<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Agrohi blog</title><description>Engineering notes from Agrohi: real-time systems, on-device machine learning, and software across web and native platforms.</description><link>https://agrohi.com/blog/</link><language>en</language><item><title>One Rust engine, four hosts: browser, Android, macOS and Linux</title><link>https://agrohi.com/blog/one-rust-engine-four-hosts/</link><guid isPermaLink="true">https://agrohi.com/blog/one-rust-engine-four-hosts/</guid><description>How nomus runs one real-time audio engine in a Chrome extension, an Android app and desktop apps for macOS and Linux, and what each platform demanded that the engine could not provide.</description><pubDate>Sun, 27 Sep 2026 00:00:00 GMT</pubDate><category>rust</category><category>audio</category><category>architecture</category><category>webassembly</category><category>android</category><category>macos</category><category>linux</category></item><item><title>20 ms end to end: real-time audio in an AudioWorklet with zero-import WASM</title><link>https://agrohi.com/blog/audioworklet-zero-import-wasm/</link><guid isPermaLink="true">https://agrohi.com/blog/audioworklet-zero-import-wasm/</guid><description>How nomus runs a million-parameter recurrent network on the browser&apos;s audio thread: a Rust engine compiled to a WebAssembly module with no imports, a flat C ABI, and a worklet that never throws.</description><pubDate>Sun, 27 Sep 2026 00:00:00 GMT</pubDate><category>webassembly</category><category>web-audio</category><category>audioworklet</category><category>rust</category><category>chrome-extension</category><category>performance</category></item><item><title>Honest evaluation: how we measured a music filter, and where it still fails</title><link>https://agrohi.com/blog/honest-evaluation-music-filter/</link><guid isPermaLink="true">https://agrohi.com/blog/honest-evaluation-music-filter/</guid><description>The metrics, test sets and selection rules behind nomus&apos;s voice-preserving music filter, the runs that collapsed, the competitor benchmark we lost on one axis, and the cases where the shipping model still fails.</description><pubDate>Sun, 27 Sep 2026 00:00:00 GMT</pubDate><category>machine-learning</category><category>evaluation</category><category>audio</category><category>source-separation</category><category>mlops</category></item></channel></rss>