Writing.
Research writing spans formal methods and runtime verification. Technical notes and field reports examine the construction of provably safe AI.
Entries
- Updates
Current Technical Program
Multiple stacks already compile policies into monitors and mediate effects. They also emit verifiable evidence. The next mile is scale and standards.
11 min - Participation
Contributing to Verification Infrastructure
Technical participation guide for public work across the SentinelOps verification stack.
7 min - Infrastructure
Building Verification Infrastructure
We are past intent. Multiple stacks already compile policies into monitors and mediate effects. They also emit verifiable evidence.
10 min - Roadmap
Verification Infrastructure Development Sequence
Safety emerges through a repeated cadence. Specification leads to proof and execution. Demonstration then supports standardisation across domains.
8 min - Technical
Mapping the Verification Space
The safety conversation is noisy. Evidence scales through properties established by proof and enforced online.
15 min - Vision
The Verifiable AI Ecosystem
Three years from now, the "AI internet" feels boringly safe. Users benefit from a fabric in which every agent action is mediated through verified controls.
12 min
Papers
- §1Architecture · 19 pages
System Architecture
Comprehensive architectural design and formal specifications for verifiable AI systems.
- §2Technical · 14 pages
Technical Guide
Detailed technical documentation covering implementation methodologies and runtime verification techniques.
- §3Whitepaper · 37 pages
Whitepaper
Strategic roadmap for trusted AI infrastructure, grounded in the programme methodology and vision.
Receive new writing.
Occasional research dispatches for technical readers.