CAPITAL VARIATION COMES TO SILICON VALLEY

Silicon Valley isn’t just a place. It’s the engine. Capital Variation is going West Coast, hard, because that’s where AI is being built.

It’s where deep tech, robotics, chips and rare-mineral supply chains converge, and where the next 100x companies are founded, funded and scaled. We’re embedding ourselves at the source:

→ Direct access to frontier AI labs

→ First-look deal flow in physical AI, semis and robotics

→ Relationships with the builders, buyers and capital shaping the future

If you want to invest in what’s next, you need to be where it’s happening. Silicon Valley, here we come!

Featuring Dr. Ben Goertzel, Founder & CEO, BGI Labs

Dr. Ben Goertzel is a foundational figure in the field. He launched the term and concept of artificial general intelligence (AGI) in 2002 and is often called the father of AGI. He is the architect of OpenCog and its successor, Hyperon, has published 260+ works including 25+ books, and as Chief Scientist at Hanson Robotics helped create Sophia, Desdemona and Grace. Today he is the Founder and CEO of BGI Labs, building the operating system for general intelligence.

His core vision centers on moving beyond large language models, which he argues cannot reach true human-level intelligence on their own, however large they get. Today’s models are brilliant but forgetful: they lose what they learned when a session ends, their reasoning can’t be audited, and they have no safe way to improve themselves. Rather than betting on scale, BGI Labs is betting on architecture, building the rest of the mind around the model so that language models become components inside a larger, more general intelligence.

🏢 BGI Labs: The Operating System for General Intelligence

The next computing platform won’t be another model. BGI Labs is a for-profit AGI company building what comes after it: an open, auditable and decentralized cognitive layer that sits above today’s rapidly commoditizing models, funded by a real enterprise business built on the same platform.

The timing is right. Hyperon is finally scalable, the models it orchestrates have become abundant and cheap, and the agent economy now needs a cognitive operating layer that no one else is building. Behind BGI Labs stands a research bench of 20+ senior engineers with 750+ publications and 9,000+ citations, most of them a decade or two inside OpenCog and its predecessors.

Learn more at bgilabs.ai

⚡ Omega: A Mind, Not A Model

Omega is BGI Labs’ first Hyperon agent, and it is already running in live demonstration. It is a persistent AI agent with long-term and working memory, explicit reasoning, planning and tool use, built to act as a control tower for whole fleets of AI agents. Omega hives organize those agents into effective teams with clear ownership, defined permissions and a complete record of who did what.

Unlike a chatbot, Omega works in one continuous cognitive loop. It remembers knowledge, goals and decisions across months of work, coordinates across models, reasoning engines, tools and human review, governs every consequential action with permissions and approvals, and improves in measured, modular steps without replacing a foundation model.

The early results speak for themselves:

• Uninterrupted public deployment: more than three months running in public across 10,000+ messages with 24 human participants, including passing an adversarial silence test.
• Long-horizon memory: 57 of 60 correct (95%) on a LongMemEval subset, above the LLM baseline.
• Interactive reasoning: 5 of 7 ARC-AGI-3 levels solved with zero game-specific engineering, with the agent building its own analysis tools.
• Live multimodal monitoring: 80% correct event identification over a two-hour session.

Omega is also the foundation for the BGI Labs commercial platform, planned for 2027, and for a wider product suite now in development, including Omega Purple, an agentic security office, and Omega Linux and Omega Android.

Benchmarks are self-run on public protocols, with the evaluation harness and third-party replication to be published.

🧠 OpenCog Hyperon: The Architecture Behind It All

Hyperon is twenty years of architecture whose substrate has only just arrived. OpenCog specified persistent memory, explicit reasoning and gated self-modification from the beginning; what was missing was the infrastructure to run it. Distributed AtomSpace and MORK made the memory scalable, MeTTa became a real, compiled language, and open-weight models made language and perception abundant and cheap. As of early 2026, Hyperon is scalable to production workloads for the first time.

• Cognitive Synergy: neural, symbolic and evolutionary methods work together and learn together within one unified system.
• The AtomSpace Metagraph: a durable, inspectable metagraph that holds the system’s knowledge, goals and decisions.
• MeTTa Language: the language Hyperon’s reasoning is written in, with programs you can test and edit at runtime.
• PLN: probabilistic reasoning that draws inferences on explicit truth values, so every conclusion can be traced.
• Bounded Self-Modification: proposed changes run under tests, sandboxing and alignment gates before they take effect.

Safety is a property of the architecture, not a policy layered on top. Every belief, goal and decision is inspectable, and permissions and human approval sit between intention and consequence. While the scaling labs try to make opaque systems legible, BGI Labs started legible.

🤖 Featuring Chris Kudla, Co-founder & CEO, Mind Children

Most humanoid robots are being built for factory work. Mind Children is building them for people. Co-founded in 2023 by Chris Kudla and Dr. Ben Goertzel, the Seattle company designs affordable, repairable humanoid robots for the settings where human connection matters most: healthcare, education and hospitality.

Its first robot, Codey, stands about three feet tall, with an expressive face and a wheeled base that keeps him upright around people. Codey can hold a conversation, check in on someone’s wellbeing, and call a person when something seems off. That makes him useful for older adults aging at home, students who need extra one-on-one practice, and hotel guests arriving late to an understaffed front desk. He’s modular, easy to service, and designed from day one for low-cost mass manufacturing.

Chris brings two decades of building physical products, from manufacturing at Toyota and chassis engineering at Honda to aerospace and consumer hardware. Dr. Ben Goertzel, Mind Children’s Chief AI Scientist, leads the software side, and previously led the AI behind Sophia at Hanson Robotics.

Codey is in prototyping today, with pilots in care and education settings planned for 2027. Mind Children has been featured in USA Today, the Los Angeles Times, NewsNation and GeekWire, and has an open investment round on Wefunder: wefunder.com/mindchildrenrobotics

Learn more at mindchildren.com

See Codey live. Mind Children is bringing Codey to Stanford for a live demo.

October 16, 2026. Don’t miss it!