Practical insights
on software and AI

Azure OpenAI retires model versions on dates customers cannot extend. Depending on a Standard deployment's upgrade setting, the feature moves to another model or stops answering, and a saved evaluation set keyed to the model version makes that a planned test run.

Read article

Projects that target .NET 10 audit transitive packages by default. With TreatWarningsAsErrors, an advisory published after the upgrade can fail restore on unchanged code, so the audit policy has to be settled before the retarget merges.

Read article

Inbound documents can carry instructions aimed at the AI that reads them. Give the reading step no authority to act, have application code check every field it returns, and treat those values as untrusted even after they pass.

Read article

Four AI governance controls a reviewer can inspect in a .NET application: what OpenTelemetry records by default, the decision record traces do not provide, where the approval check runs, and an off switch that has been tested.

Read article

Microsoft Foundry's model router picks a model for each request. For the calling application, the smallest model in the pool sets the context limit, the model subset is also the failover set, and some parameters apply only on certain models.

Read article

For .NET 8 and .NET 9 applications still in service after November 10, 2026, security fixes alone justify the move to .NET 10. What changes on build agents and servers, what two SDK warnings report, and why a workflow improvement is planned as separate work.

Read article

An example invoice workflow: the assistant reviews and recommends, an authorized employee approves the exact payment details, and the application records that approval and checks a timed-out accounting call before any retry.

Read article

App Service's built-in MCP authorization decides who reaches the server, not which tools each caller may use. How a .NET line-of-business application filters the tool list, checks permissions when a tool is called, and handles tokens.

Read article

Azure AI Search keeps its own copy of who may read each document. Some permission changes at the source, such as a parent-scope change in SharePoint, reach the index only when a resync runs; until then it answers from the older copy.

Read article

SQL Server 2025 can call an embedding model from a T-SQL statement. That moves the endpoint credential, the retry setting and the outbound REST calls into the database, and changes how batches, rollback and chunking are planned.

Read article

Adding AI to a running .NET application starts with registering a model client. Each call it makes is billed, can fail or return wrong text, and the pipeline order around it decides what a cached answer skips and what reaches the trace.

Read article

An agent's scope is the tool list registered in code, not a line in its prompt. Microsoft.Extensions.AI's published retry defaults decide what reaches the model when a tool call fails, and the finish line is a record the application can check.

Read article

A build-versus-buy decision starts with the workflow a customer would notice, then runs two diagnostics against the incumbent product: how its configured logic is developed and released, and whether one complete slice of records, associations, history, and files can be reconstructed somewhere else.

Read article

An AI assistant answers from whatever copy of the data it reads. Before adding the model, check four things: which copy it reads, whether deletes reach the index, whose permissions the query carries, and whether retrieval is logged.

Read article

SQL Server 2025 stores embeddings natively and searches them with VECTOR_DISTANCE, both generally available. The approximate DiskANN index is preview and cannot be deployed through a DacPac. Below about 50,000 candidate vectors after filtering, Microsoft's guidance points to exact search.

Read article

Support for .NET 8 and .NET 9 ends on November 10, 2026. Before committing to an upgrade date, check package compatibility, vendor commitments, deployment type, runtime, hosting, and build agents.

Read article

Where tenant scope leaks, how shared, per-schema and per-tenant databases trade isolation against cost and restore, and what support needs on screen.

Read article

What a customer-facing MCP server has to enforce when a customer's own AI assistant books or reorders: authorization on every call, holds that expire, an approval the business records itself, and reconciliation when a payment call times out.

Read article

How an AI diligence application ties each flagged issue to its source document, reopens review when a document is replaced, and holds follow-up questions until a lawyer approves them.

Read article

How a custom AI knowledge base answers from written records with a citation, how a retired source is removed from the retrieval index and checked, and why answers are filtered by the asker's permissions.

Read article

What WebMCP means for websites, how browser agents discover page tools, and how companies can prepare one high-value flow at a time.

Read article

Where an MCP server fits on a custom software platform: start with one workflow where data already leaves by hand, expose read-only resources and one drafting tool, and keep permission checks and logs in the application.

Read article

How production API integrations decide which system owns each field, handle duplicate and out-of-order webhooks, and resolve a write whose outcome is unknown before replaying it.

Read article

What to ask a C# and .NET development company: how it finds the business rules already running in your systems, when it migrates incrementally instead of rewriting, and how it tests current behavior before changing it.

Read article

Before replacing a legacy application, find the rules it quietly enforces, replace it in phases where the platform allows, and compare old and new reports side by side before cutover.

Read article