excerpt regarding openclaw and chinese AI from forthcoming book The Garden Without Gates

here is an excerpt from a forthcoming book. open claw uses and watchers might be interested or they might not like it at all. It is not posted as promotion but to engender discussion as I see some comments from time to time regarding things related to the text. The Garden Without Gates: AI in a World Under Heaven, Martin Hardie with Patrick Zhukov Bartley I have already released a graphic version and the full text is on its way https://martinhardie.substack.com/p/the-garden-without-gates-a-graphical The intro/readers note is available in full here: https://martinhardie.substack.com/p/readers-note this section comes from the Chapter 9: And Then China Happened: "... By 2018, China was the second-largest source of GitHub activity globally, despite the platform’s intermittent blocking by the Great Firewall and the 2015 “Great Cannon” DDoS attack; a tool that hijacked ordinary Chinese web traffic and turned it into a weapon to take GitHub offline for days. It was during this period that the cultural identity of Chinese developers crystallised. The 996.ICU movement’s second act connected labour organ- ising to censorship infrastructure. The Great Cannon had weaponised ordi- nary Chinese traffic against GitHub itself; now the same platform hosted a labour movement that embedded workers’ rights in open source code. The movement’s Supreme Court vindication in 2021 was not merely a labour victory. The Anti-996 License’s genuinely novel contribution was opening a gateway to what we might call for now social open source licencing. ... March 26th is Anti-996 Day. I am writing this on March 26th, 2026. The Chinese programmer who created the 996.ICU project chose this date. On this day, seven years ago, a programmer in China started a repository that 193 became one of the most-starred in GitHub history, a labour movement named for dying in the ICU. The Anti-996 License required that anyone using the code must comply with national and international labour law. Prophetically it was this morning, 26 March 2026 that I first read Steinberger’s comments regarding Europe’s ‘crippling labour regulations’. In the context of this dis- cussion these comments by the author of a tool that could be used to build the garden and who went to the factory because Europe’s labour protections were inconvenient need no further comment at this moment. The date writes itself. ... The OpenClaw Frenzy The Hudson argument arrives at a structural claim: the Western model can- not sustain open competition with publicly-directed alternatives. But the claim remains abstract until tested against a concrete case. OpenClaw — an open-source autonomous AI agent that lives in your messaging infras- tructure rather than a browser — became that test. Here is what happened when the same tool met two different systems. OpenClaw does not have the architectural ambition of DeepSeek or the corporate scale of OpenAI. It is a tool built by one developer, Pete Stein- berger, that lets users run an AI agent locally on their own machine, con- nected to their own communication channels — Signal, Telegram, Discord, WhatsApp — storing all data on their own hardware. It does not extract data for someone else’s model. It does not lock you into someone else’s plat- form. It is, in the architectural sense, the garden: infrastructure designed for use, not for capture. The garden’s creator, however, did not stay there. As OpenClaw began to take off, Steinberger was besieged with offers from corporate AI laboratories to buy the code. His default was the American free-as-in-freedom tradition: he publicly stated that he wanted the claw to remain open source. He pointed to the Chrome/Chromium model — where the open-source engine (Chromium) remains available while the proprietary browser (Chrome) captures the market, the user base, and the revenue — as the template. Open core, not locked down. Community-driven, not corporate-owned. In February 2026, he announced he was joining OpenAI and moving to the United States. Soon after, he described Europe’s labour protections against six-day weeks as “crippling labour regulations.” “In Europe I get in- sulted,” he wrote. “People shout REGULATION and RESPONSIBILITY.” In Europe, he said, that would be illegal (as it is now illegal in China, where the 996.ICU movement had secured a Supreme People’s Court ruling against the same practice). The statement that “the builder of the garden chose the factory” is too simple. The garden’s creator found the rules of the garden — regulation, responsibility, mutual obligation — inconvenient. He traded them for the factory’s promise: no constraints, no duties, just production. The domesti- cated nerd is not tragic. He is willing. He found the ideology convenient. The protections Steinberger fled are the same protections that would have covered the Kenyan data labeller, the Madagascan annotator, the 996.ICU developer. He left them. The scam compound worker never had them. The tool that could build the garden was abandoned by its own architect. As Patrick suggests, Bifo might have said that the hacker’s ethic is not political but tragic. The act of creating the tool is its own reward. The builder knows the factory will absorb what he has made; knows the tool will be co-opted, the garden paved. He builds anyway, not despite this knowledge but because the intensity of the gesture is what he sought. Redemption lies not in what the tool becomes but in the moment of its creation. The will to build is abundant. What is scarce is the will to maintain, to stay with the thing after the intensity fades, to keep the garden weeded when no one is watching, to accept regulation and responsibility as the price of a commons that lasts. But there is another reading, which our friend Fernando, a neurologist in Madrid, offered after reading this passage. As he observes, the fork is not closed. The same clinical evidence that predicts atrophy under passive con- sumption also predicts growth under active interrogation. The hacker who does not consume the tool but interrogates it, contradicts it, forces it be- yond its statistical patterns. This is Trotsky’s permanent revolution applied to the psyche: the subject who continuously refuses the passive position, who uses the tool as a dialectical mirror rather than a dopamine dispenser. The 4 percent difficulty rule, which Fernando draws from Csikszentmihalyi, holds that optimal challenge sits just beyond current capacity. The ques- tion is whether the tool reduces the challenge below that threshold or raises the floor high enough that a challenge once unreachable becomes attainable. The same tool produces both outcomes. The difference is not the code. It is the disposition of the user. Steinberger did not fail to understand the garden. He understood it well enough to build it. He simply wanted something different from what the garden demanded. The Anti-996 License — which required anyone using the code to comply with labour law — was the opposite gesture. It offered not intensity but obligation. It was harder to create, harder to celebrate, and harder to abandon. The West recorded the event as a standard acquisition story: creator of hot open-source project joins major AI company. The GitHub stars narra- tive captured the numbers — OpenClaw accumulated 275,000 stars within four months, surpassing 996.ICU’s 247,000 — but Western press reported the previous record holder as Next.js, with no mention of 996.ICU. The workers got written out of their own victory. OpenClaw had beaten the repo that had fought for the right not to die in the ICU, and the tech press called it a success story. While OpenAI absorbed the creator, Chinese municipal governments were doing something structurally different: they were subsidising the users.