Hacker News: saltwoundshttps://news.ycombinator.com/user?id=saltwoundsHacker News RSShttps://hnrss.org/hnrss v2.1.1Wed, 02 Sep 2026 09:12:41 +0000<![CDATA[New comment by saltwounds in "Can I run AI locally?"]]>I use Raycast and connect it to LM Studio to run text clean up and summaries often. The models are small enough I keep them in memory more often than not

]]>Fri, 13 Mar 2026 19:11:21 +0000https://news.ycombinator.com/item?id=47368375saltwoundshttps://news.ycombinator.com/item?id=47368375https://news.ycombinator.com/item?id=47368375<![CDATA[New comment by saltwounds in "New iPad Air, powered by M4"]]>You should be able to DFU, but when it phones home it'll require a software upgrade

]]>Mon, 02 Mar 2026 15:26:51 +0000https://news.ycombinator.com/item?id=47219179saltwoundshttps://news.ycombinator.com/item?id=47219179https://news.ycombinator.com/item?id=47219179<![CDATA[New comment by saltwounds in "Show HN: Moonshine Open-Weights STT models – higher accuracy than WhisperLargev3"]]>Streaming transcription is crazy fast on an M1. Would be great to use this as a local option versus Wispr Flow.

]]>Wed, 25 Feb 2026 01:29:18 +0000https://news.ycombinator.com/item?id=47146138saltwoundshttps://news.ycombinator.com/item?id=47146138https://news.ycombinator.com/item?id=47146138<![CDATA[New comment by saltwounds in "Show HN: The Analog I – Inducing Recursive Self-Modeling in LLMs [pdf]"]]>I haven't come across this technique before. How'd you uncover it? I wonder how it'll work in Claude Code over long conversations

]]>Fri, 16 Jan 2026 16:04:34 +0000https://news.ycombinator.com/item?id=46647924saltwoundshttps://news.ycombinator.com/item?id=46647924https://news.ycombinator.com/item?id=46647924<![CDATA[New comment by saltwounds in "Show HN: mdto.page – Turn Markdown into a shareable webpage instantly"]]>I like the idea and expiration dates. Got any example pages using either theme?

]]>Fri, 16 Jan 2026 16:00:02 +0000https://news.ycombinator.com/item?id=46647865saltwoundshttps://news.ycombinator.com/item?id=46647865https://news.ycombinator.com/item?id=46647865<![CDATA[New comment by saltwounds in "Fallout from the AWS outage: Smart mattresses go rogue"]]>Even with AC air gets trapped between the mattress, the person, and covers. Different materials help circulate air and keep a cooler feel.

]]>Tue, 21 Oct 2025 17:20:29 +0000https://news.ycombinator.com/item?id=45658510saltwoundshttps://news.ycombinator.com/item?id=45658510https://news.ycombinator.com/item?id=45658510<![CDATA[New comment by saltwounds in "Claude Skills"]]>I connect local models to MCPs with LM Studio and I'm blown away at how good they are. But the issues creep up when you hit longer context like you said.

]]>Fri, 17 Oct 2025 16:30:25 +0000https://news.ycombinator.com/item?id=45618652saltwoundshttps://news.ycombinator.com/item?id=45618652https://news.ycombinator.com/item?id=45618652<![CDATA[New comment by saltwounds in "Claude Skills"]]>OpenAI and Anthropic's real moat is hardware. For local LLMs, context length and hardware performance are the limiting factors. Qwen3 4B with a 32,768 context window is great. Until it begins filling up and performance drops quickly.

I use local models when possible. MCPs work well, but their large context injection makes switching to an online provider the no-brainer.

]]>Fri, 17 Oct 2025 16:28:32 +0000https://news.ycombinator.com/item?id=45618623saltwoundshttps://news.ycombinator.com/item?id=45618623https://news.ycombinator.com/item?id=45618623