Reliable verification systems
Test architecture, domain models, property-based and metamorphic testing, observability, and failure analysis.
MetronForge
MetronForge develops original work in applied mathematics and reliable software verification, from working hypotheses and experiments to projects and formal publications.
An independent notebook by Viktor Mikhalkin. Notes also examine papers, books, and software by other authors through clearly attributed commentary, reviews, and thematic digests.
Subscribe via RSSNew and updated notes, without an account.
Current directions
Solvers and certificates
Structural methods for linear and nonlinear systems, executable numerical certificates, and adversarial experiments.
Reconstruction and transforms
Geometric integration, Radon-type transforms, sparse reconstruction, and ideas developed through calculation and falsification.
Featured Notes
Agentic End-to-End Testing Still Needs an Independent Oracle
Updated:CommentaryDiscusses: Agentic Testing: Where Agents Fit in the E2E Testing Stack + 1 more
Slack's agent-driven E2E experiments expose a useful new execution model, but adaptation must remain separate from the authority to declare success.
Architecture Is the Set of Decisions That Constrain Future Change
Updated:CommentaryDiscusses: Software Architecture: It Might Not Be What You Think It Is
Pierre Pureur and Kurt Bittner correctly place decisions at the centre of architecture; their importance follows from reversal cost and blast radius.
Coding Agents Work Best When Verification Is Cheaper Than Reimplementation
Updated:CommentaryDiscusses: Old and new apps, via modern coding agents
Terence Tao's restored mathematical applets show where coding agents have real leverage—tasks whose outputs remain cheap for an expert to verify.
A Test Strategy Is Risk Allocation, Not a Geometric Shape
Updated:CommentaryDiscusses: Software Testing Strategies: The Complete Guide + 3 more
Pyramids, trophies, and honeycombs summarize different testing economics; a defensible strategy begins with risks, evidence, and feedback constraints.
Property-Based Testing Starts with the Property, Not the Generator
Updated:CommentaryDiscusses: Beyond Accidental Quality: Finding Hidden Bugs with Generative Testing + 5 more
Property-based testing is more than randomized input generation; its value depends on the property, distribution, oracle, shrinking, and evidence.
Chain-of-Thought Prompting Can Improve Answers; It Is Not Proof
Updated:CommentaryDiscusses: How does Chain-of-Thought (CoT) Prompting work? + 5 more
Chain-of-thought prompting can improve some multi-step answers, but a plausible rationale is not evidence that an answer is correct or faithful.
Agentic Resource Discovery Still Needs a Verification Boundary
Updated:CommentaryDiscusses: Announcing the Agentic Resource Discovery specification + 1 more
ARD can help agents find capabilities across organizations, but discovery metadata must remain outside the trusted execution boundary.
Playwright Is Mature; Migration Still Needs an Evidence Case
Updated:CommentaryDiscusses: Playwright finally hit critical mass as we enter 2026 + 2 more
Playwright's mature tooling makes it a strong default for new browser automation, but migration should be decided through a representative experiment.
Introducing MetronForge: An Independent Research Notebook
Updated:AnnouncementMetronForge is an independent research notebook for engineering observations, computational experiments, projects, and formal publications.
Recent Notes
Documentation as Code Still Needs Tests for Documentation
Updated:CommentaryDiscusses: Write documentation like you develop code
Version control and CI improve documentation workflow, but a successful build cannot establish that instructions remain true, complete, or usable.
Selecting Fewer MCP Tools Is Not the Same as Enforcing Least Privilege
Updated:CommentaryDiscusses: The GitHub MCP Server adds support for tool-specific configuration, and more + 1 more
GitHub MCP tool filtering reduces context and accidental capability exposure, but real least privilege also requires identity, authorization, and effect controls.
Token Usage Is an Input Cost, Not an Engineering Outcome
Updated:CommentaryDiscusses: From story points to tokenmaxxing: Why engineering keeps measuring the wrong things
Counting tokens repeats an old measurement mistake unless AI-assisted work is evaluated through delivery, quality, and user outcomes.
Paper Diagrams Are Maps, Not Proof Certificates
Updated:CommentaryDiscusses: A paper diagram visualizer
Dependency diagrams can make mathematical papers easier to navigate, provided their inferred structure is not mistaken for verified proof evidence.
Choose Agent Infrastructure by the First Constraint That Can Break It
Updated:CommentaryDiscusses: The Open Source Agent Toolkit in 2026
Agent stacks should be selected layer by layer, with explicit contracts for state, tools, memory, evaluation, and recovery.