Hacker News: winterscotthttps://news.ycombinator.com/user?id=winterscottHacker News RSShttps://hnrss.org/hnrss v2.1.1Wed, 29 Jul 2026 21:29:45 +0000<![CDATA[New comment by winterscott in "Show HN: A browser-based video editor that renders videos directly with FFmpeg"]]>This is a great idea. I always enjoy seeing people build practical tools like this. Looking forward to trying it out and seeing where it goes.

]]>Mon, 27 Jul 2026 08:23:41 +0000https://news.ycombinator.com/item?id=49066618winterscotthttps://news.ycombinator.com/item?id=49066618https://news.ycombinator.com/item?id=49066618<![CDATA[New comment by winterscott in "Show HN: Brolly, a plain-text weather forecast site"]]>This is actually a great format for LLM consumption too.

Structured, concise weather information without unnecessary UI noise is much easier for agents to parse and summarize.

A JSON endpoint or an MCP server on top of this would be an interesting extension.

]]>Sun, 26 Jul 2026 00:52:16 +0000https://news.ycombinator.com/item?id=49053486winterscotthttps://news.ycombinator.com/item?id=49053486https://news.ycombinator.com/item?id=49053486<![CDATA[New comment by winterscott in "Transcribe.cpp"]]>The numerical validation and WER testing are what stand out to me here. A lot of local ASR projects claim broad model support, but it is often difficult to know whether the converted models still match their reference implementations. Having one embeddable engine across Vulkan, Metal and CUDA, along with maintained language bindings, addresses a real distribution problem. How stable do you expect the C API and model format to be after v0.1? In particular, could an application eventually switch between different model families without needing model-specific preprocessing code?

]]>Mon, 20 Jul 2026 02:33:27 +0000https://news.ycombinator.com/item?id=48973779winterscotthttps://news.ycombinator.com/item?id=48973779https://news.ycombinator.com/item?id=48973779