Hybrid [Meaning: see profile] “Where will power go after AGI?” Inspired by Sreeram Kannan’s Manifest 2026 presentation Until now, @eigencloud has been widely perceived as a project focused on using blockchain to verify computation and data. But the narrative presented in this talk was far broader than any specific product or technology—it was about who will hold power after the emergence of AGI. The presentation began by exploring how humans have been able to cooperate at such large scales. It explained that what set humans apart from other species was not individual intelligence, but the ability to coordinate toward shared goals—even with strangers. Nations, corporations, and markets are ultimately the result of countless individuals following the same rules. The problem, however, is that as cooperation grows, so does the need for trust. If you cannot be sure others will honor their commitments, the scope of whom you can work with shrinks. Sreeram argued that blockchain can change this: if rules and outcomes can be publicly verified without needing to fully trust any single organization, more people can collaborate together. This line of thinking naturally led to AGI. Discussions around AGI have largely centered on safety and control—how to manage an intelligence superior to our own. Sreeram added one more critical question: “Who will control the world after AGI?” The graph presented illustrated how power and agency are distributed. Here, agency refers to the ability to influence one’s environment and outcomes according to one’s own will. He outlined two possible futures after AGI: One is a future where powerful intelligence concentrates in the hands of a few corporations and capital holders, while everyone else becomes passive users of their services. The other is a future where more people can directly use AGI to conduct research, build companies, and solve problems they care about. The video suggested that, if current trends continue, we are far more likely to head toward the first scenario—because the resources required to build AGI are already concentrated in the hands of a few. To illustrate this, he presented a five-layer structure: energy → semiconductors → data centers → AI models → applications—alongside the capital driving all of it. Energy is needed to power semiconductors; large-scale semiconductors are required to train models; advanced models then generate breakthroughs like new drugs or scientific discoveries. Initially, different companies operated at each layer. But as scale increased, boundaries blurred. Model companies began moving downstream to secure stable computational resources—data centers and semiconductors—and didn’t stop at selling models; they started building their own applications. The example of Bitcoin mining hardware companies made this dynamic easy to understand. A company that manufactured mining rigs received massive pre-orders—then, when Bitcoin’s price surged dramatically, it realized it could earn far more by using the equipment itself rather than shipping it to customers. So instead of fulfilling orders, it diverted the hardware for its own mining operations. AI model companies face similar choices. For instance, if a powerful model can discover new pharmaceuticals, it’s far more profitable to conduct the research internally and own the results than to license the model to others. What made this explanation compelling was that it did not frame power concentration merely as greed by a few corporations. It’s natural for companies with infrastructure and models to expand upstream and downstream to capture greater value. When each company repeatedly makes decisions that benefit itself, energy, semiconductors, models, scientific outcomes, and capital can all converge in the same hands. Access restrictions may also be justified on grounds of safety. But the presentation clarified that even beyond safety, there’s a strong incentive to internalize the most valuable outcomes. Ultimately, the concentration of AGI is not just a technical issue—it depends on what choices are incentivized for those who control the technology. The presentation then outlined three prerequisites for pursuing the alternative path: Open infrastructure, open science, and open capital formation. At first, these terms sounded familiar—but when viewed in light of the preceding explanation, their meaning shifted. Simply open-sourcing AI models is not enough. You need computational resources to run them and funding to continue research. Even if brilliant scientists and developers exist worldwide, without mechanisms to trust one another or pool capital, solving large-scale problems together remains impossible.Therefore, we did not treat infrastructure, science, and capital as separate entities. The explanation was that everyone must have access to computational resources, scientists and AI agent developers must be able to collaborate together, and the necessary capital must be mobilizable from around the world. This also helped me understand what EigenCloud means by coordination. It’s not simply about bringing many people together in one place. It’s about enabling strangers to collectively pool capital, divide work, verify outcomes, and hold each other accountable when commitments are broken. Currently, corporations or institutions largely fulfill these roles. While convenient to manage, decision-making power becomes concentrated in those entities. EigenCloud is seeking ways to enable collaboration without centralizing authority in any single organization, using blockchain and public verification. It did not claim to solve the large-scale problem of power concentration after AGI all at once. Instead, it explained that the focus is on building open infrastructure, testing networks for scientific research, and connecting scientists, agent developers, and capital. I appreciated this aspect too. Because it didn’t present itself as having all the answers while proclaiming grand goals. The system being built is clearly not a finished solution—but it is unmistakably an effort to leave open an alternative to the current trajectory where all power accumulates in the hands of a few. After watching this presentation, I felt it would be a mistake to view EigenCloud merely as a verification-based cloud platform. The problems EigenCloud addresses do not end with verifying whether computations were performed correctly. They extend to changing who can participate in research, who can raise capital, and who owns the results in the age of AGI. It remains uncertain whether AGI will greatly amplify the capabilities of a few—or empower many more individuals to accomplish far more than before. The clearest takeaway from this video was that the outcome will not be decided after AGI is completed. The infrastructure being built today, the flow of capital, and the methods of collaboration are already shaping that direction. Thus, the presentation felt less like an introduction to what EigenCloud is building, and more like an explanation of why such a system must be created now. #AGI #EigenCloud