When an Event Starts an Agent: Delivery, Delay and Ordering Contracts
A transactional event ledger separates delivery from valid work, testing duplicates, object versions, expiry, supersession and overlap before an execution attempt.
THE READING ROOM
Writing & investigations.
Measuring the AI Review Queue in a Reference Repository
Faster implementation can increase the review queue without increasing accepted throughput.
Modernize the Module Without Accidentally Rewriting Its Clients
A real HTTP contract distinguishes a compatible module rewrite from a plausible200response that changes nulls, rounding, status and dates.
Blazor Interfaces for Slow, Fallible AI
An AI interface can be technically responsive and still be dishonest. It can animate a progress bar while nothing measurable is progressing. It can display “cancelled” while a server continues working. It can replace a current answer with a late result from a previous request.
One Row Returned, a Large Intermediate Workload
Two SQL queries return the same one-row aggregate. Actual plans, logical reads and a bounded experiment reveal very different source work.
Treat Copilot Studio as a Dependency: A Local Direct Line Contract Harness
A real Direct Line client meets a local HTTP fixture to separate replay, partial results, errors and cancellation from business completion.
The Human in the Loop Needs a Job Description
Human oversight works only when a person has a real decision to make, sufficient context and the authority to change the outcome.
An ASP.NET Request That Works Locally and Fails After Deployment
Real TLS, trusted forwarded headers, PathBase and a missing artifact produce different deployment failures; a bounded Nginx/ASP.NET fixture separates them with actual HTTP evidence.
One Telemetry Contract Across a .NET AI Platform
A shared dashboard can make a platform look unified long before its measurements are comparable. One service records request duration in seconds. Another records milliseconds. A third starts the timer after retrieval, while the first includes it.
The Retry That Erased the Failure: Attempt History and Log Retention
A real SQLite fixture keeps failed attempts while the current operation turns successful. Identity conflicts, repeated observations and a local transaction show why a summary is not an incident history.
Does Agent Oversight Actually Stop the Effect? Mutation Testing the Control Boundary
Six deliberately broken executor variants show whether independent effect assertions detect missing actor, scope, replay, expiry, digest and version controls.
AI Readiness Is a Set of Evidence Gaps
Readiness is better expressed as unresolved questions with owners than as a decorative maturity score.