a16z: 11 major use cases to explore the intersection of AI and encryption

Source: a16z crypto; Translation: Golden Finance xiaozou

The economic model of the Internet is changing. As the open Internet collapses into a simple prompt box, we can't help but ask: Will AI lead us to an open Internet, or build a new paywall maze? And will it be highly centralized giant companies or the vast user community that controls all of this?

This is where cryptography comes in handy. We have discussed the intersection of AI and cryptography many times. In short, blockchain provides a new paradigm for building Internet services and networks - decentralized, trusted neutral, and truly owned by users. By reconstructing the economic model of the current system, blockchain can effectively check and balance the increasingly obvious centralization trend in AI systems, helping to create a more open and reliable Internet.

The idea that cryptography can optimize AI systems and vice versa is not new, but it is often vaguely defined. Some intersections are already attracting builders and users, such as verifying "proof of human" in the current era of low-cost AI. But other use cases seem to be years or even decades away. In this article, we will share 11 real-world use cases where crypto and AI converge to stimulate discussion about feasibility, challenges to be solved, and more. These cases are based on technologies being developed today, ranging from processing massive micropayments to ensuring human control over the relationship with future AI.

1、Persistent Data and Context in AI Interactions

Author: Scott Duke Kominers, a16z crypto research partner

Generative AI thrives on data, but for many applications, context (that is, the state and background information related to the interaction) is as important as data, or even more critical.

Ideally, an AI system (whether it’s an agent, a large language model interface, or something else) would remember a lot of details about the type of project you’re working on, your communication style, your preferred programming language, and so on. In reality, however, users often need to re-establish this context across interactions within a single application (e.g., every time they start a new ChatGPT or Claude session), not to mention when switching across systems.

Currently, context in generative AI applications is almost impossible to transfer between different systems.

Blockchain technology allows AI systems to transform key contextual elements into persistent digital assets. These assets can be loaded at the beginning of a session and transferred seamlessly across AI platforms. More importantly, blockchain may be the only solution that has the promise of both forward compatibility and interoperability—because these features are based on core properties of blockchain protocols.

Gaming and media are natural application scenarios: user preferences can persist across different games and environments. But the real value lies in knowledge applications (AI needs to understand the user’s knowledge structure and learning style) and specialized AI use cases such as programming. Of course, companies have long been developing custom bots with business-specific global contexts—but such contexts are often not portable, even between different AI systems used within the same organization.

Organizations are just beginning to realize this problem. The closest universal solution is custom bots with fixed persistent contexts. But context portability between users within a platform is already beginning to emerge off-chain: for example, the Poe platform allows users to rent out their own custom bots.

Bringing this activity on-chain will allow the AI ​​systems we interact with to share a context layer that contains the key elements of all digital activities. They can immediately understand our preferences and optimize the user experience more accurately. Conversely, just like the on-chain intellectual property registration mechanism, allowing AI to reference persistent on-chain context can also enable new market interactions around prompt words and information modules—for example, users can directly license or monetize their expertise while maintaining control over their data. Of course, shared context will also enable many possibilities that we have not yet imagined.

2Universal identity system for intelligent agents

Author: Sam Broner, partner of a16z crypto investment team

Identity - the authoritative credentials that record the essence of things - is the invisible infrastructure that supports today's digital discovery, aggregation and payment systems. Because the platform has enclosed this infrastructure within the wall, the identity we perceive is only part of the finished product: Amazon assigns identifiers (ASIN or FNSKU) to products, centrally displays products and assists users in discovery and payment. The same is true for Facebook: user identity forms the basis of its information flow and runs through all scenarios within the application, including Marketplace product listings, organic posts and paid ads.

With the development of AI intelligent agents, all this is about to change. When more companies use intelligent agents for scenarios such as customer service, logistics, and payment, their platforms will become less and less like single-interface applications, but will run across multiple carriers and platforms, accumulate deep context, and perform more tasks for users. But if agent identities are tied to a single marketplace, they become unusable in other important contexts (email threads, Slack channels, inside other products).

So agents need a unified “digital passport”. Without it, we can’t pay agents, verify their versions, query their capabilities, confirm who they’re acting for, or track their reputation across applications. Agent identities need to be wallets, API registries, changelogs, and social credentials—enabling any interface (email, Slack, or other agents) to recognize and interact with them in a unified way. Without a shared foundational component, “identity” means rebuilding infrastructure from scratch for every integration, discovery is always ad hoc, and context is lost when users switch channels.

We have an opportunity to design agent infrastructure from first principles. So how do we build a trusted, neutral identity layer that’s richer than a DNS record? Rather than reinventing an all-in-one platform that bundles identity with discovery, aggregation, and payment, we can enable agents to receive payments, list capabilities, and exist in multiple ecosystems without being tied to a specific platform. This is where the value of the intersection of cryptography and AI lies - blockchain networks provide permissionless composability, enabling developers to create more useful agents and better user experiences.

Currently, vertically integrated solutions such as Facebook or Amazon do provide a better user experience - the inherent complexity of building great products involves ensuring top-down coordination of all links. But this convenience comes at a high price, especially as the cost of building agent aggregation, marketing, monetization, and distribution software decreases, and the application scenarios of agents continue to expand. While it will take work to match the user experience of vertically integrated providers, a trusted and neutral agent identity layer will allow entrepreneurs to truly own their own digital passports and encourage them to innovate boldly in the areas of distribution and design.

3Forward-compatible human proof mechanism

Author: Jay Drain Jr., a16z crypto investment partner; Scott Duke Kominers, a16z crypto research partner

As AI increasingly penetrates into all types of online interactions (from deep fakes to social media manipulation, providing support for various robots and intelligent entities), it becomes increasingly difficult to distinguish whether the online interaction object is a real human. This trust crisis is not a future hidden danger, but a current reality - from the water army comment area of ​​X platform to the robots of dating software, the boundaries between real and virtual are blurring. In this environment, human proof becomes a critical infrastructure.

Digital ID cards (including centralized IDs used by the U.S. Transportation Security Administration) are a way to verify human identity. This type of ID contains all credentials that can prove personal identity, such as usernames, PIN codes, passwords, third-party authentication (such as citizenship or credit ratings), etc. The value of decentralization is clear here: when this data is stored in a centralized system, the issuer can revoke access, charge fees, or facilitate surveillance at any time. Decentralization completely reverses this power structure: users, not platform gatekeepers, control their identities, making them more secure and censorship-resistant.

Unlike traditional identity systems, decentralized proof of human mechanisms (such as Worldcoin's Proof of Human) allow users to self-custody and verify their human identity while protecting privacy and trusted neutrality. Just as a driver's license is available anywhere regardless of the time and place of issuance, a decentralized PoP (Proof of Personhood) can serve as a reusable base layer for any platform, including those that have not yet been born. In other words, blockchain-based PoP is forward-compatible because it has:

Portability: The protocol is a public standard that can be integrated by any platform. Decentralized PoP is managed through a public infrastructure and is completely controlled by users. This makes it completely portable and compatible with any platform now or in the future.

Permissionless accessibility:Platforms can choose to honor PoP IDs, without going through a gatekeeper API that may discriminate between different use cases.

The challenge in this space is adoption. While large-scale proof-of-humanity use cases have yet to emerge, we expect critical mass, early partners, and killer apps to accelerate adoption. Each application that adopts a particular digital ID standard increases the value of that ID to users, which in turn attracts more users to acquire that ID, which in turn drives more applications to integrate that ID as a means of human authentication (this network effect can form quickly due to the interoperability of on-chain IDs).

We have already seen major consumer applications in the gaming, dating, and social media sectors announce partnerships with World ID to help users confirm that they are playing, chatting, and transacting with real humans (and the specific people they expect). This year, new identity protocols such as Solana Attestation Service (SAS) have also emerged. Although it does not directly issue human proofs, SAS allows users to privately associate off-chain data (such as KYC checks or investment qualification certifications required for compliance) with Solana wallets to build decentralized identities. These signs indicate that the turning point of decentralized PoP may not be far away.

Human proof is not only about banning robots, but also about defining clear boundaries between AI agents and human networks. It enables users and applications to distinguish between human-machine interactions, creating space for better, safer, and more authentic digital experiences.

4、Decentralized Physical Infrastructure Network (DePIN) for AI

Author: Guy Wuollet, Partner of a16z crypto investment team

Although AI is a digital service, its development is increasingly constrained by physical infrastructure bottlenecks. Decentralized Physical Infrastructure Network (DePIN) - a new model for building and operating physical systems - can help popularize the computing infrastructure required for AI innovation, making it cheaper, more resilient, and more censorship-resistant.

How to achieve it? Energy and chip acquisition are two core obstacles to the development of AI. Decentralized energy can increase electricity supply, and builders also use DePIN to integrate idle chips in gaming PCs, data centers and other scenarios. These computers can jointly form a permissionless computing resource market, creating a fair competitive environment for the development of new AI products.

Other application scenarios include distributed training and fine-tuning of large language models, and distributed networks for model reasoning. Decentralized training and reasoning can significantly reduce costs because they utilize previously idle computing resources. At the same time, it provides anti-censorship capabilities to ensure that developers will not be deprived of platform access by hyperscale cloud service providers (centralized cloud service giants that provide elastically scalable computing infrastructure).

The problem of AI models being concentrated in a few companies has long existed; decentralized networks help create more cost-effective, more censorship-resistant, and more scalable AI systems.

5、Infrastructure and protection mechanisms for the interaction between AI agents, terminal service providers and users

Author: Scott Duke Kominers, a16z crypto research partner

As AI tools improve their ability to perform complex tasks and multi-level interaction chains, the demand for autonomous interaction between agents will increase significantly.

For example, an AI agent may need to obtain specific computing data, or call professional agents to perform special tasks - such as assigning statistical robots to develop and run model simulations, or enabling image generation robots in the production of marketing materials. AI agents can also create huge value by completing the entire transaction process on behalf of users - such as searching and booking air tickets based on preferences, or discovering and ordering new books of a certain type.

There is no universal inter-agent market yet, and such cross-system queries are mainly implemented through explicit API connections or are limited to closed ecosystems that support internal agent calls.

Broadly speaking, most current AI agents run in isolated ecosystems, with relatively closed APIs and lack of architectural standardization. Blockchain technology can help protocols establish open standards, which is crucial for short-term adoption. In the long term, this also supports forward compatibility: new AI agents can be seamlessly connected to existing underlying networks when they emerge. Thanks to the interoperability, open source, decentralization and easy-to-upgrade architectural characteristics, blockchain can better adapt to AI innovation iterations.

As the market develops, many companies have begun to build blockchain infrastructure for inter-agent interaction: for example, Halliday recently launched a standardized cross-chain architecture protocol that supports AI workflow interaction, ensuring that AI behavior does not deviate from user intent through protocol-level protection. Catena, Skyfire and Nevermind use blockchain to achieve autonomous payments between agents without human intervention. More similar systems are under development, and Coinbase has even begun to provide infrastructure support for these attempts.

6KeepAI/Vibe Coding Applications Synchronous

Author: Sam Broner, partner of a16z crypto investment team; Scott Duke Kominers, partner of a16z crypto research

The revolutionary progress of generative AI has made software development easier than ever before. Coding efficiency has increased by orders of magnitude, and more importantly-now that natural language programming can be used, even inexperienced developers can fork existing programs or build new applications from scratch.

However, while AI-assisted coding creates new opportunities, it also introduces a lot of entropy into and out of the program. Although "Vibe coding" abstracts away the complex dependency network at the bottom of the software, this programming method may cause hidden dangers in the functionality and security of the program as the source library and other inputs change. In addition, when people use AI to create personalized applications and workflows, the difficulty of connecting these systems with other people's systems increases. In fact, two atmospheric coding programs that perform the same task may have completely different operating logic and output structures.

For a long time, the standardization work to ensure consistency and compatibility was first undertaken by file formats and operating systems, and later by shared software and API integration. But in a world where software evolves, deforms and forks in real time, the standardization layer needs to be widely accessible and continuously upgradeable, while maintaining user trust. More importantly, AI alone cannot solve the problem of motivating people to establish and maintain these links.

Blockchain technology can solve both problems at the same time: embed user-customized software builds through protocolized synchronization layers, and dynamically update to ensure cross-platform compatibility in changes. In the past, large companies might spend millions of dollars to hire "system integrators" such as Deloitte to customize Salesforce instances, but now engineers can create customized interfaces for viewing sales information in a weekend. But as the number of customized software surges, developers need help to keep these applications running in sync.

This is similar to the current development model for open source software libraries, but with continuous updates rather than periodic releases—and an added incentive layer. Both of these are made easier with the help of cryptography. As with other blockchain-based protocols, shared ownership of the sync layer incentivizes all parties to continue to invest in improvements. Developers, users (and their AI agents), and other participants can all 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 acts as a buffer against malicious behavior. Just as Microsoft wouldn’t break the .docx file standard to avoid affecting its users and brand reputation, the shared owners of the sync layer have no incentive to introduce poor or malicious code into the protocol.

As with all software standardization architectures before it, there is huge potential for network effects in this space. As the Cambrian explosion of AI coding software continues, the network of heterogeneous systems that need to stay connected will expand dramatically. In short: Vibe coding can’t stay in sync with Vibe alone. Cryptography is the answer.

7Micropayment system supporting revenue sharing

Author: Liz Harkavy, partner of a16z crypto investment team

AI tools and agents represented by ChatGPT, Claude and Copilot provide a new and convenient way to navigate the digital world. But regardless of the pros and cons, they are shaking the economic foundation of the open Internet. The real impact has already appeared - educational platforms are facing a sharp drop in traffic due to students turning to AI tools, and several American newspapers are suing OpenAI for copyright infringement. If the incentive mechanism cannot be reconstructed, we will witness an increasingly closed Internet, with more paywalls and fewer content creators.

Policy means exist, but while the legal process is advancing, a number of technical solutions are emerging. The most promising (and complex) solution may be to embed a revenue sharing system into the network architecture: when AI-driven behavior leads to a transaction, the content source involved in the decision-making process should receive a share. Affiliate marketing ecosystems already implement similar attribution tracking and revenue distribution, and more advanced versions can automatically track and reward all contributors to the information chain - blockchain can obviously play a role in the traceability chain.

However, such systems require new infrastructure with special features: micropayment systems that can handle multi-source microtransactions, attribution protocols that fairly evaluate various contributions, and governance models that ensure transparency and fairness. Existing blockchain tools have shown potential, such as Rollup and Layer2 solutions, AI-native financial institutions Catena Labs, and financial infrastructure protocols 0xSplits, which can achieve near-zero-cost transactions and more sophisticated payment splits.

Blockchain will enable smart payment systems through the following mechanisms:

• Nanopayments can be split to multiple data providers, and a single user interaction can automatically distribute very small payments to all contributing sources through smart contracts.

• Smart contracts support executable traceable payments based on completed transactions, compensating information sources that are confirmed to have influenced purchasing decisions after the transaction occurs in a fully transparent and traceable manner

• Support complex programmable payment distribution schemes, achieve fair distribution of benefits through code-enforced rules rather than centralized decision-making, and establish trustless financial relationships between autonomous agents

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

8Blockchain as an intellectual property and traceability register

Author: Scott Duke Kominers, a16z crypto research partner

The rise of generative AI urgently requires efficient and programmable intellectual property registration and tracking mechanisms - both to ensure that the source of content can be traced and to support business models around IP access, sharing and remixing. The current IP framework relies on high-cost intermediaries and post-event accountability, which can no longer adapt to the new era of AI consuming content instantly and generating variants with one click.

We need an open and public registration system that can provide clear proof of ownership, allow IP creators to interact conveniently and efficiently, and allow AI and other network applications to connect directly. Blockchain is a perfect solution: it can complete IP registration without intermediaries, provide tamper-proof traceability proof, and allow third-party applications to easily identify, authorize and call these IPs.

When the first two eras of the Internet (and the ongoing AI revolution) were often associated with weakened intellectual property protection, the idea that technology can protect IP naturally raised many questions. The problem is that most current IP business models focus on excluding derivative works rather than incentivizing and monetizing these creations. But programmable IP infrastructure not only allows creators, franchisees, and brands to clarify IP ownership in the digital space, but also opens the door to IP sharing business models centered on digital applications such as generative AI - which actually turns the main threat of generative AI to creative work into an opportunity.

We have seen creators test new models in the NFT field early, and some companies have used NFT assets on Ethereum to achieve network effects and value accumulation under CC0 brand building. Recently, protocols and even dedicated blockchains (such as Story Protocol) built for standardized and composable IP registration and authorization have emerged. Some artists have begun to license their artistic styles and works for creative remixing through protocols such as Alias, Neura, and Titles. Incention's Emergence series allows fans to participate in the co-creation of science fiction universes and characters, and tracks the content created by each contributor through a blockchain registry built on Story.

9Web crawlers that help content creators realize their value

Author: Carra Wu, partner of a16z crypto investment team

The AI ​​agents with the best product-market fit at the moment are not programming or entertainment assistants, but web crawlers - digital agents that autonomously travel the Internet, collect data and determine link tracking paths.

It is estimated that nearly half of network traffic already comes from non-human subjects. Crawler programs often ignore the robots.txt protocol (this application is used to tell automated crawlers website access permissions, but it is actually weak in binding force), and the data they collect eventually becomes a competitive barrier for some technology giants. Worse, websites have to bear the bandwidth and CPU resource costs of these uninvited guests, as if they are serving an endless stream of anonymous data harvesters. The blocking solutions provided by CDN (content distribution network) service providers such as Cloudflare are actually remedies that should not exist.

We have pointed out that the original contract of the Internet - the economic covenant between content creators and distribution platforms - is facing collapse. The data confirms this trend: in the past 12 months, website owners have begun to block AI crawlers on a large scale. In July 2024, only about 9% of the world's top 10,000 websites blocked AI crawlers. Now that proportion has reached 37%. As website owners upgrade their defenses and user dissatisfaction accumulates, this number will continue to rise.

If you don't rely on CDNs to completely block visitors suspected of crawling, can you find a compromise? Instead of abusing a system designed for human traffic, AI crawlers may pay for the right to collect data. This is where blockchain comes in: in this scenario, each crawler agent will hold cryptocurrency and negotiate on-chain with the website's "bouncer" agent or paywall protocol through the x402 protocol (of course, the challenge is that the Robots Exclusion standard that has been used since the 1990s is deeply rooted and requires large-scale group collaboration involving CDN giants such as Cloudflare to break through).

At the same time, human users can continue to obtain content for free by verifying their real identity through World ID (see above). In this way, content creators and website owners can obtain reasonable compensation for AI training sets during the data collection stage, while humans can still enjoy the Internet of information freedom.

10、A new paradigm for advertising that is both accurate and private

Author: Matt Gleason, a16z crypto security engineer

AI has begun to change the way we shop online, but what if the ads we see every day are really useful? The reason why people hate ads is obvious: irrelevant ads are pure noise, and overly accurate AI ads (based on massive consumer data) are creepy. Other applications monetize through unskippable ad walls (such as streaming services or game levels).

Cryptography can reconstruct the advertising mechanism and solve these pain points. Personalized AI agents combined with blockchain can find a balance between "irrelevant ads" and "terrifying accuracy" - placing ads based on user-defined preferences. The key is that all this does not require global exposure of user data, and can directly compensate data sharers or ad interactors.

Required technical elements include:

• Low-fee digital payments: To compensate users for ad interactions (views/clicks/conversions), companies need to send small payments at a high frequency. This requires a high-throughput, near-zero-fee payment system.

• Privacy-preserving data verification: AI needs to be able to prove that consumers meet certain demographic characteristics. Zero-knowledge proofs can be verified while preserving privacy.

• Incentive mechanisms: If the Internet adopts a micropayment-based monetization model (e.g., <$0.05 per interaction), users can choose to watch ads in exchange for compensation, transforming the current "extraction model" to a "participation model."

Humans have been pursuing ad relevance for hundreds of years (offline) and decades (online). Reconstructing advertising from the perspective of encryption and AI will eventually make it truly useful: accurate but not shocking, achieving a win-win situation for all parties - unlocking a more sustainable and interest-aligned new incentive structure for builders and advertisers; and providing more ways for users to explore the digital world.

This will not only not devalue the value of advertising space, but will increase its value. It is also expected to subvert the current entrenched extractive advertising economy and replace it with a more humane system: users are regarded as participants, not products.

11、AI partners owned and controlled by humans

Author: Guy Wuollet, partner of a16z crypto investment team

Modern people spend more time on electronic devices than face-to-face communication, and more and more time is spent interacting with AI models and AI-screened content. These models already inherently provide some kind of companionship—whether it’s entertainment, information, hobbies, or children’s education. It’s not hard to imagine that in the near future, AI-based educational assistants, health advisors, legal assistants, and emotional companions will become the mainstream way humans interact.

The AI ​​companions of the future will be infinitely patient and deeply adaptable to the specific needs of individual users. They are not just assistants or robot servants, but are likely to develop into cherished "relationships." Therefore, it becomes critical to who owns and controls these relationships (users or companies and other intermediaries). If you have been concerned about social media content moderation and censorship in the past decade, this issue will become exponentially more complex and more personal in the future.

Anti-censorship hosting platforms (such as blockchains) provide the most powerful path to user-controllable, uncensorable AI—this is not a new argument (it has been discussed above). Although individuals can run local models or purchase their own GPUs, most people either cannot afford it or lack the technical ability.

Although it will take some time for AI companions to become ubiquitous, the technology is evolving rapidly: text-based anthropomorphic companions are already quite mature, avatars have significantly improved, and blockchain performance continues to improve. Ensuring the ease of use of censorship-resistant assistants requires a better user experience for crypto applications. Fortunately, wallets such as Phantom have greatly simplified blockchain interactions, with embedded wallets, pass keys, and account abstraction technologies allowing users to self-host wallets without having to remember mnemonics. With the help of high-throughput trustless computers (using technologies such as optimistic proofs and ZK coprocessors), it will be possible to establish meaningful and lasting relationships with digital companions.

In the near future, the focus of the discussion will shift from "when can we see realistic digital companions" to "who can control them and how to control them."