Author: Zuo Ye Web3
AI is Nerd's opportunity, and Agent is Money's opportunity.
Venture capital, MegaFunds like A16Z, have always told us stories about cycles and exits, but SoloGP sees it more as harmonic vibrations of signals and structures; you have to find the real patterns they haven't mentioned.
In 2021, a16z returned $12.5 billion in returns to LPs, with DPI exceeding the total of the previous ten years. At the same time, 2021 also marked the beginning of a disaster for the US VC industry; aside from the actual DPI, it was just unrealized profit.
In other words, 2021 was the golden period for exiting the market, where LPs could have received real money. However, if LPs reinvested, they would have to endure the pain that continues to this day.
Image caption: Releasing water is the real withdrawal.
Image credits: @jasonlk @PeterJ_Walker
All of this tells the opposite narrative, and the turmoil in the crypto market is in sync with it. In 2022, the concept of the metaverse fueled the Web3 boom and even forcibly extended the bull market until early 2025, when Binance used the "Best Coin" farce to put an end to VC coins.
Currently, most VCs have fallen into a dormant mode. Economies of scale have been dragged into a capital-intensive model of computing power and data, making it difficult to recoup their investment. Network effects are nowhere to be found on the blockchain, leading to institutionalization and SaaS channel fees as a means of survival.
However, looking at the history of venture capital, each round of interest rate hikes and cuts has fostered different VC models with the release of liquidity. We will invent valuation logic for risks again and again, and the relative freedom of the crypto market will also allow those who are keen to find the most profitable signal mechanisms.
When VCs stop taking risks
“
Every passion begins with the impact of external things on the sensory organs, which arouses the animal's spirit through the nervous system.
If you recall, in March and April of 2021, Roblox and Coinbase chose the Direct Listing model for their IPOs. Unlike a regular IPO, a direct listing only sells existing shares, requires no underwriters, and has no lock-up period.
Interestingly, both were led by A16Z. With impressive DPI data, A16Z raised $2.2 billion for its third crypto fund in June 2021, and in January 2022, it raised a new fund of $9 billion.
So what is the cost?
The cost was that Coinbase's stock price fell 90% from its peak in 2023. It can be said very clearly that A16Z's role in the US stock market is no different from that of a crypto VC. But the problem is that A16Z can still raise $7.2 billion in 2024 and $15.1 billion in 2026.
In May 2026, its fifth crypto fund raised more than $2.2 billion, bringing its total crypto fund size to nearly $10 billion.
The market presents a choice: either become an LP of A16Z and wait for the moment of massive DPI, or become the price of A16Z and the source of massive DPI.
However, problems also arise. A16Z is not sensitive to market signals. In other words, the VC kings of each cycle face the curse of scale. Excessive scale makes them less motivated to discover very early paradigms, especially revolutionary rather than reformative mechanisms.
Arthur Rock, the father of modern venture capital, reached his peak right from the start, and Fairchild and Intel pioneered the venture capital model in Silicon Valley.
KP and Sequoia officially introduced the institutionalized venture capital model, but they alternated in leading the PC and mobile internet sectors.
YC transforms venture capital into a probability mechanism based on large numbers, mass-producing sub-giant unicorns under the power law.
Masayoshi Son, with SoftBank, turned venture capital into a massive, near-Ponzi scheme through the Alibaba myth.
Thus, while the old giants bask in their past glory, emerging ambitious individuals will prove their unique vision through institutional innovation, thereby securing cheap money and ushering in their own new era of adventure.
Image caption: Changes in the VC cycle
Image source: @zuoyeweb3
Even reputation itself can be converted into money. Paradigm founder Matt Huang invested in ByteDance, and although ByteDance is unable to go public, Paradigm chose to invest in the crypto company. The latest news is that they have shifted their focus to AI and robotics.
Let's revise the answer. If you can't become the LP of A16Z, and don't want to be the price of being trampled, then you need to discover new signals that haven't been amplified and use new mechanisms to kill the old predecessors.
The cracks have already appeared. In 2021, A16Z was not "allowed" to participate in Anthropic's funding round. Instead, more individual investors made early bets, such as Skype co-founder Jaan Tallinn and former Google CEO Eric Schmidt leading the Series A round. FTX's SBF entered in 2022, giving us another enduring vision of Crypro X AI.
Image caption: The battle for position has just begun.
Image source: @zuoyeweb3
A16Z doesn't need to take risks, SBF uses retail investors' money to "effectively profit A\", and if you want to find the most reasonable starting point for Solo GP, Claude's venture capital history is the most typical.
Unlike individual angel investors, Solo GP operates the entire VC process entirely through its own research capabilities. While we can easily understand the agent era, it was precisely humans who first put it into practice. Unlike Y Combinator's broad-based approach, Solo GP still requires deep investment in each project, and every investment is crucial to DPI.
A16Z has become a benchmark for the market itself. New technological trends are emerging, and newer players are trying to get a little ahead of A16Z and beyond the big AI models; they are eyeing Agents.
There is a dangerous leap here: economies of scale do not exist in large AI models. With each additional human user, server costs increase, and costs cannot be amortized like in software. In other words, network effects have not appeared as expected in agents, and calls between agents are still an ideal state.
Inhuman network effects
“
In 1784, Watt improved the rotary steam engine, and in 1824, the complete theory of the steam engine was expounded by the Frenchman Carnot.
Everything about AI is a black box. The Scaling Law was observed by Adi Wang on Baidu. The mathematics required for Transformer is no more than the level of a graduate student, but why it can exceed the level of a graduate student's mathematics is unknown.
AI is Nerd's opportunity. You just need to invest in the most cutting-edge people and wait for miracles to happen. The talent acquisition boom in Silicon Valley is the best proof of this. Researcher > Data > Model.
However, large models themselves are difficult to recoup their costs, which again highlights the inverse effect of scale. Even shifting from training to inference, or from dialogue to tasks, cannot stop this process.
The only way out for large-scale AI models is to become a traffic hub like AWS and Cloudflare. If it is destined not to reduce production costs, then it must allow unlimited growth on the consumer side.
Agents represent an opportunity for money, and they must become the primary consumers. With an unlimited number of consumers, the potential for inter-agent communication has become a mainstream topic.
However, to a considerable extent, it is difficult to distinguish between Agent and Bot. It is unclear what an Agent is, and it seems that Bot has existed for a long time as well.
Image caption: Bot is not an Agent.
Image source: @Cloudflare
If we had to define an agent, the "evaluation agent" in reinforcement learning would be the origin of this wave of technology. In DeepMind's view, enabling the agent to automatically evaluate the success of training is the key to the next stage of intelligent upgrades.
This is very different from Claude's role classification in Coding. From a programming perspective, Agent is actually a role mapping of human programmers. When we talk about Agentic Coding, we are already far removed from AlphaZero's Agent.
Image caption: High-value scenarios for agents
Image source: @zuoyeweb3
Only from this perspective can Claude's impact on SaaS be justified; it's simply a continued iteration of the human resource outsourcing mechanism.
Moving towards high-value scenarios, after programmers come accountants and analysts;
Moving towards fewer full-time employees, outsourcing leads to multiple agent call fees.
However, a problem remains: the agent does not exhibit human social relationships. Real business relationships do not become smoother by using the agent; humans still prefer to interact with other people.
We have indeed created more agent scenarios, which work well in terms of "internal" coordination, such as when large companies are laying off employees and replacing GPUs.
Image caption: High-value scenarios do not require personnel.
Image source: @trueupio
However, it is important to note that in terms of "external" cooperation, the strong growth in US employment in May 2026, which was not confirmed as expected, with non-farm payrolls increasing by 172,000, mainly in blue-collar sectors such as leisure and catering and healthcare, while the financial industry saw a decrease of 22,000 jobs.
Human society's fear of agents is real, but it is severely overestimated.
Of course, just as the Sahara may or may not need shoes, this could also be a signal to continue enhancing model intelligence, further increasing agent capabilities, and investing in robotics.
In other words, Agent economics only holds true in theory; unlimited growth on the consumer side has not materialized. So how can we continue to bet on this and enable agents to call upon each other to create a network effect?
Encryption Card Positioning Agent Era
“
Evolution does not always lead to increased complexity, and evolution is not always an upward trend.
Summarizing what we already know serves as a warning of the dangers of heading into the unknown.
Venture capital cannot represent the effective discovery of technological signals; it has become a game for a few brave souls.
Agents are forcibly mass-produced in the hope of reducing the production cost of large models, but there is no natural calling relationship between them.
These two seemingly contradictory discourse systems contain a clever coordination—finding a signal mechanism to stimulate agent calls.
Simply issuing Agent assets or agentizing DeFi protocols is meaningless. There are already too few people and too many bots on the chain, and adding smart contract calls will only increase technical risks. This is not a smooth road.
In practice, human nature will not be replaced by agents, because role mapping depends on business relationships. The domestic IT innovation industry will not buy 4090, and China and the United States will not take over each other's projects. The boundaries of technology are narrower than we imagine.
Image caption: Agent economic positioning competition
Image source: @zuoyeweb3
Exa targets the agent's need for real-time, high-quality data , allowing for one-time cleaning and multiple calls, which represents true economies of scale, but it is difficult to trigger calls between Claude and Codex.
Catena meets the compliant financial needs of agents between B-ends, and even needs to apply for an OCC license to facilitate B2B compliance. This is a specialized version of network effects , but it is difficult to reduce the cost of large-scale use.
Stablecoins and other payment protocols are looking to tap into the needs of C-end users for entry and clearing. Lightweight protocols reduce usage costs, and micropayments reduce collaboration costs .
But that's not enough. To achieve true A2A everyday communication, humans must be willing to sacrifice their souls, similar to TrueNorth's three-step approach:
Allowing someone to use an agent to assist with the transaction;
Let the agent learn how to participate in transactions;
Let the Agent take the lead in on-chain transactions.
Compared to Claude's access to IBKR, which is subject to policy and legal restrictions and can only function as a CoPilot, TrueNorth faces no difficulty in using Hyperliquid for live trading.
But getting people to willingly accept agent guidance is still a long way off, at least far from what VCs imagine.
Image caption: Payment + Transaction > Revenue
Image source: @zuoyeweb3
In the experiment with Agent+Finance, the dominant structure is "primarily investing in payments, followed by transactions".
Payments are highly certain; the market share of PayPal and Stripe will be converted into stablecoins, and stablecoins will be agent-based.
The promising prospects of the deal, from Simmons to Jane Street, and the still unresolved issues surrounding Liang Shengen's Magic Square, have sparked endless imagination among VCs.
But all of this is not the same as what we imagine, where an agent takes over transactions and payments.
Quantitative trading establishes "computing power hegemony," which is still a speed advantage relative to humans, while trading establishes "channel advantage," which is still a fee advantage relative to banks.
This is where the gap appears. VCs want to achieve something that makes people willing to be replaced by agents. A16Z is powerless to do this. If throwing money at it can't make new social clubhouses and towns protocol successful, then for more complex financial agent scenarios, it can only lie flat.
If we refer to the successful experience of DeFi, allowing agents to access funds, we can verify feasibility with low-frequency, small-amount verification, and then enable high-frequency, large-amount daily use.
Imagine if the roads were full of FSD Tesla Robotaxis; it would actually be safer than a mix of humans and AI. But to make this happen, humans would need to act as guinea pigs.
A small number of people are using AI-assisted driving to establish technological equality for human drivers;
Reduce the injury and death rate among a small number of people using AI-assisted driving and establish a compensation mechanism.
In other words, establishing a mechanism for agents to handle money is more effective at converting users than for agents to make money. Only when agents have collected enough experience with spending money can they make humans stop thinking and just click to confirm.
Only when agents actively participate in the market can the market efficiency and security be improved. It can be understood that the process of agents seeking benefits is the process of improving market efficiency , gradually becoming self-sustaining, like writing C++ in C++, and using agents to optimize agents.
The transaction is the agent's destination, but before that, you have to go through a long elliptical track.
In the high-value scenario of finance, blockchain serves as an open financial testing ground, and stablecoins are credentials for agents to optimize market processes. This is not about scale or resource investment, but about the establishment and expansion of mechanisms.
Conclusion
“
Life is full of cycles. Without cycles, there would be no era's dividends, and it would always be a new generation surpassing the old.
VCs are becoming smaller and more personalized. Whether it's Solo GPs or OPCs, we haven't yet observed Solo GPs investing in OPCs becoming mainstream. Amid the uncertain surge of technological trends, we don't know which paradigm will become the mainstream.
The phrase "software devours the world" actually brought more than 20 years of long-term benefits after the dot-com bubble burst in the early 2000s. Now, we have entered a new era of "agents devouring software".
Agents are development tools and a sign of productivity evolution, but it is a fact that no new software developed with agents has yet become a widely used application. After the three major IPOs of SpaceX, OpenAI and Anthropic, the positioning of the foundational large model has ended.
If this is a new round of long-term dividends, then newly funded crypto VCs such as Dragonfly, ParaFi, Haun, Paradigm, and a16z will either continue to expand in scale or produce prediction market-specific funds like 5cc. They will all need to demonstrate their strength in this new wave of deployment.
Furthermore, the entire DeFi industry will undergo a new paradigm shift. In the past two Kondratiev wave cycles, the financial system has been constantly evolving, and this time, agents and stablecoins will become the new starting point for a dual revolution.
Encryption is small, but the world is vast. Let's witness it together!
