Surviving the Next Three Years of AI: The Liberman Brothers on Open Networks, Corporate Control, and Decentralization

Overview

The next three years of artificial intelligence will decide the balance of power between everyday users and a select group of trillion-dollar tech conglomerates. In this episode of Silicon Valley Girl, venture investors and tech founders David and Daniel Liberman join host Marina Mogilko to discuss the looming closing window of AI accessibility. Having previously built and sold their startup to Snap, the brothers are now focused on Gonka, a decentralized AI network designed to keep advanced models accessible and affordable for everyone. They argue that the current excitement around productivity gains masks a deeper, more concerning structural shift: the creators of the world's most powerful AI models are losing control to corporate stakeholders, setting the stage for artificial scarcity and unprecedented digital gatekeeping.

For anyone navigating the tech landscape, this discussion is an urgent wake-up call. The Libermans break down why the concept of open-source AI is rapidly disappearing, how physical-world industries like plumbing and manufacturing will eventually be absorbed by agentic systems, and why consumer action is the only remaining counterweight to an impending corporate monopoly over global compute. This episode provides a blueprint for understanding the transition from passive consumer to active participant in the decentralized future of technology.

Key Takeaways

The rapid transition of user trust from multi-source search engines to single-source AI chatbots is already fundamentally altering how humanity processes information and perceives reality.
AI founders are losing corporate control; Anthropic founder Dario Amodei reportedly owns only 1.5% of his company, leaving decisions in the hands of massive institutional stakeholders.
The top three cloud providers control nearly 70% of global compute, giving them the ultimate power to dictate pricing and access as AI integrates into every consumer application.
Truly open-source AI is virtually dead in the frontier model space, replaced by strategic open-weight releases designed to lock developers into proprietary cloud ecosystems.
Physical industries have a longer runway for adaptation than software engineering, but ultimately, every single job will face disruption within the next decade.
Decentralization, through community-run micro-data centers and sovereign hardware grids, represents the most viable path to maintaining digital freedom.

The Shift in Trust and the Slow Disruption of the Physical World

While many observers argue that artificial intelligence has yet to deliver a truly life-changing event for the average person, the Libermans point out that a massive, quiet psychological shift has already occurred. Millions of people have stopped cross-referencing information across multiple web search results, choosing instead to place absolute trust in the singular outputs of conversational agents. This shift in how we conduct research and acquire knowledge is pivotal, especially since these models are prone to hallucination and bias. When users accept a single answer without verification, whoever controls the model controls their perception of reality.

This cognitive disruption is moving faster than physical disruption, but the physical world is catching up. The Libermans share a personal anecdote about a recurring hot water issue in their home that multiple professional plumbers failed to solve over three years, repeatedly recommending expensive system replacements. By pointing a smartphone camera running ChatGPT at the piping system, the AI instantly identified a missing loop pipe. The plumbers, guided by the AI's diagnosis, resolved the issue in minutes. This demonstrates that while physical labor is still required, AI will act as an omnipresent, real-time guide in the earpieces of blue-collar workers, drastically lowering the barrier to expert-level execution.

However, this transition will not be entirely seamless. While software engineers were disrupted first because their output exists entirely in the digital realm, physical industries are protected only by the slower manufacturing times of robotics. The Libermans warn that everything will be disrupted, and every single job will be disrupted. While some professions have five to ten years to adapt as factory automation matures, others will face complete restructuring within the next twelve months.

The 1.5% Problem: Who Actually Controls Frontier AI?

One of the most startling revelations from the discussion centers on the ownership structures of frontier AI labs. The Libermans highlight that Dario Amodei, the co-founder of Anthropic and creator of the Claude model family, reportedly retains only a 1.5% equity stake in his own company. This dilution means that the visionary researchers who built these technologies and attempted to instill them with specific ethical values are no longer in control of their deployment or commercialization. Instead, control has shifted to massive corporate backers and complex long-term trusts whose ultimate allegiances lie with institutional shareholders.

When the people who built the technology are sidelined, the primary driving force of the industry shifts from public benefit to profit maximization. Today, a tiny handful of corporations—specifically Google, Microsoft, and Amazon—control roughly 65% to 70% of global cloud compute. Because compute is the foundational oxygen of the AI age, these entities hold the keys to the future of global productivity. If access to these models remains entirely centralized within these corporate clouds, these companies can easily capture the financial upside of increased human productivity simply by raising API costs and subscription fees.

The Open-Weight Illusion and Manufactured Scarcity

The tech industry frequently points to open-weight models as a safeguard against corporate monopoly, but the Libermans dismiss this as a temporary marketing strategy. Many developers believe they are building on open foundations, but open-weight is not the same as true open-source. When a company releases model weights, they are often doing so to build a developer ecosystem and challenge dominant proprietary models. Once they acquire a massive user base, they can easily close the gates on future versions, leaving developers stranded.

This pattern is already visible globally. Most major U.S. AI labs have quietly backed away from open-source releases. Meanwhile, international players often use semi-open models as a geopolitical tool, hosting them temporarily on public repositories to gain market share before funneling users toward centralized, sovereign data centers. If the industry successfully transitions to a closed ecosystem, these dominant players will gain the ability to create artificial scarcity. As the Libermans warn, if you create artificial scarcity, you can increase the price for as long as you want, locking users into expensive, inescapable digital ecosystems.

The Path to Decentralization: Reclaiming Digital Freedom

If the future of AI is left entirely to the market, the worst-case scenario will not look like a sudden, dramatic collapse, but rather a slow, imperceptible erosion of individual agency. Industry by industry will be absorbed, and users will find their daily workflows, personal data, and communication channels entirely dependent on a single corporate ecosystem. When your email, calendar, documents, and search tools are all tied to one underlying model, switching providers becomes nearly impossible due to high friction and data lock-in.

To counter this centralization, the Libermans advocate for the rapid development of decentralized AI networks. Rather than relying on massive, centralized data centers owned by tech giants, the future of open AI lies in community-driven infrastructure. This involves building grids of micro-data centers housed in residential homes, similar to how solar panels and home batteries allow individuals to disconnect from the traditional electrical grid. By distributing the computational load across thousands of individual nodes, the developer community can ensure that high-performance AI remains a cheap, universally accessible commodity that no single corporation or government can easily gatekeep or turn off.

Practical Applications

Audit Your Tool Stack: Evaluate the AI tools you use daily for coding, writing, or design. Identify where your data is siloed and actively seek out independent, interoperable alternatives to avoid ecosystem lock-in.
Support Open and Decentralized Projects: Whenever possible, run local models on your own hardware using tools like Ollama, and support decentralized compute initiatives that aim to democratize access to model training and inference.
Prepare for Agentic Workflows: Shift your focus from learning specific software UI to mastering system design, prompt engineering, and agent swarming. The ability to direct AI agents to perform complex multi-step tasks will be far more valuable than manual execution.
Upskill with Hybrid Physical-Digital Knowledge: If you work in a physical trade or real-world industry, begin integrating computer vision and real-time AI assistance into your workflow now to stay ahead of the automation curve.
Advocate for Data Sovereignty: Be vocal about user rights and data ownership. Support regulatory frameworks that protect consumer access to open weights and prevent monopolistic compute hoarding.

Final Thoughts

The critical window to shape the future of artificial intelligence is closing rapidly. If developers, creators, and consumers remain passive users of centralized corporate platforms, they will inevitably surrender their digital sovereignty to a handful of cloud monopolies capable of manufacturing artificial scarcity. True progress requires a concerted shift toward decentralized infrastructure, open-source development, and localized hardware grids. The next three years will determine whether AI becomes a universal utility that amplifies human potential, or a highly guarded toll road controlled by the world's largest corporations.

Source

Podcast: Silicon Valley Girl

Guest: David and Daniel Liberman

Channel: Silicon Valley Girl

Published: July 17, 2026