A2035 Ventures

Frontier-AI and national-security technical leaders strengthen Project 777 as it advances a compounding intelligence layer demonstrating bounded recursive self-improvement and independent runtime validation. DALLAS , Sept. 3, 2026 /PRNewswire/ -- A2035 Ventures has expanded the technical leadership around Project 777 as the company advances strategic discussions with major AI organizations. S. S. Air Force, serves as a Technical Advisor to Project 777. Sean Perry , formerly with Agentic AI at Meta Superintelligence Labs, has joined as Strategic Technical Lead .

Dunlap has advised Project 777 since April 2025, providing technical perspective as the architecture and evidence base have developed. Perry reviewed Project 777's Technical Proof Packet after leaving Meta's superintelligence organization and is now helping lead technical diligence and strategic technical engagement. "Project 777 has now demonstrated three core breakthroughs: runtime intelligence and safety, bounded recursive self-improvement, and cross-domain operation through the same core architecture. Sean and Preston bring extraordinary perspective from frontier AI and national-security systems as we strengthen the evidence and scale what already works," said TK Stohlman, Founder & Architect of Project 777.

Project 777 is a compounding intelligence layer above frontier models and agents . Its three demonstrated breakthroughs are runtime intelligence and safety , which independently governs what may advance; bounded recursive self-improvement , which allows validated improvements to persist and compound; and cross-domain architecture , which allows the same core system to operate across materially different sectors without redesign. As agents work, built-in layers continuously check their reasoning, evidence, and safety, and can stop or redirect work before it advances.

The intelligence producing the work cannot simply approve itself. This is what allows intelligence to improve and compound while the system remains under control. Google DeepMind's recent From AGI to ASI paper identifies recursive self-improvement, new architectures, and multi-agent collective intelligence as credible pathways toward artificial superintelligence. Over the past two years, Project 777 has independently built and operated a governed architecture that brings several of those dynamics together: bounded recursive self-improvement, a model-agnostic multi-agent intelligence layer, and independent runtime control.

The system carries validated intelligence, evaluation history, accepted improvements, rejected changes, and accumulated operating experience forward even as the underlying frontier models change. A2035 Ventures believes this combination represents early operating signs of safe superintelligence. Project 777 also has early commercial validation.