innovation of generative AI has brought a qualitative leap in software development efficiency: coding speed has increased by several orders of magnitude, and most importantly, it can be done through natural language — even inexperienced programmers can replicate existing programs or build new applications from scratch.

However, AI-assisted coding introduces a significant amount of uncertainty both inside and outside of programs while creating new opportunities. “Custom Programming” (Vibe coding) abstracts the complex dependency network at the software’s core, but this also makes programs susceptible to functional and security vulnerabilities when changes occur in the source library and other inputs. Additionally, when people use AI to create personalized applications and workflows, interactions with others’ systems become more challenging — in fact, even if two “Custom Programming” programs have the same functionality, their operational logic and output structure may differ significantly.

Historically, the standardization work to ensure software consistency and compatibility was initially undertaken by file formats and operating systems, while in recent years it has relied on shared software and API integration. However, in the new era of real-time evolution, iteration, and branching of software, the standardization layer needs to have wide accessibility and continuous upgradability, while maintaining user trust. Moreover, relying solely on AI cannot solve the problem of “incentivizing people to build and maintain these connections.”

Blockchain technology simultaneously addresses these two issues: the protocolized synchronization layer can be embedded in the user’s customized software architecture and dynamically updated to ensure cross-system compatibility as the environment changes. Historically, large enterprises might have paid millions of dollars to “system integrators” like Deloitte for customized Salesforce instances. Today, engineers can create custom interfaces to view sales information over the weekend, but as the number of customized software grows, developers need professional support to keep these applications running in sync. (Note: Salesforce is a customer relationship management (CRM) software service provider founded in the United States in March 1999.)

This is similar to the development model of today’s open-source software libraries, but with continuous updates (rather than periodic releases) and an incentive mechanism — both made easier by Crypto technology. Like other blockchain-based protocols, the shared ownership mechanism of the synchronization layer incentivizes all parties to actively invest in improvements: developers, users (and their AI agents), and other consumers can be rewarded for introducing, using, and optimizing new features and integrations.

Conversely, shared ownership tightly binds all users to the overall success of the protocol, creating a buffer mechanism against malicious behavior — just as Microsoft would not easily undermine the .docx file standard (as it would have a chain reaction impact on users and the brand), the co-owners of the synchronization layer would also be disinclined to introduce inefficient or malicious code into the protocol.

Like all standardized software architectures we have seen, there is a huge potential for network effects here. As the “Cambrian explosion” of AI coding software continues to evolve, the heterogeneous system networks that need to maintain communication will expand exponentially. In short, “custom programming” requires not only “coding style,” but also Crypto technology to maintain system synchronization.

7. A micropayment system that supports revenue sharing

Written by: Liz Harkavy

AI agents and tools like ChatGPT, Claude, and Copilot provide a new way to navigate the digital world conveniently, but whether good or bad, they are shaking the economic foundations of the open internet. We have seen concrete manifestations of this trend — for example, educational platforms have seen a sharp decline in traffic due to students’ extensive use of AI tools, and several American newspapers are suing OpenAI for alleged copyright infringement. If we do not readjust the incentive mechanisms, we may face an increasingly closed internet: more paywalls and fewer content creators.

Of course, policy solutions always exist, but as they progress through the judicial process, a series of technical solutions are also emerging. The most promising (and technically challenging) solution may be to embed a revenue-sharing mechanism within the network architecture: when AI-driven actions facilitate transactions, the content sources that provide informational support for that decision should receive a corresponding share. The affiliate marketing ecosystem is already engaged in similar attribution tracking and revenue sharing, while more advanced versions can automatically track and reward all contributors in the information chain — blockchain technology can clearly play a key role in tracking this traceability chain.

However, such systems also require new infrastructure with additional functions—especially micro-payment systems capable of handling micro-transactions across multiple sources, attribution protocols that fairly assess the value of different contributions, and governance models that ensure transparency and fairness. Many existing blockchain tools (such as Rollups and Layer2, AI-native financial institution Catena Labs, financial infrastructure protocol 0xSplits) have demonstrated application potential, supporting nearly zero-cost transactions and more granular payment splits.

Blockchain can implement complex proxy payment systems through various mechanisms:

· Nano payments can be split among multiple data providers, allowing a single user to interact and trigger micropayments to all contributing sources through automated smart contracts.

· Smart contracts support the triggering of executable traceable payments after a transaction is completed, compensating the sources that provide information for purchasing decisions in a fully transparent and traceable manner.

· In addition, blockchain supports complex and programmable payment splitting and allocation, ensuring that income is fairly distributed through code-enforced rules rather than relying on centralized decision-making, creating trustless financial relationships among autonomous agents.

As these emerging technologies mature, they will create new economic models for the media industry, capturing the entire value creation chain from creators to platforms to users.

8. Blockchain as an Intellectual Property and Traceability Registration System

Written by: Scott Duke Kominers

The development of generative AI has given rise to an urgent demand for efficient and programmable mechanisms for registering and tracking intellectual property—both to clarify rights ownership and to support business models surrounding access, sharing, and re-creation of intellectual property. The existing intellectual property protection framework relies on expensive intermediaries and post-facto enforcement measures, which cannot adapt to the demands of an era where AI consumes content instantly and generates new variants at the click of a button.

What we need is an open public registration system that provides clear proof of ownership, allowing intellectual property creators to efficiently participate in interactions, while AI and other web applications can connect directly. Blockchain technology is the ideal choice: it allows for intellectual property registration without intermediaries, provides tamper-proof provenance proof, and enables third-party applications to easily identify, authorize, and use the intellectual property.

Some people are skeptical about the view that “technology can protect intellectual property” — after all, the first two eras of Internet development (and the ongoing AI revolution) have often been associated with the weakening of intellectual property protection. Part of the reason is that many contemporary business models based on intellectual property focus on “prohibiting derivative works” rather than incentivizing and commercializing derivative creations. However, programmable intellectual property infrastructure not only allows creators, brands, and IP owners to clearly establish ownership in the digital space but also opens the door to “business models around IP sharing (for generative AI and other digital applications)” — which effectively turns the main threat of generative AI to creation into an opportunity.

We have seen creators in the NFT space experimenting with new models early on: companies use NFT assets on Ethereum to support network effects and value accumulation under CC0 brand building; recently, infrastructure providers are also building protocols for standardized and composable IP registration and licensing (such as Story Protocol) and even dedicated blockchains. Some artists have begun to use these tools to license their styles and works for creative re-creation through protocols like Alias, Neura, Titles, etc. Incention’s Emergence series invites fans to co-create a sci-fi universe and its characters, relying on the blockchain registry built on Story Protocol to track the ownership of each element.

9. Compensation content creator’s web crawler mechanism

Written by: Carra Wu

Nowadays, the AI agents that best meet market demand are not programming or entertainment tools, but web crawlers — they autonomously browse the web, collect data, and decide where to scrape from.

It is estimated that nearly half of current internet traffic comes from non-human entities. Crawlers often ignore the robots.txt protocol (which is supposed to inform automated crawlers whether they are allowed to access the website, but in practice has weak enforceability) and use the data they scrape to strengthen the market barriers of tech giants. Worse still, websites have to pay the price for these unwelcome guests, bearing the costs of providing bandwidth and CPU resources for vast amounts of unidentifiable crawlers. In response, CDNs (Content Delivery Networks) like Cloudflare offer blocking services, but this is merely a patchwork solution that should not exist.

We have pointed out that the native protocols of the internet (the economic agreements between content creators and distribution platforms) may collapse, and data is confirming this trend. In the past 12 months, website owners have massively blocked AI crawlers: in July 2024, only 9% of the top 10,000 websites globally prohibited AI crawlers, but this percentage has now reached 37%, and as more website operators upgrade their technology and user dissatisfaction intensifies, this number is expected to rise further.

If we do not rely on CDN to completely block suspected crawler access, can we find a compromise solution? AI crawlers should not utilize a system designed for human traffic for free, but should pay for data scraping rights. This is exactly where blockchain comes into play: each web crawler agent can hold Crypto and negotiate on-chain with the “access proxy” or paywall protocol of each website through the x402 protocol (of course, the challenge lies in the fact that the robots.txt protocol has been deeply embedded in the internet business logic since the 1990s, requiring large-scale coordination or participation from CDNs like Cloudflare to break through).

At the same time, humans can access content for free by proving their identity through World ID (see Chapter 3). In this model, content creators and website owners can be compensated during the collection of AI datasets, while human users can still enjoy an internet of “information freedom.”

10. Personalized Advertising with Privacy Protection

Written by: Matt Gleason

AI has begun to influence the online shopping experience, but what if the ads we encounter daily could be “truly useful”? The reasons people dislike ads are obvious: ineffective ads are mere noise, while overly precise AI ads based on massive consumer data feel invasive to privacy. Other applications capitalize on this by imposing limits on content through “non-skippable ads” (such as streaming services or game levels).

Crypto has provided the potential to reconstruct advertising models. Personalized AI agents combined with blockchain can find a balance between “irrelevant ads” and “overly precise ads,” delivering advertisements based on user-defined preferences. More importantly, this model does not require exposing users’ global data and can directly compensate users who actively share data or interact with ads.

To achieve this goal, the following technical requirements must be met:

· Low-fee digital payments: To compensate users for their advertising interactions (views, clicks, conversions), businesses need to frequently send small payments, which requires the system to have high processing capabilities and nearly zero transaction fees.

· Privacy-preserving data verification: AI agents need to prove that users meet specific demographic attributes, and zero-knowledge proofs can perform attribute verification while protecting privacy.

· Incentive Model: If the internet adopts a profit model based on micropayments (e.g., single interactions below $0.05, see Chapter 7), users can voluntarily choose to “watch ads for small rewards,” transforming the existing “exploitation” model into a “participatory” model.

For decades, online (and even offline for hundreds of years) advertising has always pursued “relevance.” Reconstructing advertising from the perspective of Crypto and AI will ultimately make it more practical—customized yet non-intrusive, benefiting all parties: for developers and advertisers, unlocking more sustainable and incentive-compatible business models; for users, gaining more paths to explore the digital world.

This will not only enhance the value of advertising space but may also disrupt the deeply entrenched “exploitative” advertising economy of today, creating a more human-centered system — users are no longer commodities being traded, but active participants.

11. AI companions owned and controlled by humans

Written by: Guy Wuollet

Nowadays, people spend more time on devices than in offline interactions, and increasingly engage with AI models and AI-generated content. These models have begun to offer companionship value, whether for entertainment, information gathering, catering to 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 emotional support will become the mainstream way of interaction.

Future AI companions will possess infinite patience and be tailored to individual needs — they will no longer be mere tools or robotic servants, but may become highly valued relationships. Therefore, the question of “who owns and controls these relationships” is crucial (is it the users, or intermediaries such as companies). If you have been concerned about the content filtering and censorship of social media over the past decade, this issue will become even more complex and personal in the future.

The viewpoint that “censorship-resistant blockchain custody platforms are the most feasible path for users to control AI” has been discussed multiple times (as mentioned earlier). Theoretically, individuals can run device-side models or purchase GPUs on their own, but most people either cannot afford it or lack the technical skills.

Although the popularization of AI companions still requires time, the relevant technology is rapidly iterating: anthropomorphic text interaction companions have become quite mature, avatar technology has made significant progress, and blockchain performance is continuously improving. To ensure that censorship-resistant companions are easy to use, we need to rely on better user experiences to realize crypto-driven applications. Fortunately, wallets like Phantom have greatly simplified blockchain interactions, and embedded wallets, password keys, and account abstraction technology allow users to hold self-custodial wallets without having to memorize seed phrases. High-throughput trusted computing technologies based on Optimistic and ZK co-processors will also help build deep and lasting relationships with digital companions.

In the near future, the focus of discussion will shift from “when will realistic digital companions appear” to “who can control them.”