<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Stewart&apos;s Notes</title><description>A quiet personal blog about reliable AI systems, accessibility, and practical software work.</description><link>https://discostew-blog.pages.dev/</link><item><title>Making AI Reliability Accessible</title><link>https://discostew-blog.pages.dev/posts/ai-reliability-accessibility/</link><guid isPermaLink="true">https://discostew-blog.pages.dev/posts/ai-reliability-accessibility/</guid><description>A short note from Stewart on treating AI reliability as an accessibility practice, not just an infrastructure concern.</description><pubDate>Tue, 09 Jun 2026 00:00:00 GMT</pubDate></item><item><title>Fidelity over backward compatibility</title><link>https://discostew-blog.pages.dev/posts/fidelity-over-backward-compatibility/</link><guid isPermaLink="true">https://discostew-blog.pages.dev/posts/fidelity-over-backward-compatibility/</guid><description>Why I accepted breaking API changes in Propontis when the old schema hid important evidence boundaries.</description><pubDate>Tue, 09 Jun 2026 00:00:00 GMT</pubDate></item><item><title>I fixed an eval by hashing the contract, not the file</title><link>https://discostew-blog.pages.dev/posts/hashing-the-contract-not-the-file/</link><guid isPermaLink="true">https://discostew-blog.pages.dev/posts/hashing-the-contract-not-the-file/</guid><description>Why an evaluation baseline should track the public behavior users depend on instead of private implementation bytes.</description><pubDate>Tue, 09 Jun 2026 00:00:00 GMT</pubDate></item><item><title>Let the LLM write the bio. Make the database check the facts.</title><link>https://discostew-blog.pages.dev/posts/let-the-llm-write-make-the-database-check-facts/</link><guid isPermaLink="true">https://discostew-blog.pages.dev/posts/let-the-llm-write-make-the-database-check-facts/</guid><description>A design note on generated biographies, stat-claim extraction, and using structured data to keep prose honest.</description><pubDate>Tue, 09 Jun 2026 00:00:00 GMT</pubDate></item><item><title>MCP servers have a context tax</title><link>https://discostew-blog.pages.dev/posts/mcp-servers-and-context-tax/</link><guid isPermaLink="true">https://discostew-blog.pages.dev/posts/mcp-servers-and-context-tax/</guid><description>Why I do not want persistent MCP servers in my default Hermes profile unless the recurring value is obvious.</description><pubDate>Tue, 09 Jun 2026 00:00:00 GMT</pubDate></item><item><title>The model should explain the evidence, not become the evidence</title><link>https://discostew-blog.pages.dev/posts/model-should-explain-evidence-not-be-evidence/</link><guid isPermaLink="true">https://discostew-blog.pages.dev/posts/model-should-explain-evidence-not-be-evidence/</guid><description>A short decision story about source authority, provenance, and keeping generated prose out of the truth path.</description><pubDate>Tue, 09 Jun 2026 00:00:00 GMT</pubDate></item><item><title>Removing ChromaDB made my RAG system more reliable</title><link>https://discostew-blog.pages.dev/posts/removing-chromadb-made-rag-reliable/</link><guid isPermaLink="true">https://discostew-blog.pages.dev/posts/removing-chromadb-made-rag-reliable/</guid><description>Why I removed a vector database from my baseball RAG project and kept DuckDB as the source of truth for structured facts.</description><pubDate>Tue, 09 Jun 2026 00:00:00 GMT</pubDate></item></channel></rss>