16 October 2026
I moved my GitHub Actions jobs to a self-hosted runner in Docker to save minutes. Jobs first failed, then got slower, because every job downloaded what GitHub's hosted image already has. Putting those tools in the runner image brought a deploy from 2:00 to 1:28.
30 September 2026
AI helped me build a SignalR update that was broadcast to every connected client instead of just the relevant one. The first fix it suggested, filtering in React, was cosmetic. The real fix was one line in the backend.
17 September 2026
My blog counter went through three implementations in a week: fetching the whole article list to count it, a dedicated count endpoint that still cold-started the backend, and a Cosmos DB Change Feed keeping a Cloudflare KV value current so the read never touches origin. Plus a debugging story about one secret with three different names.
10 September 2026
How I added a moderated comment system to this blog: Cosmos DB's own Change Feed as the queue, an Azure Function scoring comments via Cloudflare Workers AI, least-privilege RBAC kept out of the CI pipeline on purpose, several real bugs found the hard way in production, and a mid-flight model upgrade after the small LLM proved unreliable.
5 August 2026
Tracking down a TCP handler bug that looked like a race condition but turned out to be a stale variable surviving a handler replacement — root-caused through log correlation alone, fixed with a handler map, and a reminder that debugging discipline matters as much as tooling.
13 July 2026
Debugging Skills Part 2: using EF Core's TagWith() to label LINQ queries so they're identifiable in the generated SQL, then using Query Store to find which tagged query is driving high logical I/O or a bad execution plan.
6 July 2026
Debugging Skills Part 1: tracing a 2022 async memory-leak in a TCP handler system with a custom WithName() task-naming extension inspired by Jon Skeet, then fixing it with linked cancellation tokens.
16 May 2026
Building a reusable Azure DevOps YAML pipeline for a small .NET helper library: AI-assisted (Aider + Qwen2.5) unit tests, automated SemVer NuGet packaging to GitHub Packages, restore caching, a dotnet format style gate, and the Cache@2 quirk that needed a manual mkdir to fix.
29 April 2026
Foundry Local reaching GA raised hopes for a real Ollama replacement on Intel GPU — but both tested OpenVINO models hard-capped responses at 199 tokens regardless of the configured max_tokens, a blocker for agentic coding workflows like Aider.
21 March 2026
A streamlined Windows setup for running Aider + Ollama fully locally on an Intel GPU: zero cloud dependency, unlimited context, and a qwen2.5-coder:14b model that handles .NET refactors, new endpoints, and unit tests as a genuine local pair programmer.
17 January 2026
The full write-up (and code) for the Semantic Kernel multi-agent Hangman game, promised as a follow-up to the earlier 'finally got it working' milestone post.
15 January 2026
The first real working milestone for the Semantic Kernel Hangman game: switching to a bigger qwen2.5:32b-instruct model via IPEX-LLM Ollama and simplifying down to 2 agents (no host agent) finally got a full game to completion.
9 January 2026
Three small local models (llama3.2:3b, phi4-mini, qwen2.5:7b) and a host agent structure for the Hangman game — writing the game logic was easy, getting agents to reliably call functions and use tool output instead of hallucinating turned out to be the real challenge.