About Simone Systems Research
An independent research organization investigating the engineering, evaluation, and compute economics of autonomous AI systems.
Background & Purpose
As artificial intelligence transitions from conversational models to autonomous multi-step systems, the primary engineering bottlenecks shift from parameter scale to systems-level reliability: coordination topologies, verification bounds, resource allocation, and failure recovery.
Simone Systems Research conducts independent technical investigations aimed at developing reproducible methods and open software artifacts to address these challenges.
Leadership & Researcher
Jonathan Simone conducts independent research into AI agent orchestration, evaluation, verification, and compute efficiency. His work focuses on developing systems that distinguish plausible progress from independently verified improvement.
Research Program Areas
- Agent Orchestration: Deterministic state-machine coordination, sandboxed tool execution, and communication protocols across heterogeneous reasoning models.
- AI Evaluation & Verification: Testbench designs that isolate genuine algorithmic capability gains from benchmark leakage, metric hacking, and stochastic noise.
- Compute Economics: Quantitative trade-offs between inference cost, context compression, test-time compute, and verified task accuracy.
- Adaptive Systems: Closed-loop introspection, memory state tracking, and measured policy evolution in dynamic environments.
For reviewers
Simone Systems Research is independent and founder-led. Ten questions a reviewer asks first, each with a one-line answer and a link.
- Who is conducting the research?
- Jonathan Simone. The organization is independent and founder-led, with no institutional affiliation. GitHub →
- What questions?
- Agent evaluation, verification architecture, human control of AI coding agents, and compute efficiency. Research focus →
- What has been built?
- Five public repositories: SEED, The Council, BigBoss, The Bus, Godot AI Methodology. Projects →
- What has been measured?
- One author-run case study: 1,979 PR-linked commits on one repository, n = 1. Research note →PAPER.md →
- What remains hypothetical?
- C-004: that a machine-checked control plane improves outcomes over a bare repository. Registered, untested. seed-protocol →
- Can I reproduce any of it?
- The instrument checks run offline (seed-protocol reproduce commands, The Council's fixture demo). The census itself cannot be regenerated because its source repository is private. No independent reproduction has been filed yet. SEED reproduce →Council README →File a reproduction →
- Is the code public?
- Yes. Every repository is on GitHub under the MIT license. github.com/thisisntjon →
- Are failures retained?
- Yes. The Bus is a preserved negative result. seed-protocol keeps a RETRACTIONS ledger of withdrawn numbers. thebus →RETRACTIONS.md →
- Is there an identifiable methodology?
- Yes. Pre-registration, receipts, sabotage tests, and author-versus-verifier separation. Principles →
- How do I contact?
- Email. jon@simoneresearch.com →
Inquiries & Contact
For research inquiries, preprint discussions, or technical collaboration: