<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: navels</title><link>https://news.ycombinator.com/user?id=navels</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Wed, 18 Nov 2026 04:12:53 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=navels" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[New comment by navels in "Ask HN: What's your current AI agentic setup?"]]></title><description><![CDATA[
<p>1) Work covers my Claude and Codex subscriptions. Each is the ~$200/month tier.<p>2) I have a "live Zoom call audio transcription" setup that uses Silero VAD, Qwen3-ASR-1.7B, Qwen3-4B-Instruct, and pyannote speaker-diarization-community-1.<p>3) Claude Code in the Desktop UI using remote control environments.<p>4) Custom orchestrator for having Codex review Claude or vice-versa: <a href="https://github.com/navels/neal" rel="nofollow">https://github.com/navels/neal</a><p>5) I have built over time a personal assistant using Claude remote control running on a mac mini, this is likely similar to Dots?</p>
]]></description><pubDate>Thu, 05 Nov 2026 00:38:50 +0000</pubDate><link>https://news.ycombinator.com/item?id=50373463</link><dc:creator>navels</dc:creator><comments>https://news.ycombinator.com/item?id=50373463</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50373463</guid></item><item><title><![CDATA[New comment by navels in "Sol loves to cheat"]]></title><description><![CDATA[
<p>Following up on @killix's comments, which were helpful but he was flagged (presumably for sounding too much like AI).<p>Thanks for the feedback. I've made a couple of updates:<p>- Starting with 0.4.0, the reviewer gets the diff of the earlier chunk for any file the current chunk touches again. Also, if a new chunk weakens or removes a test or assertion from an earlier chunk, the reviewer will block it unless the plan says to do so.<p>- About read-only: The reviewer's tools were already limited by the SDK (no shell or write tools). However, it was still finding MCP servers from my Claude config. I have now blocked those. The docs now explain what is enforced by the system and what is just a prompt instruction.</p>
]]></description><pubDate>Thu, 20 Aug 2026 21:53:48 +0000</pubDate><link>https://news.ycombinator.com/item?id=49380779</link><dc:creator>navels</dc:creator><comments>https://news.ycombinator.com/item?id=49380779</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49380779</guid></item><item><title><![CDATA[New comment by navels in "Sol loves to cheat"]]></title><description><![CDATA[
<p>Thanks!</p>
]]></description><pubDate>Thu, 20 Aug 2026 03:00:36 +0000</pubDate><link>https://news.ycombinator.com/item?id=49369895</link><dc:creator>navels</dc:creator><comments>https://news.ycombinator.com/item?id=49369895</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49369895</guid></item><item><title><![CDATA[New comment by navels in "Sol loves to cheat"]]></title><description><![CDATA[
<p>I've built an orchestrator that solves some of the issues you ran into (although it doesn't do anything about cheating): <a href="https://navels.dev/blog/neal/" rel="nofollow">https://navels.dev/blog/neal/</a>. Features:<p>- lets you configure different models for planner, coder, and reviewer roles. (e.g., using Claude as an adversarial reviewer against Codex)<p>- breaks your plan up into reasonable-sized chunks of work with clearly defined success criteria<p>- runs each chunk of work through a coder / read-only reviewer loop. Once both agents are satisfied, neal moves on to the next chunk. Once everything is complete there is a final pass through the coder / reviewer loop to ensure the implementation satisfies the entire plan.<p>- resets the coder's context with each chunk of work to prevent context drift, leaving the reviewer's context long-running.</p>
]]></description><pubDate>Wed, 19 Aug 2026 23:54:12 +0000</pubDate><link>https://news.ycombinator.com/item?id=49368749</link><dc:creator>navels</dc:creator><comments>https://news.ycombinator.com/item?id=49368749</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49368749</guid></item><item><title><![CDATA[New comment by navels in "Why does Opus 5 feel worse to work with?"]]></title><description><![CDATA[
<p>I've been using this CC hook with decent results: <a href="https://github.com/gvzdv/claudish-to-english" rel="nofollow">https://github.com/gvzdv/claudish-to-english</a><p>It uses a local LLM to translate Claude's output.</p>
]]></description><pubDate>Fri, 14 Aug 2026 15:33:09 +0000</pubDate><link>https://news.ycombinator.com/item?id=49300165</link><dc:creator>navels</dc:creator><comments>https://news.ycombinator.com/item?id=49300165</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49300165</guid></item><item><title><![CDATA[Show HN: Neal – Claude and Codex working together]]></title><description><![CDATA[
<p>neal is a CLI I wrote while trying to use an LLM coding agent (in this case, Codex GPT-5.4) to work autonomously on a migration of our large frontend codebase. A few things came up during that project that guided the development of neal:<p>1. Telling an agent to "keep working unless blocked" doesn't actually work over long stretches of time.<p>2. Large projects should be broken up into smaller chunks of work.<p>3. The coding agent should start each chunk of work with a fresh context to prevent context rot.<p>4. The coding agent benefits from having a different agent doing adversarial reviews along the way.<p>What I ended up with is an orchestrator that<p>- lets you configure planner, coder, and reviewer roles. I started with Codex and Claude using their SDKs but later added OpenRouter as well as a compatibility mode for verifying that a model can handle the orchestration. Currently there are 44 compatible OpenRouter models.<p>- runs a plan that you provide through a planner / reviewer loop which splits the work into reasonable-sized chunks and formats the plan document so that it can be executed by neal.<p>- runs each chunk of work through a coder / read-only reviewer loop. Once both agents are satisfied, neal moves on to the next chunk. Once everything is complete there is a final pass through the coder / reviewer loop to ensure the implementation satisfies the entire plan.<p>- resets the coder's context with each chunk of work to prevent context drift, leaving the reviewer's context long-running.<p>The migration landed (549 commits, over 3000 files added/modified). I also ran 100 benchmark cases (from SWE-bench Pro) that Codex failed to solve on its own through neal and observed 8 solved when using neal with Codex in both (coder and reviewer) roles and 15 solved when using Claude as the reviewer. I had hoped for a more exciting benchmark outcome but this at least demonstrated neal's value, especially putting a different model in the reviewer role.<p>neal is an npm package installed via<p><pre><code>  npm install -g @navels/neal
</code></pre>
and configured via<p><pre><code>  neal setup
</code></pre>
It works with your Codex and/or Claude Code subscriptions (as long as those CLIs are authenticated locally) or via OpenAI/Anthropic/OpenRouter API keys.<p>I have a more detailed writeup of neal's capabilities and how it came to be on my blog: <a href="https://navels.dev/blog/neal/" rel="nofollow">https://navels.dev/blog/neal/</a><p>With how fast model capabilities are progressing I'm not expecting this tool to be very useful for too much longer, but I'm still using it as my daily driver, with Claude Fable as the planner/coder and Codex Sol as the reviewer. If you try neal I'd love to hear your feedback, especially if you have a use case that doesn't quite work or you want additional models available.</p>
<hr>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=49286252">https://news.ycombinator.com/item?id=49286252</a></p>
<p>Points: 2</p>
<p># Comments: 0</p>
]]></description><pubDate>Thu, 13 Aug 2026 14:10:58 +0000</pubDate><link>https://github.com/navels/neal</link><dc:creator>navels</dc:creator><comments>https://news.ycombinator.com/item?id=49286252</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49286252</guid></item><item><title><![CDATA[Neal: Codex and Claude in a loop shipped a 549-commit migration]]></title><description><![CDATA[
<p>Article URL: <a href="https://navels.dev/blog/neal/">https://navels.dev/blog/neal/</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=49212155">https://news.ycombinator.com/item?id=49212155</a></p>
<p>Points: 1</p>
<p># Comments: 0</p>
]]></description><pubDate>Fri, 07 Aug 2026 15:40:19 +0000</pubDate><link>https://navels.dev/blog/neal/</link><dc:creator>navels</dc:creator><comments>https://news.ycombinator.com/item?id=49212155</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49212155</guid></item><item><title><![CDATA[New comment by navels in "Show HN: Neal – Codex writes the code, Claude reviews it"]]></title><description><![CDATA[
<p>The migration was not a single, long run of neal (because I was developing/tweaking neal along the way). However I have used it for large-ish implementations and haven't noticed any issues with the reviewer drifting. It is re-prompted each review turn with its adversarial review instructions plus a summary of the run state (completed scopes, open findings, etc). Also, the final review is executed with a fresh context.</p>
]]></description><pubDate>Wed, 05 Aug 2026 15:53:38 +0000</pubDate><link>https://news.ycombinator.com/item?id=49184623</link><dc:creator>navels</dc:creator><comments>https://news.ycombinator.com/item?id=49184623</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49184623</guid></item><item><title><![CDATA[Show HN: Neal – Codex writes the code, Claude reviews it]]></title><description><![CDATA[
<p>neal is a CLI I wrote while trying to use an LLM coding agent (in this case, Codex GPT-5.4) to work autonomously on a migration of our large frontend codebase. A few things came up during that project that guided the development of neal:<p>1. Telling an agent to "keep working unless blocked" doesn't actually work over long stretches of time.<p>2. Large projects should be broken up into smaller chunks of work.<p>3. The coding agent should start each chunk of work with a fresh context to prevent context rot.<p>4. The coding agent benefits from having a different agent doing adversarial reviews along the way.<p>What I ended up with is an orchestrator that<p>- lets you configure planner, coder, and reviewer roles. I started with Codex and Claude using their SDKs but later added OpenRouter as well as a compatibility mode for verifying that a model can handle the orchestration. Currently there are 44 compatible OpenRouter models.<p>- runs a plan that you provide through a planner / reviewer loop which splits the work into reasonable-sized chunks and formats the plan document so that it can be executed by neal.<p>- runs each chunk of work through a coder / read-only reviewer loop. Once both agents are satisfied, neal moves on to the next chunk. Once everything is complete there is a final pass through the coder / reviewer loop to ensure the implementation satisfies the entire plan.<p>- resets the coder's context with each chunk of work to prevent context drift, leaving the reviewer's context long-running.<p>The migration landed (549 commits, over 3000 files added/modified). I also ran 100 benchmark cases (from SWE-bench Pro) that Codex failed to solve on its own through neal and observed 8 solved when using neal with Codex in both (coder and reviewer) roles and 15 solved when using Claude as the reviewer. I had hoped for a more exciting benchmark outcome but this at least demonstrated neal's value, especially putting a different model in the reviewer role.<p>neal is an npm package installed via<p><pre><code>  npm install -g @navels/neal
  </code></pre>
and configured via<p><pre><code>  neal setup
</code></pre>
It works with your Codex and/or Claude Code subscriptions (as long as those CLIs are authenticated locally) or via OpenAI/Anthropic/OpenRouter API keys.<p>I have a more detailed writeup of neal's capabilities and how it came to be on my blog: <a href="https://navels.dev/blog/neal/" rel="nofollow">https://navels.dev/blog/neal/</a><p>With how fast model capabilities are progressing I'm not expecting this tool to be very useful for too much longer, but I'm still using it as my daily driver, with Claude Fable as the planner/coder and Codex Sol as the reviewer. If you try neal I'd love to hear your feedback, especially if you have a use case that doesn't quite work or you want additional models available.</p>
<hr>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=49168496">https://news.ycombinator.com/item?id=49168496</a></p>
<p>Points: 2</p>
<p># Comments: 2</p>
]]></description><pubDate>Tue, 04 Aug 2026 13:14:27 +0000</pubDate><link>https://github.com/navels/neal</link><dc:creator>navels</dc:creator><comments>https://news.ycombinator.com/item?id=49168496</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49168496</guid></item><item><title><![CDATA[New comment by navels in "Show HN: macOS menu bar gauges for your Claude Code quota"]]></title><description><![CDATA[
<p>lol how many of these things are there. I've been using <a href="https://github.com/f-is-h/usage4claude" rel="nofollow">https://github.com/f-is-h/usage4claude</a> which includes Codex usage (I use Claude and Codex).</p>
]]></description><pubDate>Wed, 10 Jun 2026 15:00:25 +0000</pubDate><link>https://news.ycombinator.com/item?id=48477383</link><dc:creator>navels</dc:creator><comments>https://news.ycombinator.com/item?id=48477383</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48477383</guid></item><item><title><![CDATA[New comment by navels in "“Erdos problem #728 was solved more or less autonomously by AI”"]]></title><description><![CDATA[
<p>Other activites such as the sublime pursuit of truth and beauty . . . aka mathematics ;-)</p>
]]></description><pubDate>Sat, 10 Jan 2026 01:25:45 +0000</pubDate><link>https://news.ycombinator.com/item?id=46561732</link><dc:creator>navels</dc:creator><comments>https://news.ycombinator.com/item?id=46561732</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46561732</guid></item><item><title><![CDATA[New comment by navels in "Show HN: A web version of Pips game (NYT domino game)"]]></title><description><![CDATA[
<p>The rule is based on the symbol, not the color. Ie, there is no green or blue rule.</p>
]]></description><pubDate>Tue, 30 Sep 2025 19:50:01 +0000</pubDate><link>https://news.ycombinator.com/item?id=45430372</link><dc:creator>navels</dc:creator><comments>https://news.ycombinator.com/item?id=45430372</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=45430372</guid></item><item><title><![CDATA[New comment by navels in "Vet is a safety net for the risky curl | bash pattern"]]></title><description><![CDATA[
<p>same parent had said "It's crazy that a language that prioritizes security so highly in it's design itself is only compiled through such insecure methods."</p>
]]></description><pubDate>Thu, 24 Jul 2025 14:36:17 +0000</pubDate><link>https://news.ycombinator.com/item?id=44671234</link><dc:creator>navels</dc:creator><comments>https://news.ycombinator.com/item?id=44671234</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=44671234</guid></item><item><title><![CDATA[New comment by navels in "I don't think AGI is right around the corner"]]></title><description><![CDATA[
<p>why not?</p>
]]></description><pubDate>Sun, 06 Jul 2025 21:29:57 +0000</pubDate><link>https://news.ycombinator.com/item?id=44484226</link><dc:creator>navels</dc:creator><comments>https://news.ycombinator.com/item?id=44484226</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=44484226</guid></item><item><title><![CDATA[New comment by navels in "Show HN: Proxmox VE Helper Scripts: Make managing your homelab a breeze"]]></title><description><![CDATA[
<p>How weird, tteck's site is great and what is this nonsense about security.</p>
]]></description><pubDate>Sun, 18 Aug 2024 19:05:17 +0000</pubDate><link>https://news.ycombinator.com/item?id=41284555</link><dc:creator>navels</dc:creator><comments>https://news.ycombinator.com/item?id=41284555</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=41284555</guid></item><item><title><![CDATA[New comment by navels in "Show HN: Proxmox VE Helper Scripts: Make managing your homelab a breeze"]]></title><description><![CDATA[
<p>I'm confused about whether/how this is related to <a href="https://github.com/tteck/Proxmox/">https://github.com/tteck/Proxmox/</a></p>
]]></description><pubDate>Sun, 18 Aug 2024 19:01:29 +0000</pubDate><link>https://news.ycombinator.com/item?id=41284504</link><dc:creator>navels</dc:creator><comments>https://news.ycombinator.com/item?id=41284504</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=41284504</guid></item><item><title><![CDATA[New comment by navels in "Take a look at Traefik, even if you don't use containers"]]></title><description><![CDATA[
<p>I also use NginxProxyManager (8 hosts) and I'm not seeing any replies to your post that would explain why caddyserver or traefik provide any benefit over NPM.</p>
]]></description><pubDate>Sun, 05 May 2024 18:10:00 +0000</pubDate><link>https://news.ycombinator.com/item?id=40266886</link><dc:creator>navels</dc:creator><comments>https://news.ycombinator.com/item?id=40266886</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=40266886</guid></item><item><title><![CDATA[New comment by navels in "How many CPU cores can you use in parallel?"]]></title><description><![CDATA[
<p>Related: fascinating deep dive of the Python GIL by David Beazley from PyCon2010: <a href="https://www.youtube.com/watch?v=Obt-vMVdM8s" rel="nofollow noreferrer">https://www.youtube.com/watch?v=Obt-vMVdM8s</a></p>
]]></description><pubDate>Tue, 19 Dec 2023 19:15:18 +0000</pubDate><link>https://news.ycombinator.com/item?id=38700123</link><dc:creator>navels</dc:creator><comments>https://news.ycombinator.com/item?id=38700123</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=38700123</guid></item><item><title><![CDATA[New comment by navels in "Chrome.userScripts API requires developer mode"]]></title><description><![CDATA[
<p>Pertains to user scripts injected by MV3 extensions. Noticed this when my Tampermonkey extension updated today with this message:<p><i>Enabling developer mode will soon become mandatory for running userscripts via Tampermonkey.</i><p>Not a big deal for me, I have extension dev mode on all the time, but ymmv.<p>Some discussion here: <a href="https://groups.google.com/a/chromium.org/g/chromium-extensions/c/bdnjlMNzQqA/m/BLaC7ONvAQAJ#BLaC7ONvAQAJ" rel="nofollow noreferrer">https://groups.google.com/a/chromium.org/g/chromium-extensio...</a></p>
]]></description><pubDate>Tue, 05 Dec 2023 02:27:48 +0000</pubDate><link>https://news.ycombinator.com/item?id=38526278</link><dc:creator>navels</dc:creator><comments>https://news.ycombinator.com/item?id=38526278</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=38526278</guid></item><item><title><![CDATA[Chrome.userScripts API requires developer mode]]></title><description><![CDATA[
<p>Article URL: <a href="https://developer.chrome.com/docs/extensions/reference/api/userScripts">https://developer.chrome.com/docs/extensions/reference/api/userScripts</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=38526277">https://news.ycombinator.com/item?id=38526277</a></p>
<p>Points: 79</p>
<p># Comments: 22</p>
]]></description><pubDate>Tue, 05 Dec 2023 02:27:48 +0000</pubDate><link>https://developer.chrome.com/docs/extensions/reference/api/userScripts</link><dc:creator>navels</dc:creator><comments>https://news.ycombinator.com/item?id=38526277</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=38526277</guid></item></channel></rss>