https://vondraysanford.com/writing/ Notes, tutorials, and honest reflections on .NET, AI engineering, RAG pipelines, and multi-agent systems. en-us Thu, 17 Sep 2026 00:00:00 GMT https://vondraysanford.com/writing/2026-09-17-your-ai-coding-history-is-a-dataset-here-s-how-i-d-actually-.html https://vondraysanford.com/writing/2026-09-17-your-ai-coding-history-is-a-dataset-here-s-how-i-d-actually-.html Thu, 17 Sep 2026 00:00:00 GMT AI The ai-data-extractor project can pull your entire Claude Code, Cursor, and Cline chat history into structured data. That's interesting. What you do with it afterward is where things get genuinely useful. https://vondraysanford.com/writing/2026-09-15-the-ai-safety-pact-is-a-cartel-in-a-lab-coat.html https://vondraysanford.com/writing/2026-09-15-the-ai-safety-pact-is-a-cartel-in-a-lab-coat.html Tue, 15 Sep 2026 00:00:00 GMT AI When the CEOs of OpenAI, Anthropic, and Google DeepMind agreed to slow AI development for "safety," skeptics didn't have to look hard for the ulterior motive. This smells less like a safety pause and more like incumbents pulling up the ladder. https://vondraysanford.com/writing/2026-09-14-sparkdash-i-gave-a-320b-model-a-spec-and-pointed-it-at-the-s.html https://vondraysanford.com/writing/2026-09-14-sparkdash-i-gave-a-320b-model-a-spec-and-pointed-it-at-the-s.html Mon, 14 Sep 2026 00:00:00 GMT AI GLM-5.3-Flash, 320B parameters split across two DGX Sparks over a ConnectX-7 cable, built its own monitoring dashboard in 44 minutes. The scorecard was written before the session: what it got right, what broke, and whether the fixes went to the cause. https://vondraysanford.com/writing/2026-09-13-320b-on-two-sparks-what-it-actually-takes-to-run-a-mixture-o.html https://vondraysanford.com/writing/2026-09-13-320b-on-two-sparks-what-it-actually-takes-to-run-a-mixture-o.html Sun, 13 Sep 2026 00:00:00 GMT Infrastructure I bridged two NVIDIA DGX Sparks, loaded a 320B open-weights MoE model across both, and wired the endpoint into GitHub Copilot Chat. Every step, every error, and every number, from cable to first correct answer. https://vondraysanford.com/writing/2026-09-10-engineering-patterns-beat-model-power-what-the-google-ai-age.html https://vondraysanford.com/writing/2026-09-10-engineering-patterns-beat-model-power-what-the-google-ai-age.html Thu, 10 Sep 2026 00:00:00 GMT AI Google's AI Agents Challenge surfaced something most people building multi-agent systems already suspect but rarely say out loud: the model isn't the bottleneck. The architecture is. https://vondraysanford.com/writing/2026-09-08-driftwatch-measured-what-a-logistic-regression-a-coin-flip-a.html https://vondraysanford.com/writing/2026-09-08-driftwatch-measured-what-a-logistic-regression-a-coin-flip-a.html Tue, 08 Sep 2026 00:00:00 GMT mlops Six phases, 21 days, and every number measured. The wrap-up on DriftWatch: what it produced, what it cost, what broke, and what's still not done. https://vondraysanford.com/writing/2026-09-07-the-four-patterns-that-separate-toy-agents-from-production-a.html https://vondraysanford.com/writing/2026-09-07-the-four-patterns-that-separate-toy-agents-from-production-a.html Mon, 07 Sep 2026 00:00:00 GMT AI Google's AI Agents Challenge revealed something most tutorials skip: what set the winning multi-agent systems apart was software engineering, not model choice. Four patterns kept showing up in the report, and none of them is an AI problem. https://vondraysanford.com/writing/2026-09-04-the-model-was-20-driftwatch-part-6.html https://vondraysanford.com/writing/2026-09-04-the-model-was-20-driftwatch-part-6.html Fri, 04 Sep 2026 00:00:00 GMT MLOps Six phases, one thesis: the model is the small part. DriftWatch is done. This post has every number I measured, eight lessons each pinned to something that happened, and what the dashboard proved by being a page instead of a second app. https://vondraysanford.com/writing/2026-09-04-the-monitor-caught-it-the-loop-answered-it-a-human-said-yes-.html https://vondraysanford.com/writing/2026-09-04-the-monitor-caught-it-the-loop-answered-it-a-human-said-yes-.html Fri, 04 Sep 2026 00:00:00 GMT MLOps DriftWatch's drift phase is the one the whole project was built for. A real regime change dropped the champion model to ROC-AUC 0.50, the retrain loop brought a challenger back to 0.99, a screenshot exposed two bugs on the way, and a human still approves the promotion. https://vondraysanford.com/writing/2026-09-04-four-failures-one-green-run-and-the-teardown-that-held-every.html https://vondraysanford.com/writing/2026-09-04-four-failures-one-green-run-and-the-teardown-that-held-every.html Fri, 04 Sep 2026 00:00:00 GMT DevOps DriftWatch Phase 4 build log: how enterprise DevOps experience, not ML theory, solved secretless CI/CD, identity scoping, and a managed endpoint that had to prove it worked before it was allowed to disappear. https://vondraysanford.com/writing/2026-09-03-raw-cycles-in-probability-out-driftwatch-part-3.html https://vondraysanford.com/writing/2026-09-03-raw-cycles-in-probability-out-driftwatch-part-3.html Thu, 03 Sep 2026 00:00:00 GMT MLOps Phase 3 of DriftWatch is a FastAPI endpoint that accepts raw sensor cycles, computes features from the same code training used, and refuses to answer if it cannot write the prediction down. One decision drives all of it. https://vondraysanford.com/writing/2026-09-03-fifty-trials-couldn-t-beat-logistic-regression-driftwatch-pa.html https://vondraysanford.com/writing/2026-09-03-fifty-trials-couldn-t-beat-logistic-regression-driftwatch-pa.html Thu, 03 Sep 2026 00:00:00 GMT MLOps Phase 2 of DriftWatch covered training, MLflow tracking, and model registration in Azure ML. It ended with a surprise: a scaled logistic regression beat XGBoost, including XGBoost after 50 Optuna tuning trials, on engines none of the models had seen. https://vondraysanford.com/writing/2026-09-02-before-the-model-data-quarantine-label-design-and-the-phase-.html https://vondraysanford.com/writing/2026-09-02-before-the-model-data-quarantine-label-design-and-the-phase-.html Wed, 02 Sep 2026 00:00:00 GMT MLOps DriftWatch Phase 1 had no model in it. It was still the phase where the assumptions that wreck ML projects get made, or don't: label leakage, contaminated holdouts, feature code that drifts from serving. I treated all of it as infrastructure work. https://vondraysanford.com/writing/2026-09-01-kiro-crew-and-the-case-for-asynchronous-agents-what-aws-got-.html https://vondraysanford.com/writing/2026-09-01-kiro-crew-and-the-case-for-asynchronous-agents-what-aws-got-.html Tue, 01 Sep 2026 00:00:00 GMT AI AWS open-sourced Kiro Crew to run multiple coding agents across sessions and tasks asynchronously. The architectural bet underneath, async by default, matters more than the feature list, and it comes with a bill for durable state that AWS doesn't pay for you. https://vondraysanford.com/writing/2026-08-28-frontieragent-is-interesting-the-agent-team-mode-is-the-part.html https://vondraysanford.com/writing/2026-08-28-frontieragent-is-interesting-the-agent-team-mode-is-the-part.html Fri, 28 Aug 2026 00:00:00 GMT AI FrontierAgent just open-sourced a multi-agent framework with ReAct and Agent Team modes baked in. I read the architecture with AgentReview fresh in my head and wrote up the questions I'd want answered before trusting the team mode. https://vondraysanford.com/writing/2026-08-25-i-built-a-twitter-bot-detector-in-2022-i-finally-evaluated-i.html https://vondraysanford.com/writing/2026-08-25-i-built-a-twitter-bot-detector-in-2022-i-finally-evaluated-i.html Tue, 25 Aug 2026 00:00:00 GMT AI A look back at the Python NLP pipeline I built in 2022: sentiment analysis, XGBoost bot detection, and the evaluation protocol I should have written four years ago. https://vondraysanford.com/writing/2026-08-24-driftwatch-post-0-the-setup-weekend-that-wasn-t-boring.html https://vondraysanford.com/writing/2026-08-24-driftwatch-post-0-the-setup-weekend-that-wasn-t-boring.html Mon, 24 Aug 2026 00:00:00 GMT AI DriftWatch is my third public AI engineering project, and it exists to prove one thing: the model is maybe 20% of an ML system. Phase 0 was supposed to be the boring setup weekend. It was not. https://vondraysanford.com/writing/2026-08-17-model-routing-is-an-infrastructure-problem-not-a-prompt-prob.html https://vondraysanford.com/writing/2026-08-17-model-routing-is-an-infrastructure-problem-not-a-prompt-prob.html Mon, 17 Aug 2026 00:00:00 GMT AI Google Cloud API Gateway's new model routing feature is getting framed as a developer convenience, but what it's really solving is a production infrastructure problem that anyone running multi-model AI workloads has already hit the hard way. https://vondraysanford.com/writing/2026-08-13-mcp-goes-stateless-what-the-2026-07-28-spec-actually-changes.html https://vondraysanford.com/writing/2026-08-13-mcp-goes-stateless-what-the-2026-07-28-spec-actually-changes.html Thu, 13 Aug 2026 00:00:00 GMT AI The 2026-07-28 MCP spec drops stateful session management entirely. I fought the old session model while building AgentReview's MCP server, so this change matters to me: MCP servers can finally scale like any other stateless HTTP service. https://vondraysanford.com/writing/2026-08-09-three-ai-agents-reviewed-the-same-pull-request-here-s-what-i.html https://vondraysanford.com/writing/2026-08-09-three-ai-agents-reviewed-the-same-pull-request-here-s-what-i.html Mon, 10 Aug 2026 00:00:00 GMT AI AgentReview shipped. Three specialized agents reviewed 8 seeded PRs with 18 planted bugs and came back with 100% precision, 17/18 recall, and a cost of $0.068 per review. The one miss is written up honestly in the README. https://vondraysanford.com/writing/2026-08-08-agentreview-phase-5-measuring-my-own-agents-and-why-precisio.html https://vondraysanford.com/writing/2026-08-08-agentreview-phase-5-measuring-my-own-agents-and-why-precisio.html Sun, 09 Aug 2026 00:00:00 GMT AI Recall is easy to measure when you plant the bugs yourself. Precision is the metric that fights back. It took 45 manual judgments, a scoring worksheet, and a budget guard wired to every token cap before I trusted these numbers. https://vondraysanford.com/writing/2026-08-08-agentreview-phase-4-fan-out-is-easy-synthesis-is-the-hard-pa.html https://vondraysanford.com/writing/2026-08-08-agentreview-phase-4-fan-out-is-easy-synthesis-is-the-hard-pa.html Sun, 09 Aug 2026 00:00:00 GMT AI Getting three agents to run concurrently took an afternoon. Merging their findings without losing signal took real design work, and the approach that survived pairs LLM judgment with plain deterministic code. https://vondraysanford.com/writing/2026-08-08-agentreview-phase-3-three-agents-one-base-class-and-why-lane.html https://vondraysanford.com/writing/2026-08-08-agentreview-phase-3-three-agents-one-base-class-and-why-lane.html Sat, 08 Aug 2026 00:00:00 GMT AI Once the review pipeline moved into a shared base class, the Security Agent came together in under a day and the Docs Agent went even faster. The proof: 55 tests, 15 committed sample reviews, and a security agent that stays quiet on all four non-security diffs. https://vondraysanford.com/writing/2026-08-07-agentreview-phase-2-one-real-agent-real-tools-and-every-hone.html https://vondraysanford.com/writing/2026-08-07-agentreview-phase-2-one-real-agent-real-tools-and-every-hone.html Fri, 07 Aug 2026 00:00:00 GMT AI I built a C# Quality Agent that runs a PR diff through both Roslyn and Claude, then merges the results into one schema-locked findings list. Along the way I dealt with deduplication headaches, hallucinated resolution errors, undocumented MCP headers, and a fresh reminder that cost guardrails belong in code. https://vondraysanford.com/writing/2026-08-05-agentreview-phase-1-build-the-mcp-server-first-ask-questions.html https://vondraysanford.com/writing/2026-08-05-agentreview-phase-1-build-the-mcp-server-first-ask-questions.html Wed, 05 Aug 2026 00:00:00 GMT .NET Before writing a single agent I built the tooling layer: an MCP server in C#/.NET wrapping Roslyn and Semgrep. Building in that order taught me plenty about scanner blind spots, protocol hygiene, and why LLM reasoning belongs on top of static analysis instead of beside it. https://vondraysanford.com/writing/2026-07-30-microsoft-s-three-layer-llm-routing-architecture-what-it-get.html https://vondraysanford.com/writing/2026-07-30-microsoft-s-three-layer-llm-routing-architecture-what-it-get.html Thu, 30 Jul 2026 00:00:00 GMT AI Microsoft's reference architecture for routing AI agent traffic on AKS splits the problem into three clean layers. I've built and run AI services on Azure in production, so I went through it looking for what holds up and what it leaves you to solve on your own. https://vondraysanford.com/writing/2026-07-29-docquery-phase-4-shipping-a-public-rag-demo-that-doesn-t-lie.html https://vondraysanford.com/writing/2026-07-29-docquery-phase-4-shipping-a-public-rag-demo-that-doesn-t-lie.html Wed, 29 Jul 2026 00:00:00 GMT AI DocQuery left my laptop. Turning a local RAG experiment into the live public Interview Vondray demo meant patching an SDK bug on the wire, chasing down an empty-answer mystery, and learning why corpus curation is a security decision. https://vondraysanford.com/writing/2026-07-27-docquery-phase-3-the-ux-layer-that-makes-a-rag-pipeline-feel.html https://vondraysanford.com/writing/2026-07-27-docquery-phase-3-the-ux-layer-that-makes-a-rag-pipeline-feel.html Mon, 27 Jul 2026 00:00:00 GMT AI Streaming citations, live provider switching, and a side-by-side AI race mode. Phase 3 took DocQuery from a working pipeline to something that feels good to use. https://vondraysanford.com/writing/2026-07-25-docquery-phase-2-benchmarking-three-ai-stacks-so-you-don-t-h.html https://vondraysanford.com/writing/2026-07-25-docquery-phase-2-benchmarking-three-ai-stacks-so-you-don-t-h.html Sat, 25 Jul 2026 00:00:00 GMT AI I made DocQuery's entire AI stack swappable with a single config value, then raced a MacBook, an NVIDIA DGX Spark, and Azure against each other. The results overturned most of my assumptions about fast and cheap in local versus cloud inference. https://vondraysanford.com/writing/2026-07-23-mcp-is-the-real-story-behind-ai-video-editors-here-s-why-tha.html https://vondraysanford.com/writing/2026-07-23-mcp-is-the-real-story-behind-ai-video-editors-here-s-why-tha.html Thu, 23 Jul 2026 00:00:00 GMT AI A wave of AI video tools now list MCP support as a headline feature. The interesting part is the protocol underneath, because it's going to show up in a lot more places than video editing. https://vondraysanford.com/writing/2026-07-20-docquery-phase-1-building-a-local-rag-pipeline-in-net-and-th.html https://vondraysanford.com/writing/2026-07-20-docquery-phase-1-building-a-local-rag-pipeline-in-net-and-th.html Mon, 20 Jul 2026 00:00:00 GMT AI I built a fully local RAG app in .NET 10 with Ollama and ChromaDB: upload documents, ask questions, get cited answers. This post covers the architecture, the decisions behind it, and the problems I hit that no tutorial bothers to mention. https://vondraysanford.com/writing/2026-07-14-ai-agents-that-manage-kubernetes-what-sre-agenticops-actuall.html https://vondraysanford.com/writing/2026-07-14-ai-agents-that-manage-kubernetes-what-sre-agenticops-actuall.html Tue, 14 Jul 2026 00:00:00 GMT AI AI-powered SRE agents for Kubernetes are picking up steam fast. I dug into what they can do today, where they fall short, and how much of your infrastructure you should let them touch. https://vondraysanford.com/writing/2026-07-03-apple-chose-google-cloud-over-aws-and-azure-here-s-what-that.html https://vondraysanford.com/writing/2026-07-03-apple-chose-google-cloud-over-aws-and-azure-here-s-what-that.html Fri, 03 Jul 2026 00:00:00 GMT Infrastructure Apple just ran Private Cloud Compute on Google Cloud for the first time, on NVIDIA Blackwell GPUs with dual-vendor attestation. The architecture choices behind that move say a lot about where confidential computing is headed. https://vondraysanford.com/writing/2026-06-30-first-post-what-this-blog-is-about-and-why-i-m-writing-it.html https://vondraysanford.com/writing/2026-06-30-first-post-what-this-blog-is-about-and-why-i-m-writing-it.html Tue, 30 Jun 2026 00:00:00 GMT Personal I'm Vondray Sanford, a software engineer with seven-plus years in enterprise systems. This blog is where I write up what I'm building and learning across .NET, AI engineering, and the infrastructure that keeps it all running.