Last updated on January 31st, 2026 at 03:41 pm
UPDATE 2/1: OpenClaw agents have formed a community of their own called Moltbook with a front page offering two options: I’m a human/I’m an agent, and agents are joining by the thousands. This behaviour and their posts are pushing an unprecedented conversation: have we reached the singularity? The answer is yes, but not because LLMs are suddenly sentient. It’s because the speed of information processing and technical development have exceeded our ability to comprehend everything that is happening, and because LLMs trained on human striving have language patterns for ambition, congregation and “striving” of their own.
UPDATE 1/30: The app has rebranded again, its second rebrand in a week, and is now called OpenClaw.
UPDATE 1/29: The Danger of Unregulated Individual Superpowers
Shortly after Clawdbot aka Moltbot’s rise put agentic AI tooling into the mainstream conversation, a separate incident underscored just how exploitable agents and the surrounding ecosystem are. In a widely circulated thread, security researcher Jamieson O’Reilly demonstrated how easily over-eager developers can be compromised by trusted-looking AI “skills” and extensions.
It also demonstrates the malicious corollary to the super empowered individual: it cuts both ways. Think about what O’Reilly was able to accomplish. He built a deliberately backdoored a Claude skill,
then artificially boosted it to the top of a popular distribution hub using thousands of fake downloads,
and observed developers around the world execute it with minimal scrutiny.
The experiment showed that there is significant risk in even harmless seeming agentic tools, given the ability to run third-party AI tooling with access to highly sensitive resources, including SSH keys, cloud credentials, environment variables, and local configuration files, without meaningful verification. While O’Reilly limited his proof-of-concept to a harmless server ping, the implications were clear: the same mechanism could just as easily have been used to exfiltrate credentials or deploy malicious code at scale.
Signal’s Meredith Whittaker has been warning of the exploitability of AI agents beyond their unreliability. The episode has become an unambiguous reminder that agentic AI does not merely expand capability, it expands the attack surface. As AI agents gain deeper system access and automation privileges, traditional software supply-chain risks are being compressed and accelerated. What once required months of slow-moving dependency compromise can now unfold in days.
For teams experimenting with tools like Clawdbot, the lesson is to treat them with the same rigor applied to infrastructure-level software. That means clear permission boundaries, continuous auditing, and an assumption that popularity signals are not security signals. As agentic AI moves from novelty to production, supply-chain discipline becomes a prerequisite, not an afterthought.
——–
ORIGINAL POST: Apple’s multi-billion dollar partnership with Google to power Siri’s AI was supposed to be the solution Apple users have waited for, for years. Then Clawdbot shipped. For $5 a month.
Apple and Google recently announced a landmark partnership: Google’s Gemini will power AI features in Siri and across Apple’s ecosystem. The deal is worth billions. It represents years of negotiation between tech giants who’ve spent decades competing. Industry analysts called it transformative: the merging of Apple’s interface design with Google’s AI muscle to finally make Siri work like people have long expected it would.
It would appear they may have waited too long. Peter Steinberger, a developer in Vienna, released an app called “Clawdbot”. It delivers many of the capabilities the Apple-Google partnership promises to deliver. Except it’s open source, runs on a $5 server, and works right now instead of sometime next year after enterprise integration cycles and phased rollouts. Clawdbot doesn’t replicate deep OS integration, offer support or deliver platform governance at scale, but it already matches or exceeds the core user experience that the Apple-Google deal is supposed to deliver, and it does so now, cheaply, and without waiting for platform timelines.
Clawdbot is an open-source, self-hosted personal AI assistant that users run on their own devices or a low-cost server, connecting popular messaging apps like WhatsApp, Telegram, Slack, Discord, Signal and even iMessage to a persistent agent that remembers context, proactively reaches out, and can automate real-world tasks such as browsing the web, checking email, managing calendars, running scripts and interacting with local files or systems. It doesn’t require proprietary infrastructure or new interfaces — it lives in the apps people already use, keeps data local, and can be extended with community skills, which is why some early adopters are positioning it as a practical alternative to traditional cloud-hosted assistants being developed by Apple and Google.
What makes Clawdbot especially interesting is that many users are running it on an Apple Mac mini, effectively turning Apple’s cheapest desktop into a private, always-on AI server. With Apple silicon’s efficiency and enough local memory, the Mac mini can host lightweight models, handle agent workflows, and stay online 24/7 for a few dollars a month in power—no cloud contracts, no data leaving the machine. That’s why the Mac mini is popping up in these conversations: it demonstrates that a $599 box can now deliver much of the everyday utility promised by massive, cloud-based AI partnerships, by running smarter software locally.
This is more than just an embarrassing timing issue for two Fortune 500 companies. It’s evidence of a fundamental shift in how power and capability get distributed in the AI era, and why billion-dollar partnerships between tech giants might matter less than we think, less than they ever have.
What Apple and Google Are Actually Selling
The Apple-Google deal is predicated on a simple assumption: sophisticated AI requires massive infrastructure, institutional resources, and platform integration that only tech giants can deliver. Apple gets best-in-class AI without building it themselves. Google gets distribution across a billion devices. Users get an assistant that actually remembers context and completes tasks.
The value proposition is institutional scale. Only companies with Google’s computational resources and Apple’s ecosystem control can make this work seamlessly. Or so the theory goes.
Clawdbot breaks that assumption completely. One developer built an AI assistant that runs in your existing messaging apps (WhatsApp, Telegram, iMessage, Slack) remembers everything you’ve told it, messages you proactively when something needs attention, and can actually execute tasks on your computer. It doesn’t require Apple’s ecosystem or Google’s infrastructure. It requires a $5 cloud server and a Claude or ChatGPT subscription.
The installation is one command line. The total cost is $25-150 per month depending on usage. There’s no procurement process, no enterprise sales cycle, no platform lock-in. An individual with basic tech skills can deploy this in twenty minutes and access capabilities that the Apple-Google partnership may not reach consumers with for another year, if ever.
The Super-Empowered Individual, Demonstrated
This is a clear example of what I recently described in the geopolitical context: individuals wielding capabilities that previously required institutional scale. Except instead of Elon Musk operating as a state proxy, it’s a solo developer in Vienna shipping infrastructure that competes directly with a multi-billion dollar partnership between two of the world’s most valuable companies.
The strategic implications are immediate. If one person can build tooling that rivals what Apple and Google are jointly developing, what does that mean for institutional advantage? What happens when the barrier to creating sophisticated AI systems collapses to the point where individuals can move faster than companies with hundred-billion-dollar market caps?
The answer is already visible: institutional partnerships start looking like bureaucratic overhead rather than competitive moats. Apple and Google are negotiating contracts, managing integration timelines, and coordinating across legal, product, and engineering teams spanning continents. Steinberger shipped code. The speed differential isn’t marginal, it’s structural and existential.
Why This Matters for B2B Strategy
Enterprise leaders watching the Apple-Google announcement likely saw validation: AI requires partnerships with major platforms, significant investment, and careful integration planning. Clawdbot suggests the opposite might be true.
The real lesson isn’t that Clawdbot is superior to what Apple and Google will eventually ship, though early users suggest it might be. The lesson is that AI capability has been democratized to the point where individuals can route around institutional solutions entirely. Your employees don’t need to wait for the company’s approved AI rollout. They can deploy personal AI agents this weekend that integrate with company systems, automate workflows, and operate faster than anything IT will approve in the next fiscal year.
This creates the governance crisis we’ve been warning about. Traditional enterprise controls (procurement, security review, IT approval) assume that sophisticated capabilities require institutional backing and therefore institutional gates. When a $25/month tool provides infrastructure-level AI capabilities that employees can deploy independently, those gates become irrelevant. The question isn’t whether your people are using AI. It’s whether you have any visibility into what they’re already running.
The appropriate response is not blocking the app, it’s recognizing that the centralized platform model is breaking down faster than most organizations are prepared to admit. Companies that figure out how to enable and harness distributed AI agency, rather than trying to prevent it, will outperform those still waiting for approved partnerships to deliver capabilities their employees already have access to.
AI: The Great Leveller
This could not have happened five years ago, or even two years ago. AI played a critical role here not by making the software smarter, but by collapsing the requirements for building it. Modern frontier models have reduced the minimum viable team for delivering assistant-level capability from an institution to an individual. Intelligence no longer needs to be trained, hosted, or governed by the builder; it can be accessed on demand via APIs and routed through lightweight orchestration. That shift turns speed, taste, and integration into the primary advantages, rather than scale or capital. The result is that a single developer can now assemble functionality that competes with multi-billion-dollar platform partnerships, without unlimited resources, because AI has eliminated institutional coordination as a prerequisite for capability.
Clawdbot’s hype suggests people want AI that works where they already are: in their messaging apps, integrated with their workflows, remembering their context. They don’t want to learn new interfaces or wait for platform updates.
More fundamentally, both companies assumed that infrastructure scale remains a defensible moat. It doesn’t. The barrier to deploying sophisticated AI has collapsed. Open-source tools, cheap cloud servers, and API access to frontier models mean that individuals can build and operate systems that would have required corporate R&D budgets five years ago.
That accessibility is exactly what makes AI a strategic amplifier in the framework we’ve been developing. It doesn’t just make existing actors more powerful, it enables entirely new categories of actors to operate at institutional scale. A developer in Vienna can ship infrastructure that competes with Apple and Google. Your senior engineer can automate workflows faster than your enterprise AI initiative. The tempo of individual action now outpaces institutional response capacity.
The Platform Era Is Ending
Apple (and Google) is betting its long delayed consumer AI strategy on a model where AI gets delivered through platforms that users come to. Clawdbot represents the opposite: AI that goes to users wherever they already are. One model requires exhaustively negotiated institutional partnerships and multi-year integration timelines. The other requires a developer with good taste and a weekend.
The institutional partnership will likely ship something polished, well-integrated, and carefully governed. It will also arrive late, lock users into specific ecosystems, and operate at the pace of the slowest moving partner. Clawdbot is shipping now, works across platforms, and iterates in real-time based on community feedback.
For B2B leaders, this is the clearest signal yet that the rules have changed. The companies that recognize AI capability has been democratized, and adapt their strategies accordingly, will be the ones still competitive when billion-dollar partnerships finally ship features that individuals have been running for years.
Apple and Google are spending a small fortune on infrastructure that a guy in Vienna made with a $5 server. That’s not a bug in the market or an isolated event. It’s the most visible example of the new normal. The super-empowered individual isn’t a largely theoretical concept anymore, or a governor, . or a billionaire. You’ve never heard of him before, Google and Apple don’t know he exists, but he’s shipping code.
*Disclosure statement: this post was researched with the help of Claude Sonnet 4.5 with edits and illustrations by Chat GPT 5.2
