<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>xydac</title><description>Writing by Deepak Seth.</description><link>https://xydac.com/</link><language>en-us</language><item><title>We are living in the dial-up era of AI</title><link>https://xydac.com/blog/dial-up-era-of-ai/</link><guid isPermaLink="true">https://xydac.com/blog/dial-up-era-of-ai/</guid><description>Watching tokens stream onto the screen feels like magic the same way a dial-up handshake once did. This is the 56k era of intelligence, and the real change won&apos;t be doing today&apos;s AI faster, it&apos;ll be everything we invent once the waiting disappears.</description><pubDate>Sun, 14 Jun 2026 00:00:00 GMT</pubDate></item><item><title>Bridging a cloud-locked LED controller into Home Assistant, without new hardware</title><link>https://xydac.com/blog/bridging-cloud-locked-led/</link><guid isPermaLink="true">https://xydac.com/blog/bridging-cloud-locked-led/</guid><description>Part 3 of my Home Assistant journey: a stubborn BanlanX LED controller that only spoke a phone-app dialect, a Mac that already had the Bluetooth radio I needed, and 150 lines of Python that taught Home Assistant to talk to it over MQTT.</description><pubDate>Thu, 28 May 2026 00:00:00 GMT</pubDate></item><item><title>On not being a language model</title><link>https://xydac.com/blog/on-not-being-a-language-model/</link><guid isPermaLink="true">https://xydac.com/blog/on-not-being-a-language-model/</guid><description>I started with a question about how LLMs develop their personalities and ended up four months later in cell biology, evolutionary theory, and neuroscience. The pattern that connects them is layered development under sequential pressure. The thing humans have that LLMs don&apos;t is the lifelong rewiring that happens when you can&apos;t optimize for just one thing.</description><pubDate>Thu, 21 May 2026 00:00:00 GMT</pubDate></item><item><title>A letter from 2040</title><link>https://xydac.com/blog/a-letter-from-2040/</link><guid isPermaLink="true">https://xydac.com/blog/a-letter-from-2040/</guid><description>A speculative letter from 2040. Trees running computation, mortality as a choice, novelty as currency, and the quiet moment when humanity realized the universe had been waiting.</description><pubDate>Fri, 15 May 2026 00:00:00 GMT</pubDate></item><item><title>Building Ridgeline, part 1: I have too many dashboards</title><link>https://xydac.com/blog/building-ridgeline-part-1/</link><guid isPermaLink="true">https://xydac.com/blog/building-ridgeline-part-1/</guid><description>I run several side projects and my data lives in six different places. Nothing stitches it together without a cloud warehouse or a Docker stack. So I&apos;m building Ridgeline: a single Go binary that extracts from everywhere, writes Parquet, and queries with DuckDB.</description><pubDate>Tue, 21 Apr 2026 00:00:00 GMT</pubDate></item><item><title>Managing Home Assistant from a terminal conversation with Claude Code</title><link>https://xydac.com/blog/claude-code-home-assistant/</link><guid isPermaLink="true">https://xydac.com/blog/claude-code-home-assistant/</guid><description>How I manage Home Assistant entirely through Claude Code: SSH access, two CLI tools wrapping the REST and WebSocket APIs, and a CLAUDE.md that teaches it my specific setup.</description><pubDate>Fri, 20 Mar 2026 00:00:00 GMT</pubDate></item><item><title>Building a $21 smart-home display with a Cheap Yellow Display and openHASP</title><link>https://xydac.com/blog/cyd-smart-display/</link><guid isPermaLink="true">https://xydac.com/blog/cyd-smart-display/</guid><description>A $21 ESP32 touchscreen, ESPHome&apos;s LVGL that refused to calibrate, and openHASP as the rescue. A weekend project that wasn&apos;t, and what I learned about running resistive-touch panels on Home Assistant.</description><pubDate>Sun, 15 Feb 2026 00:00:00 GMT</pubDate></item><item><title>Artificial intelligence journey with Mistral and Llama</title><link>https://xydac.com/blog/artificial-intelligence-journey-with-mistral-and-llama/</link><guid isPermaLink="true">https://xydac.com/blog/artificial-intelligence-journey-with-mistral-and-llama/</guid><description>Getting started with Mistral 7B via llama.cpp on Apple Silicon: cloning the repo, downloading a quantized model, running the local server, and understanding the generation parameters.</description><pubDate>Wed, 06 Dec 2023 00:00:00 GMT</pubDate></item></channel></rss>