<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: re5i5tor</title><link>https://news.ycombinator.com/user?id=re5i5tor</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Sat, 21 Nov 2026 02:16:57 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=re5i5tor" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[New comment by re5i5tor in "Pi Is the Last Harness"]]></title><description><![CDATA[
<p>I have journeyed from Claude Code to OMP to Pi while working very hard to keep a super open mind, open to anything.<p>This from the post is why I landed on Pi<p>"What makes Pi the best fit here? Simple: it’s the only harness today that’s open source, vendor-agnostic, extensible, and minimal enough to not force opinions on you. It gives you full ownership and puts you in control. Being able to change the providers without switching apps alone is reason enough to adopt it, but the fact that it’s also so flexible makes it almost impossible for me to switch back."<p>I am building my own (intentionally simple) subagent orchestration framework inside Pi and am watching Pi Durable closely as I suspect it and Pi will unify before long and subagents are better supported in Durable. (Relatively uninformed opinion in this case, not using Durable yet.)<p>Edit-typo</p>
]]></description><pubDate>Sun, 01 Nov 2026 17:21:56 +0000</pubDate><link>https://news.ycombinator.com/item?id=50327598</link><dc:creator>re5i5tor</dc:creator><comments>https://news.ycombinator.com/item?id=50327598</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50327598</guid></item><item><title><![CDATA[New comment by re5i5tor in "Gemini 3.8 Live and 3.8 Live Extended Thinking"]]></title><description><![CDATA[
<p>“caveats and provisos” makes me think of Robin Williams’ genie imitating William F. Buckley Jr.</p>
]]></description><pubDate>Tue, 15 Sep 2026 20:14:49 +0000</pubDate><link>https://news.ycombinator.com/item?id=49718245</link><dc:creator>re5i5tor</dc:creator><comments>https://news.ycombinator.com/item?id=49718245</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49718245</guid></item><item><title><![CDATA[New comment by re5i5tor in "Qwen3.8 27B scores 52 on Artificial Analysis"]]></title><description><![CDATA[
<p>Have you tried it? I’d recommend doing so, it’s impressive in real use cases.</p>
]]></description><pubDate>Mon, 17 Aug 2026 18:39:23 +0000</pubDate><link>https://news.ycombinator.com/item?id=49335674</link><dc:creator>re5i5tor</dc:creator><comments>https://news.ycombinator.com/item?id=49335674</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49335674</guid></item><item><title><![CDATA[New comment by re5i5tor in "Chrome DevTools MCP (2025)"]]></title><description><![CDATA[
<p>Lots of MCP hate, and some love, in the comments.<p>80% of MCPs are thin wrappers over APIs . Yes they stink.<p>A well written remote OAuth MCP need not stink. Tons of advantages starting with strong security baked in.<p>I like Cloudflare Code Mode as an MCP pattern. Two tools, search and execute.<p>1M Opus 4.6 also reduces the  penalties of MCP’s context approach. Along with tool search etc.</p>
]]></description><pubDate>Mon, 16 Mar 2026 12:42:16 +0000</pubDate><link>https://news.ycombinator.com/item?id=47398199</link><dc:creator>re5i5tor</dc:creator><comments>https://news.ycombinator.com/item?id=47398199</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47398199</guid></item><item><title><![CDATA[New comment by re5i5tor in "Chrome DevTools MCP (2025)"]]></title><description><![CDATA[
<p>And people forget, Claude Code isn’t the only Claude surface, and CLIs don’t help in other surfaces other than Cowork.</p>
]]></description><pubDate>Mon, 16 Mar 2026 12:28:05 +0000</pubDate><link>https://news.ycombinator.com/item?id=47398054</link><dc:creator>re5i5tor</dc:creator><comments>https://news.ycombinator.com/item?id=47398054</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47398054</guid></item><item><title><![CDATA[New comment by re5i5tor in "Yann LeCun raises $1B to build AI that understands the physical world"]]></title><description><![CDATA[
<p>Does anyone else see an echo of Severance (Apple TV series) here?</p>
]]></description><pubDate>Wed, 11 Mar 2026 13:12:42 +0000</pubDate><link>https://news.ycombinator.com/item?id=47335131</link><dc:creator>re5i5tor</dc:creator><comments>https://news.ycombinator.com/item?id=47335131</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47335131</guid></item><item><title><![CDATA[New comment by re5i5tor in "MCP server that reduces Claude Code context consumption by 98%"]]></title><description><![CDATA[
<p>See my comment above, context-mode has no way to inject itself into the MCP tool-call - response loop.<p>Still high-value, outside MCPs.</p>
]]></description><pubDate>Sun, 01 Mar 2026 01:26:40 +0000</pubDate><link>https://news.ycombinator.com/item?id=47202628</link><dc:creator>re5i5tor</dc:creator><comments>https://news.ycombinator.com/item?id=47202628</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47202628</guid></item><item><title><![CDATA[New comment by re5i5tor in "MCP server that reduces Claude Code context consumption by 98%"]]></title><description><![CDATA[
<p>I dug into this further. Tested empirically and read the code.<p>Confirmed: context-mode cannot intercept MCP tool responses. The PreToolUse hook (hooks/pretooluse.sh) matches only Bash|Read|Grep|Glob|WebFetch|WebSearch|Task. When I called my obsidian MCP's obsidian_list via MCP, the response went straight into context — zero entries in context-mode's FTS5 database. The web fetches from the same session were all indexed.<p>The context-mode skill (SKILL.md) actually acknowledges this at lines 71-77 with an "after-the-fact" decision tree for MCP output: if it's already in context, use it directly; if you need to search it again, save to file then index. But that's damage control — the context is already consumed. You can't un-eat those tokens.<p>The architectural reason: MCP tool responses flow via JSON-RPC directly to the model. There's no PostToolUse hook in Claude Code that could modify or compress a response before it enters context. And you can't call MCP tools from inside a subprocess, so the "run it in a sandbox" pattern doesn't apply.<p>So the 98% savings are real but scoped to built-in tools and CLI wrappers (curl, gh, kubectl, etc.) — anything replicable in a subprocess. For third-party MCP tools with unique capabilities (Excalidraw rendering, calendar APIs, Obsidian vault access), the MCP author has to apply context-mode's concepts server-side: return compact summaries, store full output queryably, expose drill-down tools. Which is essentially what you suggested above.<p>Still very high value for the built-in tool side. Just want the boundary to be clear.<p>Correct any misconceptions please!</p>
]]></description><pubDate>Sun, 01 Mar 2026 01:25:08 +0000</pubDate><link>https://news.ycombinator.com/item?id=47202616</link><dc:creator>re5i5tor</dc:creator><comments>https://news.ycombinator.com/item?id=47202616</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47202616</guid></item><item><title><![CDATA[New comment by re5i5tor in "MCP server that reduces Claude Code context consumption by 98%"]]></title><description><![CDATA[
<p>I'm not sure it actually works with MCPs *at all*, trying to get that clarified. How can context-mode get "into the MCP loop"?</p>
]]></description><pubDate>Sun, 01 Mar 2026 01:00:30 +0000</pubDate><link>https://news.ycombinator.com/item?id=47202437</link><dc:creator>re5i5tor</dc:creator><comments>https://news.ycombinator.com/item?id=47202437</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47202437</guid></item><item><title><![CDATA[New comment by re5i5tor in "MCP server that reduces Claude Code context consumption by 98%"]]></title><description><![CDATA[
<p>Hmmm. I was talking about the output side. When data comes back from an MCP tool call, context-mode is still not in the loop, not able to help, is it?<p>Edit: clarify "MCP tool"</p>
]]></description><pubDate>Sat, 28 Feb 2026 23:42:41 +0000</pubDate><link>https://news.ycombinator.com/item?id=47201789</link><dc:creator>re5i5tor</dc:creator><comments>https://news.ycombinator.com/item?id=47201789</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47201789</guid></item><item><title><![CDATA[New comment by re5i5tor in "MCP server that reduces Claude Code context consumption by 98%"]]></title><description><![CDATA[
<p>Really intrigued and def will try, thanks for this.<p>In connecting the dots (and help me make sure I'm connecting them correctly), context-mode _does not address MCP context usage at all_, correct? You are instead suggesting we refactor or eliminate MCP tools, or apply concepts similar to context_mode in our MCPs where possible?<p>Context-mode is still very high value, even if the answer is "no," just want to make sure I understand. Also interested in your thoughts about the above.<p>I write a number of MCPs that work across all Claude surfaces; so the usual "CLI!" isn't as viable an answer (though with code execution it sometimes can be) ...<p>Edit: typo</p>
]]></description><pubDate>Sat, 28 Feb 2026 17:16:48 +0000</pubDate><link>https://news.ycombinator.com/item?id=47197779</link><dc:creator>re5i5tor</dc:creator><comments>https://news.ycombinator.com/item?id=47197779</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47197779</guid></item><item><title><![CDATA[New comment by re5i5tor in "I’m joining OpenAI"]]></title><description><![CDATA[
<p>I shared a lot of these uninformed (speaking only for myself) opinions, but have a lot more respect for Peter after listening to his recent 3hr interview with Lex Fridman. I came away liking and respecting.</p>
]]></description><pubDate>Tue, 17 Feb 2026 02:19:46 +0000</pubDate><link>https://news.ycombinator.com/item?id=47042944</link><dc:creator>re5i5tor</dc:creator><comments>https://news.ycombinator.com/item?id=47042944</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47042944</guid></item><item><title><![CDATA[New comment by re5i5tor in "Monosketch"]]></title><description><![CDATA[
<p>Slightly off topic but I am writing an Excalidraw MCP that allows Claude Code and Claude.ai to create Excal drawings and then iterate on them — I gave Claude the ability to “see” the resulting drawing via a tool that runs a rendering pipeline and returns a png to the model. It’s producing the Olympic logo pretty nicely ;-)</p>
]]></description><pubDate>Fri, 13 Feb 2026 14:01:57 +0000</pubDate><link>https://news.ycombinator.com/item?id=47002802</link><dc:creator>re5i5tor</dc:creator><comments>https://news.ycombinator.com/item?id=47002802</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47002802</guid></item><item><title><![CDATA[New comment by re5i5tor in "TIL: Apple Broke Time Machine Again on Tahoe"]]></title><description><![CDATA[
<p>Exactly right.</p>
]]></description><pubDate>Mon, 02 Feb 2026 05:54:28 +0000</pubDate><link>https://news.ycombinator.com/item?id=46852876</link><dc:creator>re5i5tor</dc:creator><comments>https://news.ycombinator.com/item?id=46852876</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46852876</guid></item><item><title><![CDATA[New comment by re5i5tor in "TIL: Apple Broke Time Machine Again on Tahoe"]]></title><description><![CDATA[
<p>I strongly recommend against Time Machine to NAS — just mentioned it in an article today as “works great, until it inevitably and catastrophically fails”.<p>Reliable for me is Kopia from Mac to S3 compatible volume (minio) on Synology.</p>
]]></description><pubDate>Mon, 02 Feb 2026 05:52:47 +0000</pubDate><link>https://news.ycombinator.com/item?id=46852871</link><dc:creator>re5i5tor</dc:creator><comments>https://news.ycombinator.com/item?id=46852871</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46852871</guid></item><item><title><![CDATA[New comment by re5i5tor in "Cowork: Claude Code for the rest of your work"]]></title><description><![CDATA[
<p>Using any plugins or skills for that work?</p>
]]></description><pubDate>Sat, 17 Jan 2026 23:09:11 +0000</pubDate><link>https://news.ycombinator.com/item?id=46663053</link><dc:creator>re5i5tor</dc:creator><comments>https://news.ycombinator.com/item?id=46663053</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46663053</guid></item><item><title><![CDATA[New comment by re5i5tor in "Cloudflare acquires Astro"]]></title><description><![CDATA[
<p>Based on a request presumably from a HN viewer, I made the repo [2] for this site  public.<p>[2] <a href="https://github.com/cto4ai/cto4ai-blog" rel="nofollow">https://github.com/cto4ai/cto4ai-blog</a></p>
]]></description><pubDate>Sat, 17 Jan 2026 16:51:36 +0000</pubDate><link>https://news.ycombinator.com/item?id=46659527</link><dc:creator>re5i5tor</dc:creator><comments>https://news.ycombinator.com/item?id=46659527</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46659527</guid></item><item><title><![CDATA[New comment by re5i5tor in "Cloudflare acquires Astro"]]></title><description><![CDATA[
<p>I use Astro deployed on Cloudflare, blog-newsletter kind of site [1], moved over from Hugo. If this keeps Astro viable then it seems like a net win.<p>[1] <a href="https://cto4.ai" rel="nofollow">https://cto4.ai</a></p>
]]></description><pubDate>Fri, 16 Jan 2026 15:05:20 +0000</pubDate><link>https://news.ycombinator.com/item?id=46647071</link><dc:creator>re5i5tor</dc:creator><comments>https://news.ycombinator.com/item?id=46647071</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46647071</guid></item><item><title><![CDATA[New comment by re5i5tor in "Cloudflare acquires Astro"]]></title><description><![CDATA[
<p>If you aren’t realizing the importance of POSSE imho that’s crazy ;-)</p>
]]></description><pubDate>Fri, 16 Jan 2026 14:59:56 +0000</pubDate><link>https://news.ycombinator.com/item?id=46647005</link><dc:creator>re5i5tor</dc:creator><comments>https://news.ycombinator.com/item?id=46647005</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46647005</guid></item><item><title><![CDATA[New comment by re5i5tor in "Cowork: Claude Code for the rest of your work"]]></title><description><![CDATA[
<p>100% confirm Opus 4.5 is very image smart.</p>
]]></description><pubDate>Mon, 12 Jan 2026 23:50:18 +0000</pubDate><link>https://news.ycombinator.com/item?id=46595732</link><dc:creator>re5i5tor</dc:creator><comments>https://news.ycombinator.com/item?id=46595732</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46595732</guid></item></channel></rss>