Research & perspectives
Research I lead, and what I am learning from building production systems.
My research work sits inside Thynkr Systems and is grounded in the products we build and operate. This page will also hold technical writing as it is published.
§01Current research direction
Research & development
THYNKR RSI — Governed Recursive Self-Improvement
Exploring how AI systems can observe their performance, identify improvements and evaluate better approaches through controlled experimentation and accountable human governance.
I lead the initiative’s architectural direction. It builds on two foundations that already exist at Thynkr Systems, Brain and WorQ. The improvement engine itself is planned research, not a deployed capability.
Research questions
- How should an AI system establish a trustworthy baseline before it recommends change?
- Which evidence makes an improvement proposal worth testing?
- How do controlled experiments stay reproducible across models and workflows?
- Where must human approval remain, however capable the system becomes?
§02Perspectives
Writing in preparation.
No articles, papers or talks are published yet. The topics below are planned; when a piece is published it will be listed here with its date.
- 01Governed recursive self-improvementPlanned
- 02Multi-model orchestrationPlanned
- 03AI agent assurancePlanned
- 04Practical AI evaluationPlanned
- 05Software architecturePlanned
- 06Autonomous systems and human oversightPlanned
- 07Production engineering lessonsPlanned
Research or engineering collaboration.
I welcome conversations with engineering teams and researchers working on model orchestration, measurable AI evaluation, agent governance and controlled experimentation.