From Exit to Occupancy: Why Crypto VCs Target Agent Networks
author:Sae Web3
Original title: Water Release Is Real Exit: When Encrypted VC Card Agent Network Effect
AI is an opportunity for Nerd, and Agent is an opportunity for money
Venture capital, A16Z, etc. MegaFunds have always told us stories about cycles and exits, but in Solo GP's opinion, it's more like harmonic vibrations of signals and structures; you need to find the real rules they haven't mentioned.
In 2021, a16z returned $12.5B of revenue for LPs. The DPI was higher than the sum of the previous ten years. At the same time, 2021 was also the beginning of a disaster in the US VC industry. Leaving aside the actual DPI, it was just a floating profit.
In other words, 2021 is a golden time to quit, and LPs can actually get real money, but if LPs get back involved, they will have to experience the pain that continues to this day.
Photo Caption: Discharging water is the only way to really quit
Photo credit: @jasonlk @PeterJ_Walker
All of this tells the opposite story, and the shock in the crypto market coincided with this. The 2022 metaverse concept fueled Web3 and even forced the bull market to continue. Until early 2025, Binance put an end to VC Coin with the “girlfriend coin” farce.
Today, most VCs have fallen into a silent model. Economies of scale are being dragged into a capital model that is heavy on computing power and data. There is no way to recover the costs. There is no way to talk about network effects on the chain, and towards institutionalization and SaaS channels.
However, looking at the history of venture capital, with each cycle of interest rate hikes and interest rate cuts, the water released will cultivate a different VC model. We will invent risky valuation logic over and over again, and the relative freedom of the crypto market will also allow interested people in this market to discover the most profitable signal mechanism.
When VC's No Longer Risky
“
Every passion begins with the impact of external objects on the sensory organs, causing the animal spirit to move through nerves.
If you still have an impression, in March and April 2021, Roblox and Coinbase chose the Direct Listing model. Unlike regular IPOs, direct listings only sell old shares, no underwriters, and no lock-up period.
Interestingly, both are led by A16Z. According to the impressive DPI data, in June 2021, A16Z raised $2.2 billion for the third crypto fund, and in January 2022, A16Z raised $9 billion in new funds.
So what's the cost?
The cost is that Coinbase's stock price fell 90% from its high in 2023. It can be stated very clearly that A16Z's role in US stocks is no different from that of crypto VC, but the problem is that A16Z can still raise $7.2 billion in 2024 and $151 billion in 2026.
Even in May 2026, its fifth crypto fund raised more than $2.2 billion, and its crypto fund's historical amount was close to $10 billion.
The market gave the choice between being the LPs of the A16Z, the amazing DPI waiting for the moment the water was released, or the cost of the A16Z and the source of the amazing DPI.
However, problems also followed. A16Z was not sensitive to market signals. In other words, the VC kings of every cycle faced the curse of scale. Due to excessive scale, they were not motivated enough to discover ultra-early paradigms, especially revolutions rather than improvement mechanisms.
Arthur Rock, the father of modern venture capital, debuted at its peak, and Fairchild and Intel launched the venture capital model in Silicon Valley;
KP and Sequoia officially introduced an institutional venture capital model, but they alternately took the lead in PC and mobile internet;
YC turned venture capital into probability under the big number mechanism to mass-manufacture sub-giant unicorns under power law;
Sun Zhengyi took SoftBank to turn venture capital into a large-scale similar Ponzi game through Ali's myth;
Just like this, when old giants indulge in their past glory, emerging ambitists will prove their unique vision through institutional innovation, and then get cheap money and start a new era of adventure for themselves.
Photo Caption: Changes in the VC Cycle
Photo credit: @zuoyeweb3
Reputation itself can even be exchanged for money. Paradigm founder Matt Huang voted for ByteDance. Although Byte cannot be listed, Paradigm chose to use cryptography. The latest news is that they have switched to AI and robots.
Let's fix the answer. If you can't be the LP of the A16Z, and you don't want to be the price of being trampled on, then you need to discover new signals that have not been amplified and use new mechanisms to destroy old seniors.
A gap has already emerged. In 2021, A16Z was not “allowed” to participate in Anthropic's financing. Instead, more individual investors placed early bets. For example, Skype co-founder Jaan Tallinn and former Google CEO Schmidt led the A round. FTX's SBF entered in 2022, giving us another kind of Crypro X AI that has endured for a long time.
Photo Caption: The card game has just begun
Photo credit: @zuoyeweb3
A16Z doesn't need to take risks. SBF “effectively benefits A\” with retail money. If you want to find the most reasonable starting point for Solo GP, Claude's venture capital history is the most typical.
Unlike personal angels, Solo GP runs the entire VC entirely with self-research ability. It's easy for us to understand in the Agent era, but it was just humans who first implemented it. Unlike the YC broad casting model, Solo GP still requires deep investment in every project, and every move is important to DPI.
A16Z has become an indicator of the market itself, new technological trends have emerged, and the updated participant ideas have managed to come a little earlier than A16Z's. Other than the big AI model, they are focusing on Agents.
There is a dangerous leap forward here, that is, economies of scale cannot exist in large AI models. For every additional human user, server costs increase, and costs cannot be diluted like software. In other words, network effects do not occur as scheduled in Agents, and calls between agents are still an ideal state.
Inhuman network effects
“
Watt improved the rotary steam engine in 1784, and the complete theory of the steam engine in 1824 was explained by the Frenchman Carnot.
Everything about AI is a black box. Scaling Law was observed by Adi Wang at Baidu. The math required for Transformers does not exceed the level of graduate students, but it is unknown why it can surpass the math level of graduate students.
AI is Nerd's opportunity. You only need to give money to a group of people at the cutting edge and wait for a miracle to happen. Popular talent acquisitions in Silicon Valley are the best proof. Inventions > Data > Model.
However, the big model itself is difficult to recoup the costs. The opposite of scale effects is emphasized once again. Even if we shift from training to reasoning, and even from dialogue to tasks, we cannot stop this process.
The only way out for the big AI model is to become a traffic center like AWS and CloudFlare. If production-side costs cannot be reduced, then the consumer side must grow infinitely.
Agents are an opportunity for money. Agents must become consumers. Unlimited agents X unlimited spending. This is the root cause of agents calling each other into the mainstream of the topic.
However, to a large extent, agents and bots are hard to tell apart, and it's hard to tell what an agent is; it seems like bots have existed for a long time.
Photo Caption: Bot is not an Agent
Photo credit: @Cloudflare
If you don't have to define an agent, the “evaluation agent” in reinforcement learning is the origin of this wave of technology. In DeepMind's idea, making the agent automatically evaluate the success of training is the key to the next intelligent upgrade.
This is very different from Claude's role division from Coding. From a programming perspective, Agent is actually a character mapping for human programmers. When we talk about Agentic Coding, it's far from the topic of AlphaZero's Agent.
Photo Caption: Agent High Value Scenario
Photo credit: @zuoyeweb3
It is only from this perspective that Agent generation, Claude's impact on SaaS makes sense; it is nothing more than a continuous iteration of the human resource outsourcing mechanism:
Move towards high-value scenarios, where programmers are followed by accountants and analysts;
Move towards fewer full time employees, and outsourcing is followed by multiple agent invocation fees.
However, one problem still exists. Agents don't show human social relationships. Real business relationships don't go smoothly without using Agent, and humans still like to deal with people.
We did create more Agent scenarios, which worked well in terms of “internal” coordination; for example, it was very easy to use when large manufacturers cut staff and replaced GPUs.
Photo Caption: Don't want people in high-value scenes
Photo credit: @trueupio
However, in terms of “external” collaboration, it is important to note that it has not been confirmed as expected. In May 2026, there was a strong increase in US employment, and non-farm payrolls increased by 172,000, mainly in blue-collar fields such as casual dining and medical care. Instead, the financial industry lost 22,000 people.
Human society's fear of Agents is real, but it's grossly overrated.
Of course, just as Sahara needed shoes, this could also be a signal to continue to enhance the model's intelligence, continue to increase Agent abilities, and invest in robotics.
In other words, Agent economics is only theoretically founded, and unlimited growth on the consumer side has not come true. If you continue to bet, how can agents use each other to create a network effect?
The era of cryptographic card agents
“
Evolution doesn't always lead to increased complexity, and evolution isn't always an upward trend.
Summarizing what we know, we are warning the danger of the unknown.
Venture capital does not represent the effective exploration of technical signals; it has become a game for a few brave people;
Agents were forced to be manufactured in large quantities, hoping to reduce production costs for large models, but the call relationship between them did not naturally occur.
These two seemingly contradictory discourse systems include ingenious coordination — searching for signaling mechanisms to stimulate Agent invocation.
There is no point in simply issuing Agent assets or agenting DeFi protocols. There are inherently few people and many bots on the chain, and superimposing smart contract calls will only increase technical risk. This path is not easy.
Practically speaking, human primacy will not be replaced by agents, because character mapping relies on business relationships. Xinchuang doesn't buy 4090, China and the US don't take over each other, and the boundaries of technology are narrower than we think.
Photo Caption: Agent Economy Card Contest
Photo credit: @zuoyeweb3
Exa targets the Agent's need for realtime+high-quality data, cleaning once, calling multiple times. This is a real economy of scale, but it's difficult to trigger calls between Claude and Codex.
Catena meets the financial requirements of agents between the B-side, and even requires applying for an OCC license to facilitate B2B compliance. This is a specialized network effect, but it is difficult to reduce large-scale usage costs.
However, the various types of payment agreements represented by stablecoins want to cut the demand for C-end entry+clearing exports. Lightweight agreements reduce usage costs, and micropayments reduce collaboration costs.
But that's not enough. In order to achieve ultimate A2A daily communication, humans must be willing to give their souls, similar to the three steps of TrueNorth:
Have them use an Agent to assist in transactions;
Let Agent learners participate in transactions;
Let agents dominate on-chain transactions.
Compared to Claude's access to IBKR's policy and legal restrictions, it can only be some kind of CoPilot, and TrueNorth is not difficult to use the actual version of Hyperliquid.
But being willing to be guided by agents is still far away, at least farther from VC's imagination.
Photo Caption: Payments+Transactions > Earnings
Photo credit: @zuoyeweb3
In the experiment with Agent+ finance, “mainly investment and payment, followed by investment and transaction” occupied the absolute mainstream structure.
Payments are very certain. Both PayPal and Stripe's market shares will be stablecoins, and stablecoins will all be agentized.
The deal has broad prospects, from Simmons to Jane Street, and Leung Sing's unending love fantasy, sparking VC's endless imagination.
But all of this isn't the same thing as we imagine an Agent taking over transactions and payments.
What is quantitatively established is “hegemony in computing power,” and it is still a speed advantage over humans. The establishment of transactions is a “channel advantage,” and it is still a preferential rate compared to banks.
As a result, a divide has arisen. VC has to contribute to one thing, making people willing to be actively replaced by agents. A16Z can't do anything about this. Spending money can't contribute to the success of the new social Clubhouse and Towns Protocol, then it can only directly flatten the more complicated financial agent scenario.
If you refer to the successful experience of DeFi, let the Agent access funds and verify the viability with low frequency and small amounts, then can it be used frequently and at large amounts on a daily basis.
As you can imagine, if all the roads were FSD Tesla Robotaxis, then it would be safer than a mix of humans/AI, but in order to make this happen, humans would instead be needed as test objects:
A small number of people use AI to assist driving and establish equal rights for human drivers in science and technology;
Reduce the casualty rate of a small number of people using AI-assisted driving, and establish a compensation mechanism.
In other words, establishing a mechanism for agents to make money will be easier to convert users than agents to make money. Only when agents gather enough experience with money can humans stop thinking and click to confirm along the way.
Only when Agents actively participate in the market can the efficiency and safety of the market be improved. It can be understood that the process of agents seeking revenue is a process of improving market efficiency. Gradually, they write C++ in C++ and use Agents to optimize agents.
The deal is where Agent ends, but before that, you have to go through a long elliptical runway.
In the high-value financial scenario, blockchain is an open financial testing ground, and stablecoins are proof for agents to optimize market processes. This has nothing to do with scale and resource investment; this is about the establishment and expansion of mechanisms.
epilogue
“
There is no cycle in life. Without a cycle, there will be no dividends of our times; a new generation will always outperform the old.
VC is getting smaller and more personal. Whether it's Solo GP or OPC, we haven't seen Solo GP investing in OPC become mainstream. Under the uncertain wave of technology, we don't know what paradigm will become mainstream.
“Software eats up the world”. After the internet bubble in early 2000, it instead ushered in long-term dividends for more than 20 years. Now it is a new era of “agents devouring software.”
The Agent itself is a development tool and a sign of productivity evolution, but no new software developed by Agent has become a national application. This is also true. After the three IPOs of SpaceX, OpenAI, and Anthropic, the large pedestal model card slot ended.
If it's a new round of long-term dividends, then newly raised crypto VCs such as Dragonfly, ParaFi, Haun, Paradigm, and a16z will continue to expand in scale, or whether they will run out of 5cc to predict market-specific racetrack funds, all of which will show strength in a new deployment frenzy.
Even, the entire DeFi industry will experience a new paradigm update. Financial system innovation has continued to change over the past two health cycles, and this time, Agents and stablecoins have become a new starting point for the dual revolution.
Crypto is very small, the world is big, let's witness it!
Twitter:https://twitter.com/BitpushNewsCN
Compare the TG exchange group:https://t.me/BitPushCommunity
Compare TG subscriptions:https://t.me/bitpush
