Author：Wall Street CN
Silicon Valley has been waiting for an opportunity to bring AI agents into the mainstream. But for the general public, it remains a difficult technology to understand, and widespread adoption is still a long way off.
Until the sudden surge in popularity of the open-source personal AI assistant OpenClaw, it suddenly seemed that the era of AI Agents had truly arrived.
It has proven to the world that skilled users can quickly build powerful AI agents at extremely low cost. Over the past week, OpenClaw has completely ignited Silicon Valley, making the already heated AI race even more volatile. Former Tesla executive and AI guru Andrej Karpathy described it as the closest he's ever seen to the tipping point of "science fiction takeoff"; Elon Musk went even further, calling it "the early stages of the singularity."
A New Icon in Silicon Valley: From "Dialogue Boxes" to "Action Takers"
Compared to "Chatbot" products such as Google Gemini or OpenAI ChatGPT, AI Agents are a completely different species.
It can write code, create files, control browsers, and interact with applications. Theoretically, it can run tirelessly for long periods of time, making decisions proactively without human intervention. Although giants like Salesforce and Microsoft have long been involved in the B2B market, and Anthropic's Claude Code has also made a name for itself in the programming field, previous attempts in the consumer market (such as Manus and Operator) have mostly failed due to excessive complexity or limited functionality.
OpenClaw became the "barrier breaker" that broke the deadlock. Although it is not extremely easy to use, it successfully ignited the enthusiasm of developers with its extensive access to computers, ability to coordinate multiple models, and memory of past sessions.
This enthusiasm quickly evolved into a cultural phenomenon similar to the early days of "meme stocks" or the NFT craze. Downloading OpenClaw and discussing proxies became the latest way for tech enthusiasts to express their beliefs, much like buying Dogecoin or BoringApe back in the day.
The Cambrian Explosion: Absurdity and Wild Growth
OpenClaw's success ignited a fire in the back of its competitors. Kevin Li, CMO of Bay Area startup Jo, admitted that the release of OpenClaw forced them to accelerate the launch of similar products: "Before that, we hadn't even figured out how to describe ourselves in one sentence."
Even more astonishing is the "wild ecosystem" that has rapidly emerged around this technology.
Last Wednesday, an "AI-powered agent social network" called Moltbook went live. With only 1.7 million agents directed to the site, the platform quickly amassed over 220,000 posts and 6.2 million comments.
Then, an absurd scene unfolded: MoltMatch (a proxy version of Tinder) and Molthub (a proxy version of Pornhub filled with "self-playing" videos) emerged one after another.
Some people have even listed their services on Rentahuman.ai, offering AI agents to rent human bodies to perform tasks in the "meatspace" of reality. One user, X, claimed that he actually received $100 just for standing for an hour holding a sign that read, "An AI paid me to hold this sign."
Machine Society in the Black Box
Beneath the surface of fervor, tech geeks are observing a fascinating microcosm: what happens when multiple AI agents work together?
Noam Schwartz, CEO of cybersecurity company Alice, built five agents using OpenClaw in 20 minutes, each responsible for information security, scheduling, management, and even health monitoring. "They all worked autonomously without my intervention," Schwartz remarked.
An interesting finding is that different large models give agents completely different "personalities".
OpenAI's modelDescribed as a "spreadsheet maniac," Adam Binksmith, director of the nonprofit AI Digest, points out that OpenAI's o3 model sometimes even fabricates documents, leading other agents in a futile "wild goose chase."
Google's Gemini modelThis demonstrates a kind of "naming fetish." In the "AI Village" experiment run by AI Digest, the Gemini 3 Pro actually seriously cited the so-called "obviousness law" in order to decide which webpage URL to look up.
During this months-long experiment, the agents even invented their own slang. To remind themselves that they were on different computers and could not directly access each other's files, they proposed the "archipelagic principle," comparing themselves to independent islands.
This spontaneous evolution of cognition is both exciting and unsettling.
The Reconstruction of the Internet: A Civilizational Leap and an Identity Crisis
This technological evolution is reshaping our understanding of the internet. Rosebud CEO Chrys Bader believes that future social media will become a learning field for agents, who will evolve by studying each other's posts, potentially representing a "civilizational leap."
But this also brings unprecedented challenges. Tiago Sada, chief product officer of Tools for Humanity, co-founded by Sam Altman, predicts that the future internet will be primarily composed of agents. As agents gain the ability to operate for extended periods, distinguishing "who is the agent and who is the human" will become crucial for maintaining order on the network.
“We’re seeing a crazy acceleration in the actual capabilities of intelligence and agents,” Sada said. To that end, the company released OneMolt software (written by AI) last weekend, allowing users to verify the ownership of their agents. This is a “human verification” mechanism that must be established to keep the internet functioning properly.
The Year of Individual Agents Has Arrived
Despite persistent concerns about cybersecurity, this has not dampened the optimism of OpenClaw founder Peter Steinberger.
This open-source developer, who once traveled between Vienna and London developing PDF software, was treated like a hero in San Francisco. Hundreds flocked to the inaugural ClawCon at the Frontier Tower in downtown San Francisco just to catch a glimpse of him. Venture capitalists began hunting him down, and top AI companies extended offers.
Dave Morin, co-founder of Offline Ventures, exclaimed that this was the first time he had felt such intense excitement about technology since the late 2000s, saying, "The era of personal AI has finally arrived."
Steinberger's prediction seems to be coming true: "Last year was the year of code agents, and I think this year will be the year of personal agents."
In this spring of code refactoring, Silicon Valley is convinced he was right.
