About
Jacob designs AI infrastructure for enterprises that need institutional memory. He is currently building Project Symmetry, a multi-agent system with a five-layer episodic memory architecture grounded in cognitive science.
What We Build
Project Symmetry is a private AI infrastructure platform that functions as a digital firm. A coordinated fleet of specialized agents handles research, analysis, strategy, and execution, working together on complex problems across sessions with full institutional memory of everything the firm has ever learned. The platform is designed so that any agent type can be built and deployed for any domain. The architecture is operational; agent fleet expansion and client-facing deployment are on the roadmap.
The platform's core differentiator is a five-stage memory pipeline grounded in cognitive science. Each experience moves through a deliberate encoding and consolidation process before becoming part of the agent's permanent knowledge base.
Engrams are stored in a temporal knowledge graph: a web of typed connections rather than a flat list. When an agent begins a task, it traverses the graph from the current context outward, surfacing only the memories with direct relational bearing on this conversation, this person, this domain. As the archive grows, token cost stays bounded. The agent doesn't read its memory. It navigates it.
Specialized agents for trading, research, content, legal analysis, orchestration, and risk management. Each operates within a defined scope, contributes to shared memory, and coordinates through a custom communication protocol. The platform is extensible by design: any agent type can be built and deployed for any domain.
Each agent operates through a structured decision hierarchy that sequences signals, applies circuit breakers, and routes by regime. This prevents cascade failures from conflicting indicators and input overload. No decision leaves the system uncontested: dedicated agents serve as quality gates, contrarian reviewers, compliance checkers, and specialty researchers. Variant strategies run in shadow evaluation against live decisions; parameters that clear a performance threshold are promoted automatically, and rollback is one command. The result is a multi-agent system with the epistemic hygiene of a structured analytical process.
The trading system runs Bayesian parameter optimization, hybrid decision architecture, and a two-tier risk governance framework against 25 years of historical market data. Fully autonomous, daily operations.
A content generation agent is in parallel development, built around the most complex implementation of the memory architecture to date. The intent: a system that develops an authentic, evolving voice, informed by accumulated craft memory across every session. Currently in alpha.
Methodology
Most AI systems are built by engineers who have spent their careers inside the discipline. Project Symmetry was built by someone who spent his career outside it: through intelligence operations, organizational leadership, quantitative finance, and product delivery, applying the same analytical framework to a new domain.
The cognitive science grounding (Tulving, Tonegawa, Müller and Pilzecker) was validated after the architecture existed, not before. The goal was a system that remembered the way organizations actually need to remember. The research confirmed the design.
The core discipline is applied epistemology: how information is collected, validated, weighted, and stored under conditions where error has consequences. Those principles transfer directly to the problem of what AI agents should know, how confidently they should know it, and how much to trust what they produce.
Contact
J. I. Ashley Consulting is not currently accepting new clients. For partnership inquiries, press, or investment conversations: