Technical diligence
Before an investment, acquisition or partnership: separate what a product demonstrably does from what its materials claim, with sources cited and gaps stated.
Independent research, software, and technical advisory
Connected intelligence.
Inspectable evidence.
Assertion Labs builds and evaluates software for questions where the evidence matters, from AI systems and security architecture to technical diligence.
Source / claim / evidence
Every conclusion traceable to what supports it.
Every conclusion traceable to what supports it.
Introducing
CliqueMap An Assertion Labs product
In developmentConsequential technical questions are answered from scattered material: design documents, configuration, vendor statements and earlier reviews. CliqueMap is being built to connect that source material, the claims it makes and the relationships between systems, so that you can inspect what supports a conclusion and what remains uncertain.
It is in development. We are open to a conversation about a scoped pilot for a technical diligence or architecture investigation.
Illustrative example · synthetic data
Dataset v1.0.0
Can the support assistant read customer billing records?
Evidence for C1 · asserted in S3 (Vendor security FAQ)
-
contradicts
assistant-svc: allow billing:ReadInvoiceS2 · Service role policy, rev 14 -
supports
The assistant reads ticket history only.S1 · Architecture overview, v3 · 2 Mar
Entities: Support assistant Billing API
Limitations. Policy revision 14 may differ from what is deployed. Deployed role bindings were not inspected.
Where evidence matters
Questions we examine.
AI and security architecture
How an AI-enabled system reaches data, tools and other services, which boundaries hold, and which security claims the available evidence actually supports.
Research investigations
Bounded questions in program analysis, compilers, formal methods and software security, with reproducible evidence and documented assumptions.
Working together
Start with a bounded question.
Initial engagement
Technical Risk & Evidence Diagnostic
What can we substantiate about this system, where are the important unknowns, and what should we examine next?
Continuing engagements: briefings, diligence and research work packages, each scoped to the question at hand.
01 / Advisory
02 / Advisory
03 / Advisory
04 / Advisory
AI systems & security
Examine how AI applications access data, exercise authority and interact with other systems. Evaluate the architecture and the evidence behind its security claims.
Executive & board briefings
Clear technical context for consequential decisions. Focused preparation, a concise written brief, and a discussion of the questions management should be able to answer.
Independent technical diligence
Examine a product’s architecture, security claims and technical assumptions before a consequential investment, acquisition or product decision.
Research & assurance
Bounded investigations, analysis prototypes and independent evaluation in program analysis, compilers, formal methods and AI-enabled software systems.
Research & engineering
Ideas grounded in the work.
Explore publications in software analysis and verification, alongside independent software projects that put technical ideas into practice.
Explore the research record →Browse company projects →Insights
Questions worth examining.
Notes on software, evidence and technical judgment. New writing appears here after review; the public research record is available now.
Visit insights →The principal
Independent by design.
Assertion Labs is Ryan Williams’ company for independent research, software products and technical advisory work. Publications and earlier experience are his; company work is described separately.
About Ryan →A conversation about the work.
A short description of the problem, its timing and intended audience is enough. Engagements are subject to conflict and availability review.
Discuss a project →