<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Surya Kant Sahu</title><link>https://ojus1.github.io/</link><description>Notes on machine learning: transformers, learning theory, audio, and running models in places they were not built for.</description><item><title>Attention, from the softmax up</title><link>https://ojus1.github.io/posts/attention-from-the-softmax-up/</link><guid>https://ojus1.github.io/posts/attention-from-the-softmax-up/</guid><pubDate>Wed, 04 Nov 2026 00:00:00 GMT</pubDate><description>Scaled dot-product attention one piece at a time, with a temperature dial you can turn to watch a query make up its mind.</description></item><item><title>A language model in this tab</title><link>https://ojus1.github.io/posts/a-language-model-in-this-tab/</link><guid>https://ojus1.github.io/posts/a-language-model-in-this-tab/</guid><pubDate>Wed, 21 Oct 2026 00:00:00 GMT</pubDate><description>A small instruction-tuned model running inside the page, on your GPU through WebGPU or on your CPU, with nothing sent to a server.</description></item><item><title>Low-rank updates, plotted</title><link>https://ojus1.github.io/posts/low-rank-updates-plotted/</link><guid>https://ojus1.github.io/posts/low-rank-updates-plotted/</guid><pubDate>Thu, 08 Oct 2026 00:00:00 GMT</pubDate><description>What LoRA's rank actually buys you, as pixels and as curves. A sketch of a sweep, with charts you can hover, filter and smooth.</description></item><item><title>Watching an optimiser move</title><link>https://ojus1.github.io/posts/watching-an-optimiser-move/</link><guid>https://ojus1.github.io/posts/watching-an-optimiser-move/</guid><pubDate>Wed, 23 Sep 2026 00:00:00 GMT</pubDate><description>Gradient descent with momentum on a surface you can click, and a short film of a diffusion model turning noise into a spiral.</description></item></channel></rss>
