Our CEO, Dr. @bengoertzel, delivered a keynote at the Ai4 conference in Las Vegas, examining what it will take to reach AGI beyond scaling large language models. He identified reasoning, persistent memory, continual learning, creativity, and grounding in the world as central capabilities for systems that can generalize beyond their training. “The goal is to make real thinking machines,” he said. Dr. Goertzel described an AGI that can enter a novel environment, work out what is happening, ask questions when necessary, and learn from what it encounters. Such a system needs to form abstractions, generalize from experience, retain what it has learned, and connect knowledge across time. Memory was a major focus. “Lifelong learning depends on lifelong memory,” Dr. Goertzel said. He discussed OmegaClaw, which combines a continuously running agentic loop with symbolic working memory and long-term declarative and episodic memory, allowing an agent to retain experience over extended periods and build knowledge through continued interaction. The keynote also addressed what it means for such a system to be beneficial. “Alignment is ill-posed unless you ask: aligned with what?” Dr. Goertzel noted that human values are dynamic and heterogeneous, making alignment a question that extends beyond instruction following and behavioral constraints. The system's design must account for values, self-understanding, and moral agency. Dr. Goertzel also discussed the architecture being developed through Hyperon and OmegaClaw. Hyperon combines neural, symbolic, and evolutionary paradigms, while OmegaClaw brings persistent symbolic memory into a continuously running agentic system. He described these approaches as part of a broader effort to develop AGI beyond a single-model paradigm. He also addressed the infrastructure underlying AGI development, emphasizing open-source and decentralized approaches that allow a broader range of people and organizations to participate in developing and governing advanced AI systems.

Our CEO, Dr. @bengoertzel, joined @juliankissa on The Beyond Tomorrow Podcast to discuss the neural-symbolic path to human-level AGI, the geopolitics of AI, the transition to superintelligence, and how open-source development, decentralized infrastructure, and governance can contribute to achieving a beneficial Singularity.

We introduce OpenWater, a fully open, decentralized approach to addressing deepfakes and related issues through digital provenance. The proposal addresses a fundamental limitation of AI-based content detection. Detectors can identify patterns associated with known generators, but they do not establish an artifact’s history. OpenWater records signed claims about how digital content was created, modified, and published. Interoperable credentials and durable watermarks preserve the link between an artifact and its provenance records, even when conventional metadata is removed. OpenWater also keeps the provenance infrastructure distributed across independent services rather than placing it under a single authority. Individuals, institutions, and AI systems can apply their own standards when evaluating the available evidence. The same framework can extend to training data, model outputs, and agent actions, giving AI systems access to provenance information about the data and processes they encounter.

How does the scale of current AI infrastructure investment affect the economics of the AI industry, and what happens if AGI arrives on a different timeline than expected? In a new analysis, our CEO, Dr. @bengoertzel, examines three possible trajectories for AI progress and the financial obligations created by today’s infrastructure buildout. His central argument is that a financial crisis could emerge across fast, moderate, and pessimistic AI scenarios, meaning that a future market correction would not necessarily tell us much about whether the underlying technology is progressing toward AGI. The analysis then considers what changes if decentralized AGI succeeds. A global network using heterogeneous hardware and token-based funding would have a different capital structure from the debt-financed infrastructure being built by major technology companies, with no comparable refinancing gates and a capacity to absorb hardware released during a downturn.

Our CEO, Dr. @bengoertzel, sat down with Dan Turchin on the AI and the Future of Work podcast to discuss the path to Beneficial General Intelligence. The conversation examines what AGI requires beyond LLM-centric systems, including reasoning grounded in persistent memory, architectures capable of examining themselves, self-modeling and other-modeling, and values that can be expressed symbolically. Dr. Goertzel explains how Hyperon-based systems enable these capabilities through the integration of neural, symbolic, and evolutionary paradigms.