Founder, Fronesis Labs — building deterministic audit infrastructure for agentic AI systems
(DCL · cryptographic verification · AI compliance)
AI Systems Architect · Siberia, Russia
Patent applicant (6 patent filings with Rospatent) · Independent researcher
Most agentic systems trust their agents on the output alone.
I build the layer that checks whether the work was actually done.
Core thesis: An AI agent doesn't execute a task — it generates text that looks like execution. Without a deterministic verification layer, there is no difference between a completed task and a fabricated one.
This is not a model alignment problem. It is an architecture problem.
Deterministic Commitment Layer (DCL) addresses it by introducing three primitives:
- Commitment Protocol — cryptographic binding to inputs and declared tool use before execution begins
- Execution Trace Verification — independent signing of every tool call, decoupled from the agent's self-report
- Cryptographic Audit Log — an immutable chain linking commitment → trace → result, replayable and tamper-evident
If you can't replay the execution, you can't audit it.
If you can't audit it, you can't deploy it in a regulated environment.
Working implementation of the DCL evaluation pipeline:
- Jupyter notebooks with end-to-end verification examples
- FastAPI endpoints for commitment and trace verification
- Docker containerization + CI/CD
- Python · FastAPI · Docker · GitHub Actions
Six patent applications filed through Rospatent covering the DCL technology stack.
IP terms negotiated: core DCL architecture remains author's property under license.
Representative systems:
- DCL core — deterministic commitment and trace verification protocol
- Audit chain primitives — cryptographic log construction for multi-agent pipelines
- Compliance adapter layer — mapping DCL outputs to FSTEK and regulated-industry requirements
-
Deterministic verification layer vs. fabricated execution in agentic systems
-
CVE-HUM trilogy (in progress) — applying cybersecurity frameworks to analyze civilizational vulnerabilities. First volume: No Threats Detected. Submitted to Zer0 Books, MIT Press.
- Pilot deployments — regulated industries where agent auditability is a hard requirement (finance, legal, government)
- Standards discussion — researchers and architects working on deterministic governance or formal verification for AI
- Whitepaper feedback — DCL full technical specification in preparation, early access on request
- 🐱 GitHub: DariRinch
- 📝 Habr: @DariRinch
- 💼 Fronesis Labs: github.com/Fronesis-Labs
Verification is not a feature.
It is a precondition for deployment.
— Dari Rinch