<rss version="2.0" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Hacker News: everlier</title><link>https://news.ycombinator.com/user?id=everlier</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Tue, 21 Apr 2026 06:28:17 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=everlier" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[New comment by everlier in "My first impressions on ROCm and Strix Halo"]]></title><description><![CDATA[
<p>owning GGUF conversion step is good in sone circumstances, but running in fp16 is below optimal for this hardware due to low-ish bandwidth.<p>It looks like context is set to 32k which is the bare minimum needed for OpenCode with its ~10k initial system prompt. So overall, something like Unsloth's UD q8 XL or q6 XL quants free up a lot of memory and bandwidth moving into the next tier of usefulness.</p>
]]></description><pubDate>Sun, 19 Apr 2026 02:44:19 +0000</pubDate><link>https://news.ycombinator.com/item?id=47821402</link><dc:creator>everlier</dc:creator><comments>https://news.ycombinator.com/item?id=47821402</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47821402</guid></item><item><title><![CDATA[New comment by everlier in "Ask HN: Who is using OpenClaw?"]]></title><description><![CDATA[
<p>I tried, really really hard but then I realised that I essence it's a poorly written agentic coding assistant that wastes a lot of tokens antropomorphising itself while forcing me to debug via WhatsApp instead of normal tools. So I leaned into that and made OpenCode my general assistant, it worked much better in this aspect.</p>
]]></description><pubDate>Wed, 15 Apr 2026 21:23:32 +0000</pubDate><link>https://news.ycombinator.com/item?id=47785424</link><dc:creator>everlier</dc:creator><comments>https://news.ycombinator.com/item?id=47785424</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47785424</guid></item><item><title><![CDATA[New comment by everlier in "Trusted access for the next era of cyber defense"]]></title><description><![CDATA[
<p>Changes as we speak, z.ai is the first one to show differential pricing</p>
]]></description><pubDate>Tue, 14 Apr 2026 21:55:12 +0000</pubDate><link>https://news.ycombinator.com/item?id=47772000</link><dc:creator>everlier</dc:creator><comments>https://news.ycombinator.com/item?id=47772000</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47772000</guid></item><item><title><![CDATA[New comment by everlier in "Mi – agentic harness in 30 lines of JavaScript"]]></title><description><![CDATA[
<p>After seeing a recent video from Theo, I wanted to see how far I can take a harness contained in just 30 lines of JavaScript. Turns out - far enough to be useful, it handles simple tasks just fine, works with both cloud and local models, uses just three tools (but can do with a single one, frankly speaking), cleanly handles detached commands or cancellation mid-run, has non-interactive mode and can be run with NPX.</p>
]]></description><pubDate>Mon, 13 Apr 2026 13:13:51 +0000</pubDate><link>https://news.ycombinator.com/item?id=47751489</link><dc:creator>everlier</dc:creator><comments>https://news.ycombinator.com/item?id=47751489</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47751489</guid></item><item><title><![CDATA[Mi – agentic harness in 30 lines of JavaScript]]></title><description><![CDATA[
<p>Article URL: <a href="https://github.com/av/mi">https://github.com/av/mi</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=47751488">https://news.ycombinator.com/item?id=47751488</a></p>
<p>Points: 3</p>
<p># Comments: 1</p>
]]></description><pubDate>Mon, 13 Apr 2026 13:13:51 +0000</pubDate><link>https://github.com/av/mi</link><dc:creator>everlier</dc:creator><comments>https://news.ycombinator.com/item?id=47751488</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47751488</guid></item><item><title><![CDATA[New comment by everlier in "The Quality Wall of AI Adoption"]]></title><description><![CDATA[
<p>Thanks! Caustics shader is one of the treasures in our current landing :-)</p>
]]></description><pubDate>Wed, 08 Apr 2026 21:11:31 +0000</pubDate><link>https://news.ycombinator.com/item?id=47696337</link><dc:creator>everlier</dc:creator><comments>https://news.ycombinator.com/item?id=47696337</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47696337</guid></item><item><title><![CDATA[The Quality Wall of AI Adoption]]></title><description><![CDATA[
<p>Article URL: <a href="https://jitera.com/blog/in-vs-through/">https://jitera.com/blog/in-vs-through/</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=47696234">https://news.ycombinator.com/item?id=47696234</a></p>
<p>Points: 1</p>
<p># Comments: 2</p>
]]></description><pubDate>Wed, 08 Apr 2026 21:03:11 +0000</pubDate><link>https://jitera.com/blog/in-vs-through/</link><dc:creator>everlier</dc:creator><comments>https://news.ycombinator.com/item?id=47696234</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47696234</guid></item><item><title><![CDATA[New comment by everlier in "Pi: Another AI agent toolkit, but this one is interesting"]]></title><description><![CDATA[
<p>Pi is a great set of libraries, I would tend to say its underappreciated previously, but now it's fairly mainstream</p>
]]></description><pubDate>Mon, 30 Mar 2026 23:34:15 +0000</pubDate><link>https://news.ycombinator.com/item?id=47581004</link><dc:creator>everlier</dc:creator><comments>https://news.ycombinator.com/item?id=47581004</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47581004</guid></item><item><title><![CDATA[Show HN: AI for Your Team]]></title><description><![CDATA[
<p>The product was there for the past two years, but we're only now approaching a release on global market. In a nutshell, it's a workspace for non-technical people to work with AI without forcing anyone at the hard end of the slop, as the agents artifacts are shared and available to everyone in the team equally. Jitera agents have all the typical features: memory, skills, MCP, scheduled tasks, access to enterprise data storage, but go a step beyound in customization and integrations.</p>
<hr>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=47580992">https://news.ycombinator.com/item?id=47580992</a></p>
<p>Points: 1</p>
<p># Comments: 0</p>
]]></description><pubDate>Mon, 30 Mar 2026 23:31:40 +0000</pubDate><link>https://jitera.com/</link><dc:creator>everlier</dc:creator><comments>https://news.ycombinator.com/item?id=47580992</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47580992</guid></item><item><title><![CDATA[New comment by everlier in "OpenAI Plans Launch of Desktop 'Superapp'"]]></title><description><![CDATA[
<p>They are playing catch up with Anthropic's in this functionality. Claude's app was unified and extended with new use-cases since pretty much the very beginning and OpenAIs approach is just a reflection of their attempt too shoot for all targets once (with independent teams, of course).</p>
]]></description><pubDate>Sat, 21 Mar 2026 00:08:42 +0000</pubDate><link>https://news.ycombinator.com/item?id=47462577</link><dc:creator>everlier</dc:creator><comments>https://news.ycombinator.com/item?id=47462577</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47462577</guid></item><item><title><![CDATA[Interface Hall of Shame (1999)]]></title><description><![CDATA[
<p>Article URL: <a href="http://hallofshame.gp.co.at/mshame.htm">http://hallofshame.gp.co.at/mshame.htm</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=47462538">https://news.ycombinator.com/item?id=47462538</a></p>
<p>Points: 4</p>
<p># Comments: 0</p>
]]></description><pubDate>Sat, 21 Mar 2026 00:04:01 +0000</pubDate><link>http://hallofshame.gp.co.at/mshame.htm</link><dc:creator>everlier</dc:creator><comments>https://news.ycombinator.com/item?id=47462538</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47462538</guid></item><item><title><![CDATA[A History of Local LLMs]]></title><description><![CDATA[
<p>Article URL: <a href="https://av.codes/blog/local-llms-history/">https://av.codes/blog/local-llms-history/</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=47461517">https://news.ycombinator.com/item?id=47461517</a></p>
<p>Points: 4</p>
<p># Comments: 0</p>
]]></description><pubDate>Fri, 20 Mar 2026 22:24:14 +0000</pubDate><link>https://av.codes/blog/local-llms-history/</link><dc:creator>everlier</dc:creator><comments>https://news.ycombinator.com/item?id=47461517</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47461517</guid></item><item><title><![CDATA[New comment by everlier in "OpenCode – Open source AI coding agent"]]></title><description><![CDATA[
<p>OpenCode is an awesome tool.<p>Many folks from other tools are only getting exposed to the same functionality they got used to, but it offers much more than other harnesses, especially for remote coding.<p>You can start a service via `opencode serve`, it can be accessed from anywhere and has great experience on mobile except a few bugs. It's a really good way to work with your agents remotely, goes really well with TailScale.<p>The WebUI that they have can connect to multiple OpenCode backends at once, so you may use multiple VPS-es for various projects you have and control all of them from a single place.<p>Lastly, there's a desktop app, but TBH I find it redundant when WebUI has everything needed.<p>Make no mistakes though, it's not a perfect tool, my gripes with it:<p>- There are random bugs with loading/restoring state of the session<p>- Model/Provider selection switch across sessions/projects is often annoying<p>- I had a bug making Sonnet/Opus unusable from mobile phone because phone's clock was 150ms ahead of laptop's (ID generation)<p>- Sometimes agent get randomly stuck. It especially sucks for long/nested sessions<p>- WebUI on laptop just completely forgot all the projects at 
one day<p>- `opencode serve` doesn't pick up new skills automatically, it needs to be restarted</p>
]]></description><pubDate>Fri, 20 Mar 2026 22:22:45 +0000</pubDate><link>https://news.ycombinator.com/item?id=47461500</link><dc:creator>everlier</dc:creator><comments>https://news.ycombinator.com/item?id=47461500</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47461500</guid></item><item><title><![CDATA[New comment by everlier in "Ask HN: Is anyone else experiencing AI fatigue?"]]></title><description><![CDATA[
<p>What I feel is more like "too much of a good thing", and too many people that want quick riches concentrating on this field after the web3 fiasco. Both of these seems more like a systematic societal issues rather than the problem with technology itself.</p>
]]></description><pubDate>Fri, 20 Mar 2026 12:31:34 +0000</pubDate><link>https://news.ycombinator.com/item?id=47453645</link><dc:creator>everlier</dc:creator><comments>https://news.ycombinator.com/item?id=47453645</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47453645</guid></item><item><title><![CDATA[New comment by everlier in "40 Months of Prompt Injection"]]></title><description><![CDATA[
<p>A dossier on major events in agentic security since the first release of ChatGPT on November 17th, 2022. Fun fact, first prompt injection was documented three days before that date.</p>
]]></description><pubDate>Fri, 20 Mar 2026 12:28:43 +0000</pubDate><link>https://news.ycombinator.com/item?id=47453620</link><dc:creator>everlier</dc:creator><comments>https://news.ycombinator.com/item?id=47453620</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47453620</guid></item><item><title><![CDATA[40 Months of Prompt Injection]]></title><description><![CDATA[
<p>Article URL: <a href="https://openguard.sh/blog/40-months-of-prompt-injections/">https://openguard.sh/blog/40-months-of-prompt-injections/</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=47453619">https://news.ycombinator.com/item?id=47453619</a></p>
<p>Points: 4</p>
<p># Comments: 1</p>
]]></description><pubDate>Fri, 20 Mar 2026 12:28:43 +0000</pubDate><link>https://openguard.sh/blog/40-months-of-prompt-injections/</link><dc:creator>everlier</dc:creator><comments>https://news.ycombinator.com/item?id=47453619</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47453619</guid></item><item><title><![CDATA[There Is No Firewall for English]]></title><description><![CDATA[
<p>Article URL: <a href="https://openguard.sh/blog/english-firewall/">https://openguard.sh/blog/english-firewall/</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=47442822">https://news.ycombinator.com/item?id=47442822</a></p>
<p>Points: 7</p>
<p># Comments: 0</p>
]]></description><pubDate>Thu, 19 Mar 2026 17:26:16 +0000</pubDate><link>https://openguard.sh/blog/english-firewall/</link><dc:creator>everlier</dc:creator><comments>https://news.ycombinator.com/item?id=47442822</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47442822</guid></item><item><title><![CDATA[New comment by everlier in "Node.js needs a virtual file system"]]></title><description><![CDATA[
<p>We call it a slip slop at work, it's ok to slip some slop if it's "our" slop :-)</p>
]]></description><pubDate>Tue, 17 Mar 2026 17:40:31 +0000</pubDate><link>https://news.ycombinator.com/item?id=47415818</link><dc:creator>everlier</dc:creator><comments>https://news.ycombinator.com/item?id=47415818</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47415818</guid></item><item><title><![CDATA[New comment by everlier in "Kagi Small Web"]]></title><description><![CDATA[
<p>Never thought about this, but it makes sense they don't want a better local search, just for users to rely more on their product. It's messed up - so much time and human potential wasted on poor search and ads.</p>
]]></description><pubDate>Tue, 17 Mar 2026 17:35:27 +0000</pubDate><link>https://news.ycombinator.com/item?id=47415763</link><dc:creator>everlier</dc:creator><comments>https://news.ycombinator.com/item?id=47415763</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47415763</guid></item><item><title><![CDATA[New comment by everlier in "The Webpage Has Instructions. The Agent Has Your Credentials"]]></title><description><![CDATA[
<p>Great observation! OpenGuard as a project aims to solve exactly this problem - a protocol-level "firewall" for LLM APIs.</p>
]]></description><pubDate>Mon, 16 Mar 2026 13:31:46 +0000</pubDate><link>https://news.ycombinator.com/item?id=47398793</link><dc:creator>everlier</dc:creator><comments>https://news.ycombinator.com/item?id=47398793</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47398793</guid></item></channel></rss>