An Evidence-Aware Agent Framework Upgrade Tracker in .NET
A public NuGet evidence tracker compares three actual function-invocation contracts across two pinned packages while leaving general compatibility unknown.
THE AUTHOR
Marcin Borecki
Developer, entrepreneur and business angel. Writing about AI, software, data and the evidence behind technical decisions.
About the author →THE READING ROOM
Writing & investigations.
Following Async Work Without Confusing It With the Thread Stack
A real .NET 11 trace reveals logical async chains, a short-process buffer trap and the limits of method mapping and observation-cost measurements.
The Allocation That Disappears: Escape Analysis from .NET 10 to .NET 11
One compiled C# assembly, two pinned runtimes, and nine allocation scenarios. Real IL, JIT listings and measurements show when an object disappears, when nullable boxing moves off the managed heap, and why a local delegate can still allocate.
Runtime Async in .NET 11: Who Builds the State Machine Now?
Runtime async changes where suspension is represented. A real two-library .NET 11 lab shows smaller compiler artifacts and lower allocation on some paths, while a controlled suspension case resists a universal speedup claim.
The Evidence Pack Behind an AI Investment Thesis
An engineering evidence pack can clarify uncertainty without pretending to decide the financial merits of an investment.
From Demo to Product Decision: A Lab for Accepted Work Orders
A real ASP.NET and Microsoft.Extensions.AI fixture compares two proposal paths through one validator, preserves every failed revision, and makes explicit review the boundary of accepted work.
What Becomes Scarce When Software Gets Cheaper to Build?
Cheaper software creation can increase the value of choosing well, earning attention and accepting responsibility for maintained outcomes.
A Minimal Azure Agent Platform with Testable Operating Boundaries
A small local reference makes admission, retry, journal state and shutdown observable; a compiled Azure candidate preserves the distinction between mapping and deployment.
A Read-Only SQL Diagnostic Agent Needs an Enforced Boundary
A diagnostic agent’s read-only claim is tested through actual SQL permission denials, fixed query templates and evidence-based hypothesis rejection.
Tool Calls You Can Audit: Deterministic Controls Around Agents
An agent can explain why it wants to perform an action without establishing that it may perform it. It can produce a well-formed tool call without establishing that the arguments are safe. It can report success without proving that the external effect occurred exactly once.
Follow the Agent Tool to the Rows the Database Actually Reads
A real framework tool reaches SQL Server, where two Index Seek plans return the same answer while reading different amounts of data.
Executable Deployment Recipes with an Honest Verification Date
A versioned local deployment recipe records its prerequisites, owned resources, observed HTTP results and cleanup, making its verification date a claim about an actual run.