Eigenstate

Jason Shen / Quant Developer and AI Systems Builder

Jason Shen

I build systems where state, memory, and feedback cannot be hand-waved.

Thesis

Trading systems trained the discipline: truth, drift, reconciliation, auditability. I now apply that discipline to AI memory and continuity.

Systems

Systems I Build

Case files, not project cards.

FILE SLOT

Case SYSTEM 01

proof artifact v1

Trading Infrastructure

Execution, reconciliation, monitoring, and provider-confirmed truth under live trading pressure.

OPEN CASE FILE →

FILE SLOT

Case SYSTEM 02

exploration shell

AgentOS

Filesystem memory, context construction, scheduler-driven proactivity, and human review.

OPEN CASE FILE →

FILE SLOT

Case SYSTEM 03

case-file shell

AI Workflow Runtime

Composable multi-step agent work with explicit handoffs, tools, verification, and artifacts.

OPEN CASE FILE →

System Map

Source, core, validation.

Content Surfaces

Notes, essays, lab.

Ansatz

Field notes for hypotheses, constraints, and system observations.

Eigenvalues

Durable essays on state, memory, systems, and judgment.

Impuls

Small experiments and benches when a system needs to be tested.

Operating Principles

Three rails.

  1. P01

    Truth

    Provider-confirmed truth beats internal state.

  2. P02

    Memory

    Memory should be inspectable before it becomes persistent.

  3. P03

    Boundary

    Frontstage experience and backstage infrastructure should stay distinct.

Collaboration Entry

Start with context.

This surface stays open where architectural choices are explicit:

  • agentic systems and execution architecture
  • decision systems for risk, review, and fallback behavior
  • character-system exploration in its current shape
Open the contact surface