When the AI ​​gold rush meets Crypto, AI agents make it easier for encryption products to "fly into the homes of ordinary people"

Original title: "AI, Crypto, and America"

Author: Derek Edws , Managing Partner of Collab+Currency

Compiled by: Zen, PANews

I am no longer convinced that our traditional strategic options can cope with the new realities of the future—especially given the alarming acceleration of artificial intelligence technology and its disruptive potential for the American workforce.

The United States must begin moving toward a new framework to accommodate new labor realities. This starts with understanding two of the most disruptive technological innovations of the past two decades: AI and Crypto.

AI

Thirty years ago, Internet search engines first emerged. Their value is deceptively simple: retrieving the world's data. Twenty-five years later, OpenAI’s ChatGPT is the fastest-growing consumer application in history. The product has 300 million weekly active users and more than 10 million monthly paid subscribers. In less than two years, annual product revenue was approximately $3 billion.

What is the value of these artificial intelligence systems? Combine "retrieve" with "complete". Simply put:

1. Request any job using natural language;

2. Let computer intelligence handle the work.

In the past three months, we can already clearly see that these products have rapidly improved in their ability to complete work requests, and the areas covered are also expanding - text, mathematics, audio, video, geometry, programming, etc. .

So far, the scaling assumptions of AI systems still hold true: more data, more computing power, and better models. Additionally, new dimensions of scaling are being explored, with innovations like test-time compute emerging. Last month, OpenAI's latest inference model achieved a score of 25% on a cutting-edge mathematics benchmark, problems that only highly specialized mathematicians can solve.

These next-generation inference models are adapting to new tasks and producing observable results. At the same time, new advances in robotics are enabling machines to increasingly perform complex physical tasks. Intelligent and human-like hardware will never tire, never take a vacation, or never go on strike.

This release of technological productivity is not unprecedented - the industrial revolution in the 18th century and the digital revolution in the 20th century greatly reduced costs, improved efficiency, and changed various forms of labor markets and economic structures.

But artificial intelligence appears to be a different kind of technological revolution. These systems can create value in unprecedented ways, beyond traditional cost structures. Anthropic CEO Dario Amodei believes that artificial intelligence may bring us nearly 50 to 100 years of biological progress in the next 5 to 10 years, simply by transferring most of our human labor structure to artificial intelligence systems.

As the infrastructure that supports the growth of these systems continues to evolve, I believe global humanity will see two trend lines emerge:

• The cost of all mental labor will converge toward the cost of running an AI system;

• The cost of all manual labor will tend to be the cost of the robot's mechanical parts.

The promise of this technology is clear:

• Unlimited results;

• Prices are significantly reduced;

• Across every value category.

For the purposes of this article, we assume that the above trends hold true. As someone who believes deeply that the next generation of Americans should have the same opportunities as those who came before them, I have two lingering questions: First, how will the United States win the AI ​​revolution with superiority? Second, how can individual Americans participate in the benefits of future AI systems and realize their potential to disrupt human labor?

cryptocurrency

In 2008, Satoshi Nakamoto released the Bitcoin white paper. The paper proposes a new, gamified accounting system in which globally distributed computers work together to reach a shared digital consensus on the Bitcoin ledger.

Today, Bitcoin has become the most powerful supercomputer in the world. The network is far larger than the networks of the likes of Amazon, Google and Microsoft combined. However, after fifteen years of development, blockchain still seems clunky:

• Poorly written code may have vulnerabilities that could lead to theft or account balances being reset to zero;

• User errors in managing private keys were the biggest cause of cryptocurrency theft last year;

• Compared with traditional web applications, decentralized applications are more difficult to use and have extremely high user churn rates.

Despite these limitations, interaction with cryptocurrencies is at an all-time high. One study estimates that 40% of the U.S. population now owns cryptocurrencies, up from 30% in 2023. About 24,000 developers around the world are actively involved in the development of blockchain and blockchain applications every month, which is much higher than the number of only 1,000 developers per month ten years ago.

Despite all these limitations, why are cryptocurrencies still growing? I believe this is because of five unique characteristics of cryptocurrencies that work together and cannot be replicated by any other database architecture:

1. Digital ownership. The blockchain database is global, fully auditable, community-owned, tamper-proof, and operates 24/7. Through the blockchain, individuals can own any digital object on the Internet, establishing a global digital property rights system for the first time.

2. Coordinate incentives. By automating the execution of contracts, blockchain-based protocols can leverage programmable incentives to coordinate new forms of digital work on the network. These incentive mechanisms can be used for those who use the product or service, provide economic security, contribute core code, provide market supply or demand, recommend others to use the product, etc.

3. Frictionless micropayments. Today, most Internet companies must adopt subscription, bundling, and advertising-supported business models because transaction costs are too high. This limitation restricts Internet business model innovation and hinders the emergence of consumer-friendly alternatives. Blockchain databases enable digital payments to be made globally in a frictionless, low-cost manner - bypassing the inefficiencies of traditional payment systems and eliminating the risk of chargebacks. Additionally, any cryptoasset can be divided into arbitrarily small units.

4. Shared standards. By leveraging shared settlement standards on the blockchain, various protocol tokens, stablecoins, applications, games, and financial services can fit together seamlessly, just like Lego bricks can be connected at will.

5. Distributed security. Blockchain networks are typically distributed across multiple nodes around the world, eliminating single points of failure. This decentralized architecture makes it more difficult for malicious actors to attack these network systems because they need to control a majority of nodes simultaneously.

Today, the cryptocurrency economy has a market capitalization of approximately $3.6 trillion and covers multiple emerging areas.

Over the next decade, I believe that the cryptocurrency economy will be repriced and rise significantly—driven primarily by two major trends at the intersection of artificial intelligence and cryptocurrency: (1) AI and crypto infrastructure; (2) AI and crypto infrastructure Encryption applications.

AI and Crypto Infrastructure

To understand the current landscape of AI infrastructure, it’s helpful to look back at historical events and look for similarities.

In 1849, the California Gold Rush quickly attracted massive investment. Hundreds of new roads were built for rapid transportation. San Francisco's port became one of the busiest in the world, shipping gold miners, goods, and tools around the world. A strong banking and financial system emerged to support the needs of emerging global businesses. Infrastructure investments at that time laid the foundation for the region's future as an economic powerhouse.

Today, 175 years later, the world is witnessing a similar gold rush, this time with the goal of creating artificial general intelligence (AGI). This time, however, the infrastructure to support AI is not limited to a specific region; these networks of data, computing power and electricity are being built by competitors around the world.

Unsurprisingly, the capital and computing power required to train, optimize and deploy AI infrastructure is extremely expensive and only a few companies can afford it. A conservative estimate is that the cost of training GPT-3 exceeds US$4 million, while the training cost of GPT-4 exceeds US$60 million.

More capital, more computing power, better performance.

While I am extremely proud and supportive of America’s contributions to AI in its traditional corporate form, I also believe it is necessary to acknowledge its structural limitations:

• Capture value. While centralized companies are able to use venture capital to drive important AI innovations, the economic benefits of these products are limited to a small number of shareholders, limiting the broader impact on society.

• Proprietary knowledge. Advances in technology frameworks tend to be concentrated within centralized companies, limiting others’ access to key breakthroughs—and information now flows around the clock at Internet speeds.

• Opaque system. Centralizing AI in opaque, closed, centralized systems makes it difficult for independent verifiers to audit a company’s practices around data collection, security, and accountability.

• Closed competition. The huge computing resources required to develop advanced artificial intelligence pose a huge obstacle to new products entering the market, and only a few well-funded companies can continuously break through technical bottlenecks.

By combining America’s AI infrastructure with the five unique characteristics of cryptocurrencies—digital ownership, coordinated incentives, frictionless micropayments, shared standards, and distributed security—I believe we can mitigate the negative impacts of centralized AI. , and restore the long-standing spirit of competition in the U.S. capital market. Additionally, by combining U.S. AI infrastructure with cryptocurrencies, I also believe it will lead to better performing AI systems: (a) greater transparency, (b) and greater transparency among millions of U.S. participants in the future. Fair title.

(a) Better performance

To understand the breakthroughs that AI can achieve without massive financial backing, take a look at the DeepSeek team. Two weeks ago, the China-based research group released DeepSeek-V3, a 670B parameter model that performs on par with many closed-source SOTA models, including GPT-4o and Claude-Sonnet-3.5. DeepSeek has not received any venture capital so far.

As open source projects like Bitcoin and Ethereum have proven, by distributing programmable incentives to a global pool of contributors, a qualified workforce and computing network can be greatly enhanced, creating a force far more powerful than a single lab or centralized system. From this perspective, creating a system that rewards AI labor and computing networks is not that different from creating a system that rewards Bitcoin labor and computing networks.

A few examples:

• For improved training data, cryptographic networks can reward humans for contributing high-quality annotated datasets—including private data, proprietary intelligence, or information not captured through traditional web scraping.

• For more powerful computing networks, cryptographic networks can incentivize individuals and organizations to contribute computing power through a decentralized marketplace — quickly building a global network of machines without upfront capital investment.

• For more efficient model training, open source developers can provide customized contributions, improvements, and optimizations to existing models in exchange for corresponding rewards. Additionally, when code and model weights are made public, hundreds of researchers and developers can simultaneously publish improvements, debug issues, fine-tune custom models and agents, and create new applications based on them—all over the network The incentive mechanism is coordinated.

Over time, I believe this type of broad collaboration pioneered by decentralized projects like Nous Research, Prime Intellect, and Bittensor will exceed what can be achieved within well-resourced private companies.

(b) Greater transparency

Open source AI models allow the research community to take a comprehensive look at their training process, architecture, and behavior and make improvements. This transparency helps identify potential risks or biases early on, leading to more reliable systems that people can trust. By leveraging blockchain in this process, the entire process of creating, rewarding, and improving AI protocols can remain transparent and auditable.

(c) Fairer ownership

Designed for every vertical in the AI ​​technology stack, the crypto network will establish a fairer ownership structure than existing centralized models. Through programmatic incentives, all crypto protocol contributors and participants are transparently compensated.

In addition, the complete market formed around various types of work in the artificial intelligence infrastructure stack will promote more refined competition in various categories of artificial intelligence design fields. Categories and subcategories such as data, compute, training, and deployment can all compete and accrue value in independent environments.

However, it’s not just AI infrastructure that ultimately benefits. I believe that AI agents will completely reorganize the current trajectory of global cryptocurrency adoption – across all application verticals.

AI and Cryptocurrency Applications

The complexity of cryptocurrency applications has long been seen as a significant barrier to widespread adoption. Over the past fifteen years, blockchain has required users to participate in complex approval procedures, manage private keys, and understand complex UI patterns that are beyond the grasp of most Internet users.

However, with the advent of proxy technology, these user patterns are changing rapidly. If you think of AI models as reactive infrastructure, responding based on previous training data, then you can think of AI agents as active applications, integrating models into new architectures to accomplish narrow goals.

Simply put, AI agents use underlying models to automatically think, automatically plan, and automatically execute actions. What needs to be understood is that agents are different from the “robots” we used to know. Unlike robots, AI agents can reason on demand. They can analyze their performance, adjust strategies, and solve complex tasks that sometimes require hundreds, and in the future, thousands of unique steps.

In September 2024, I met with one of my portfolio founders who was building an AI agent protocol for blockchain navigation. This protocol is called Wayfinder. From his phone, and using a few simple natural language prompts, I deployed a frontend and token contract that replicated Bitcoin's monetary policy onto the BASE blockchain and used ETH cross-chained from the Ethereum mainnet. The entire process took less than four prompts and took a total of five minutes to complete.

Startup apps like Wayfinder illustrate an important trend: AI agents will mediate cryptocurrencies’ longstanding technical frictions. Over the next 12 months, agents will transform the complex structure of blockchain into seamless natural language interactions, increasing the accessibility of the protocol, protecting users from their own errors, and helping developers deploy more secure code , and greatly reduce the churn rate of consumers in complex decentralized products and services.

What’s more, a key management network will extend all of these agent capabilities, allowing agents to seamlessly perform tasks across blockchains without human involvement. A global namespace network will enable each agent’s actions to span all blockchains and be connected to a human identity.

Simply put:

Crypto Brokers make it easier to build or use any crypto product, no matter which blockchain it is on.

• In the field of decentralized finance, agents will compress the financial friction of cryptocurrencies into the risk needs of each user by simplifying natural language instructions.

• In the crypto gaming space, agent workflows will power personalized and ownable creations in the crypto economy, more complex non-player characters, and on-chain experiences tailored to players.

• In a decentralized organization, humans will agree on overall policy goals and constraints and allow agents to perform in various dimensions of business, protocol, or administrative functions.

These guiding roles and advantages have brought revolutionary breakthroughs from zero to one for all encryption applications. Millions of new users will join this way, and no sector will be left out.

AI, cryptocurrencies and ownership

When understanding the grand trends before us, it is important to look back and remember the lessons from our history. For much of human history, the ability to secure and defend resources meant survival itself. The modern property rights system is the product of millions of years of this evolutionary pressure. The concept of property rights is so fundamental to the human experience that it is enshrined in the U.S. Constitution (Fifth Amendment). America’s Founding Fathers viewed property rights as a cornerstone of our system of governance and way of life.

Economists have also long argued that strong property rights are the cornerstone of economic growth. They are critical to enabling individuals to safely generate income, store wealth, and leverage those assets for credit and investments over time.

Several studies also support this idea. A study of more than 100 countries from 1990 to 2002 showed that countries with stronger property rights grew faster than those with weaker property rights, in part because they were better able to promote technological growth and improvements in resource allocation.

From the perspective of property rights, blockchain is a technology with core competitiveness. They are the most powerful technical foundation for digital information in the world, enabling immutable record keeping, cryptographically secure ownership, distributed security, and programmatic execution of rights through smart contracts.

As the United States enters the era of digital intelligence, blockchain can also serve as a standard environment for all AI infrastructure and applications, ensuring that American AI systems can benefit from the structural support of the five unique characteristics of cryptocurrency. Historically, the United States has created unprecedented opportunities, both for individuals and as a nation—from freedom from colonial rule, to the Constitution's promise of individual freedom, to the fight against segregation, to fierce market competition and entrepreneurship.

Today, standing on the threshold of AGI, I believe the United States has a significant opportunity to further solidify its leadership along these same dimensions. Combining U.S. AI policy goals with cryptocurrencies will inspire unprecedented levels of individual participation in open source networks, driving incentivized contributions at all levels of the AI ​​stack. Broad participation in our AI markets will stimulate competition, encourage new forms of bottom-up mobilization, and lead to wider social impacts.