<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"
	xmlns:content="http://purl.org/rss/1.0/modules/content/"
	xmlns:wfw="http://wellformedweb.org/CommentAPI/"
	xmlns:dc="http://purl.org/dc/elements/1.1/"
	xmlns:atom="http://www.w3.org/2005/Atom"
	xmlns:sy="http://purl.org/rss/1.0/modules/syndication/"
	xmlns:slash="http://purl.org/rss/1.0/modules/slash/"
	>

<channel>
	<title>Gea-Suan Lin&#039;s BLOG</title>
	<atom:link href="https://blog.gslin.org/feed/" rel="self" type="application/rss+xml" />
	<link>https://blog.gslin.org</link>
	<description>幹壞事是進步最大的原動力</description>
	<lastBuildDate>Tue, 22 Sep 2026 18:52:44 +0000</lastBuildDate>
	<language>en-US</language>
	<sy:updatePeriod>
	hourly	</sy:updatePeriod>
	<sy:updateFrequency>
	1	</sy:updateFrequency>
	<generator>https://wordpress.org/?v=7.1.2</generator>
<atom:link rel="hub" href="https://pubsubhubbub.appspot.com"/>
<atom:link rel="hub" href="https://pubsubhubbub.superfeedr.com"/>
<atom:link rel="hub" href="https://websubhub.com/hub"/>
<atom:link rel="self" href="https://blog.gslin.org/feed/"/>
<site xmlns="com-wordpress:feed-additions:1">21326247</site>	<item>
		<title>WordPress 突然冒出來的 RCE 問題</title>
		<link>https://blog.gslin.org/archives/2026/09/23/13223/wordpress-%e7%aa%81%e7%84%b6%e5%86%92%e5%87%ba%e4%be%86%e7%9a%84-rce-%e5%95%8f%e9%a1%8c/</link>
					<comments>https://blog.gslin.org/archives/2026/09/23/13223/wordpress-%e7%aa%81%e7%84%b6%e5%86%92%e5%87%ba%e4%be%86%e7%9a%84-rce-%e5%95%8f%e9%a1%8c/#respond</comments>
		
		<dc:creator><![CDATA[Gea-Suan Lin]]></dc:creator>
		<pubDate>Tue, 22 Sep 2026 18:52:44 +0000</pubDate>
				<category><![CDATA[Blog]]></category>
		<category><![CDATA[Computer]]></category>
		<category><![CDATA[Murmuring]]></category>
		<category><![CDATA[Network]]></category>
		<category><![CDATA[Security]]></category>
		<category><![CDATA[Software]]></category>
		<category><![CDATA[advisory]]></category>
		<category><![CDATA[blog]]></category>
		<category><![CDATA[code]]></category>
		<category><![CDATA[execution]]></category>
		<category><![CDATA[rce]]></category>
		<category><![CDATA[remote]]></category>
		<category><![CDATA[security]]></category>
		<category><![CDATA[wordpress]]></category>
		<guid isPermaLink="false">https://blog.gslin.org/?p=13223</guid>

					<description><![CDATA[突然看到「Unauthenticated path traversal in page-template resolution leading to conditional RCE (via)」這包，居然是可以指定讀檔 RCE： An unauthenticated attacker can make get_page_template() page-template resolution include a chosen readable local .php file outside the active theme directories. If relevant pre-conditions for both the server environment and the active theme are met, this can lead to RCE. 是有 pre-condition 不是隨意檔案沒錯，但還是很... 而且 &#8230; <a href="https://blog.gslin.org/archives/2026/09/23/13223/wordpress-%e7%aa%81%e7%84%b6%e5%86%92%e5%87%ba%e4%be%86%e7%9a%84-rce-%e5%95%8f%e9%a1%8c/" class="more-link">Continue reading<span class="screen-reader-text"> "WordPress 突然冒出來的 RCE 問題"</span></a>]]></description>
										<content:encoded><![CDATA[<p>突然看到「<a href="https://github.com/WordPress/wordpress-develop/security/advisories/GHSA-7hp8-65ch-5whp">Unauthenticated path traversal in page-template resolution leading to conditional RCE</a> (<a href="https://news.ycombinator.com/item?id=49803959">via</a>)」這包，居然是可以指定讀檔 <a href="https://en.wikipedia.org/wiki/Remote_code_execution">RCE</a>：</p>
<blockquote><p>An unauthenticated attacker can make <code>get_page_template()</code> page-template resolution include a chosen readable local .php file outside the active theme directories. If relevant pre-conditions for both the server environment and the active theme are met, this can lead to RCE.</p></blockquote>
<p>是有 pre-condition 不是隨意檔案沒錯，但還是很... 而且 patch 是 2016 年的 4.7 到現在的版本，應該還是因為 4.7 是目前還有在 security maintenance 的關係，更早應該也有？</p>
<p>發現的研究人員 Robert Ressl 的網站 <a href="https://ressl.ch/">https://ressl.ch/</a> 看起來是用 AI 來挖資安研究員，而且還有自己弄 uncensored model 放到 <a href="https://en.wikipedia.org/wiki/Hugging_Face">Hugging Face</a> 上讓大家用，不確定這次的 issue 是不是也是掃出來的？</p>
]]></content:encoded>
					
					<wfw:commentRss>https://blog.gslin.org/archives/2026/09/23/13223/wordpress-%e7%aa%81%e7%84%b6%e5%86%92%e5%87%ba%e4%be%86%e7%9a%84-rce-%e5%95%8f%e9%a1%8c/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
		<post-id xmlns="com-wordpress:feed-additions:1">13223</post-id>	</item>
		<item>
		<title>OpenAI 也推出了 GPT-6 Sol 與 Luna：也跟著降價了</title>
		<link>https://blog.gslin.org/archives/2026/09/23/13222/openai-%e4%b9%9f%e6%8e%a8%e5%87%ba%e4%ba%86-gpt-6-sol-%e8%88%87-luna%ef%bc%9a%e4%b9%9f%e8%b7%9f%e8%91%97%e9%99%8d%e5%83%b9%e4%ba%86/</link>
					<comments>https://blog.gslin.org/archives/2026/09/23/13222/openai-%e4%b9%9f%e6%8e%a8%e5%87%ba%e4%ba%86-gpt-6-sol-%e8%88%87-luna%ef%bc%9a%e4%b9%9f%e8%b7%9f%e8%91%97%e9%99%8d%e5%83%b9%e4%ba%86/#respond</comments>
		
		<dc:creator><![CDATA[Gea-Suan Lin]]></dc:creator>
		<pubDate>Tue, 22 Sep 2026 18:42:00 +0000</pubDate>
				<category><![CDATA[API]]></category>
		<category><![CDATA[Computer]]></category>
		<category><![CDATA[Murmuring]]></category>
		<category><![CDATA[Network]]></category>
		<category><![CDATA[Service]]></category>
		<category><![CDATA[ai]]></category>
		<category><![CDATA[cost]]></category>
		<category><![CDATA[gpt]]></category>
		<category><![CDATA[gpt-6]]></category>
		<category><![CDATA[language]]></category>
		<category><![CDATA[large]]></category>
		<category><![CDATA[llm]]></category>
		<category><![CDATA[luna]]></category>
		<category><![CDATA[model]]></category>
		<category><![CDATA[openai]]></category>
		<category><![CDATA[price]]></category>
		<category><![CDATA[pricing]]></category>
		<category><![CDATA[sol]]></category>
		<guid isPermaLink="false">https://blog.gslin.org/?p=13222</guid>

					<description><![CDATA[Anthropic 的 Opus 5.5 推出沒多久 (參考上一篇「Claude Opus 5.5，殺 Fable 5.1 + 降價...」)，OpenAI 這邊也補上 GPT-6 的另外兩條產品線了：「Introducing GPT‑6 Sol and Luna」。 文章的第一個主題就是價錢，比 GPT-5.6 的 Sol 與 Luna 產品線便宜 50% 以上： 其他的就是比 GPT-5.6 的 Sol &#38; Luna 好的一些說明，不過這次就沒有 Terra 了？ 這樣好像能理解 SpaceXAI 那邊的 Grok 4.7 延期發表情況了... 多撐了兩個禮拜但還是弄不出比較像樣的東西？ 然後依照慣例準備要 reset 了：(https://x.com/thsottiaux/status/2102463847714247142) 不過這最近大家都丟新東西出來，下一輪應該就會要等比較久了，大概十月底？]]></description>
										<content:encoded><![CDATA[<p><a href="https://en.wikipedia.org/wiki/Anthropic">Anthropic</a> 的 Opus 5.5 推出沒多久 (參考上一篇「<a href="https://blog.gslin.org/archives/2026/09/23/13221/claude-opus-5-5%ef%bc%8c%e6%ae%ba-fable-5-1-%e9%99%8d%e5%83%b9/">Claude Opus 5.5，殺 Fable 5.1 + 降價...</a>」)，<a href="https://en.wikipedia.org/wiki/OpenAI">OpenAI</a> 這邊也補上 GPT-6 的另外兩條產品線了：「<a href="https://openai.com/index/introducing-gpt-6-sol-and-luna/">Introducing GPT‑6 Sol and Luna</a>」。</p>
<p>文章的第一個主題就是價錢，比 GPT-5.6 的 Sol 與 Luna 產品線便宜 50% 以上：</p>
<picture><source type="image/webp" srcset="https://i.gslin.com/s/1790102053-88b42f71.webp" /><img decoding="async" src="https://i.gslin.com/s/1790102053-88b42f71.png" alt="" /></picture>
<p>其他的就是比 GPT-5.6 的 Sol &amp; Luna 好的一些說明，不過這次就沒有 Terra 了？</p>
<p>這樣好像能理解 <a href="https://en.wikipedia.org/wiki/SpaceXAI">SpaceXAI</a> 那邊的 Grok 4.7 延期發表情況了... 多撐了兩個禮拜但還是弄不出比較像樣的東西？</p>
<p>然後依照慣例準備要 reset 了：(<a href="https://x.com/thsottiaux/status/2102463847714247142">https://x.com/thsottiaux/status/2102463847714247142</a>)</p>
<picture><source type="image/webp" srcset="https://i.gslin.com/s/1790102416-83dc64c6.webp" /><img decoding="async" src="https://i.gslin.com/s/1790102416-83dc64c6.png" alt="" /></picture>
<p>不過這最近大家都丟新東西出來，下一輪應該就會要等比較久了，大概十月底？</p>
]]></content:encoded>
					
					<wfw:commentRss>https://blog.gslin.org/archives/2026/09/23/13222/openai-%e4%b9%9f%e6%8e%a8%e5%87%ba%e4%ba%86-gpt-6-sol-%e8%88%87-luna%ef%bc%9a%e4%b9%9f%e8%b7%9f%e8%91%97%e9%99%8d%e5%83%b9%e4%ba%86/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
		<post-id xmlns="com-wordpress:feed-additions:1">13222</post-id>	</item>
		<item>
		<title>Claude Opus 5.5，殺 Fable 5.1 + 降價...</title>
		<link>https://blog.gslin.org/archives/2026/09/23/13221/claude-opus-5-5%ef%bc%8c%e6%ae%ba-fable-5-1-%e9%99%8d%e5%83%b9/</link>
					<comments>https://blog.gslin.org/archives/2026/09/23/13221/claude-opus-5-5%ef%bc%8c%e6%ae%ba-fable-5-1-%e9%99%8d%e5%83%b9/#comments</comments>
		
		<dc:creator><![CDATA[Gea-Suan Lin]]></dc:creator>
		<pubDate>Tue, 22 Sep 2026 18:24:21 +0000</pubDate>
				<category><![CDATA[API]]></category>
		<category><![CDATA[Computer]]></category>
		<category><![CDATA[Murmuring]]></category>
		<category><![CDATA[Network]]></category>
		<category><![CDATA[Programming]]></category>
		<category><![CDATA[Service]]></category>
		<category><![CDATA[ai]]></category>
		<category><![CDATA[anthropic]]></category>
		<category><![CDATA[claude]]></category>
		<category><![CDATA[cost]]></category>
		<category><![CDATA[fable]]></category>
		<category><![CDATA[language]]></category>
		<category><![CDATA[large]]></category>
		<category><![CDATA[llm]]></category>
		<category><![CDATA[model]]></category>
		<category><![CDATA[opus]]></category>
		<category><![CDATA[pricing]]></category>
		<guid isPermaLink="false">https://blog.gslin.org/?p=13221</guid>

					<description><![CDATA[如同最近傳言的，Anthropic 在準備 Opus 5.5：「Claude Opus 5.5」，Hacker News 上已經有蠻多討論了，像是 max effort 的 benchmark：「Claude Opus 5.5 (Adaptive Reasoning, Max Effort, Default Fallback) Intelligence, Performance &#38; Price Analysis (via)」，以及 xhigh effort 的 benchmark：「Claude Opus 5.5 (Adaptive Reasoning, Xhigh Effort, Default Fallback) Intelligence, Performance &#38; Price Analysis」。 先從官方的 benchmark 表格來看，居然是殺 Claude Fable 5.1： 不過下面的 footprint 有點玄機，看起來 Anthropic 這邊是拿 max &#8230; <a href="https://blog.gslin.org/archives/2026/09/23/13221/claude-opus-5-5%ef%bc%8c%e6%ae%ba-fable-5-1-%e9%99%8d%e5%83%b9/" class="more-link">Continue reading<span class="screen-reader-text"> "Claude Opus 5.5，殺 Fable 5.1 + 降價..."</span></a>]]></description>
										<content:encoded><![CDATA[<p>如同最近傳言的，<a href="https://en.wikipedia.org/wiki/Anthropic">Anthropic</a> 在準備 Opus 5.5：「<a href="https://www.anthropic.com/claude-opus-5-5">Claude Opus 5.5</a>」，Hacker News 上已經有蠻多討論了，像是 max effort 的 benchmark：「<a href="https://artificialanalysis.ai/models/claude-opus-5-5">Claude Opus 5.5 (Adaptive Reasoning, Max Effort, Default Fallback) Intelligence, Performance &amp; Price Analysis</a> (<a href="https://news.ycombinator.com/item?id=49804316">via</a>)」，以及 xhigh effort 的 benchmark：「<a href="https://artificialanalysis.ai/models/claude-opus-5-5-xhigh">Claude Opus 5.5 (Adaptive Reasoning, Xhigh Effort, Default Fallback) Intelligence, Performance &amp; Price Analysis</a>」。</p>
<p>先從官方的 benchmark 表格來看，居然是殺 Claude Fable 5.1：</p>
<picture><source type="image/webp" srcset="https://i.gslin.com/s/1790100514-f0bf1e89.webp" /><img decoding="async" src="https://i.gslin.com/s/1790100514-f0bf1e89.png" alt="" /></picture>
<p>不過下面的 footprint 有點玄機，看起來 Anthropic 這邊是拿 max effort，但其他家不是，但交叉看一下 <a href="https://artificialanalysis.ai/">Artificial Analysis</a> 的資料，xhigh effort 的的確是領先...</p>
<p>另外是降價降的不少，標準的 I/O tokens 與 cache write 先來 20% off，然後 cache read 這邊降了 60%，這對 coding agent 應用來說會有蠻大的改進：</p>
<picture><source type="image/webp" srcset="https://i.gslin.com/s/1790100949-d12a4868.webp" /><img decoding="async" src="https://i.gslin.com/s/1790100949-d12a4868.png" alt="" /></picture>
<p>另外看了一下 <a href="https://claude.ai/new#settings/usage">Usage</a> 頁面，給了一張 reset 券：</p>
<picture><source type="image/webp" srcset="https://i.gslin.com/s/1790101117-493b0462.webp" /><img decoding="async" src="https://i.gslin.com/s/1790101117-493b0462.png" alt="" /></picture>
<p>剛好這個連休可以拿來做 side project？</p>
]]></content:encoded>
					
					<wfw:commentRss>https://blog.gslin.org/archives/2026/09/23/13221/claude-opus-5-5%ef%bc%8c%e6%ae%ba-fable-5-1-%e9%99%8d%e5%83%b9/feed/</wfw:commentRss>
			<slash:comments>1</slash:comments>
		
		
		<post-id xmlns="com-wordpress:feed-additions:1">13221</post-id>	</item>
		<item>
		<title>Fable 5 品質下降的問題</title>
		<link>https://blog.gslin.org/archives/2026/09/22/13220/fable-5-%e5%93%81%e8%b3%aa%e4%b8%8b%e9%99%8d%e7%9a%84%e5%95%8f%e9%a1%8c/</link>
					<comments>https://blog.gslin.org/archives/2026/09/22/13220/fable-5-%e5%93%81%e8%b3%aa%e4%b8%8b%e9%99%8d%e7%9a%84%e5%95%8f%e9%a1%8c/#respond</comments>
		
		<dc:creator><![CDATA[Gea-Suan Lin]]></dc:creator>
		<pubDate>Mon, 21 Sep 2026 18:50:22 +0000</pubDate>
				<category><![CDATA[API]]></category>
		<category><![CDATA[Computer]]></category>
		<category><![CDATA[Murmuring]]></category>
		<category><![CDATA[Network]]></category>
		<category><![CDATA[Service]]></category>
		<category><![CDATA[ai]]></category>
		<category><![CDATA[anthropic]]></category>
		<category><![CDATA[degrade]]></category>
		<category><![CDATA[fable]]></category>
		<category><![CDATA[language]]></category>
		<category><![CDATA[large]]></category>
		<category><![CDATA[llm]]></category>
		<category><![CDATA[model]]></category>
		<category><![CDATA[performance]]></category>
		<category><![CDATA[service]]></category>
		<guid isPermaLink="false">https://blog.gslin.org/?p=13220</guid>

					<description><![CDATA[在 https://x.com/Lon/status/2101793422487204027 這邊看到 Anthropic 的 Fable 5 的品質下降的問題 (via)，作者是用 subscription 版本測試： 不確定 Fable 5.1 的情況，先記錄一下...]]></description>
										<content:encoded><![CDATA[<p>在 <a href="https://x.com/Lon/status/2101793422487204027">https://x.com/Lon/status/2101793422487204027</a> 這邊看到 <a href="https://en.wikipedia.org/wiki/Anthropic">Anthropic</a> 的 Fable 5 的品質下降的問題 (<a href="https://news.ycombinator.com/item?id=49789224">via</a>)，作者是用 subscription 版本測試：</p>
<picture><source type="image/webp" srcset="https://i.gslin.com/s/1790016407-24307b2b.webp" /><img decoding="async" src="https://i.gslin.com/s/1790016407-24307b2b.png" alt="" /></picture>
<p>不確定 Fable 5.1 的情況，先記錄一下...</p>
]]></content:encoded>
					
					<wfw:commentRss>https://blog.gslin.org/archives/2026/09/22/13220/fable-5-%e5%93%81%e8%b3%aa%e4%b8%8b%e9%99%8d%e7%9a%84%e5%95%8f%e9%a1%8c/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
		<post-id xmlns="com-wordpress:feed-additions:1">13220</post-id>	</item>
		<item>
		<title>Kagi 的搜尋數量也大幅降低了...</title>
		<link>https://blog.gslin.org/archives/2026/09/22/13218/kagi-%e7%9a%84%e6%90%9c%e5%b0%8b%e6%95%b8%e9%87%8f%e4%b9%9f%e5%a4%a7%e5%b9%85%e9%99%8d%e4%bd%8e%e4%ba%86/</link>
					<comments>https://blog.gslin.org/archives/2026/09/22/13218/kagi-%e7%9a%84%e6%90%9c%e5%b0%8b%e6%95%b8%e9%87%8f%e4%b9%9f%e5%a4%a7%e5%b9%85%e9%99%8d%e4%bd%8e%e4%ba%86/#respond</comments>
		
		<dc:creator><![CDATA[Gea-Suan Lin]]></dc:creator>
		<pubDate>Mon, 21 Sep 2026 18:44:59 +0000</pubDate>
				<category><![CDATA[Computer]]></category>
		<category><![CDATA[Murmuring]]></category>
		<category><![CDATA[Network]]></category>
		<category><![CDATA[Search Engine]]></category>
		<category><![CDATA[Service]]></category>
		<category><![CDATA[WWW]]></category>
		<category><![CDATA[engine]]></category>
		<category><![CDATA[kagi]]></category>
		<category><![CDATA[query]]></category>
		<category><![CDATA[search]]></category>
		<category><![CDATA[service]]></category>
		<guid isPermaLink="false">https://blog.gslin.org/?p=13218</guid>

					<description><![CDATA[Mastodon 上面看到 Kagi 使用量降低很多的情況：(https://mastodon.social/@Goston/117290817396924941) 我看了一下我自己的情況也是類似，之前甚至有到 2000 的，現在已經掉到 600 上下了： 不過看了一下用戶數，看起來還是有慢慢的在上升，現在到 75k 左右了？算了一下 YoY 大約 45%～50%？ 可以參考去年六月底寫的「Kagi 訂閱超過 50k 用戶」，huh？]]></description>
										<content:encoded><![CDATA[<p><a href="https://en.wikipedia.org/wiki/Mastodon_(social_network)">Mastodon</a> 上面看到 <a href="https://en.wikipedia.org/wiki/Kagi">Kagi</a> 使用量降低很多的情況：(<a href="https://mastodon.social/@Goston/117290817396924941">https://mastodon.social/@Goston/117290817396924941</a>)</p>
<picture><source type="image/webp" srcset="https://i.gslin.com/s/1790015744-78ea1aea.webp" /><img decoding="async" src="https://i.gslin.com/s/1790015744-78ea1aea.png" alt="" /></picture>
<p>我看了一下我自己的情況也是類似，之前甚至有到 2000 的，現在已經掉到 600 上下了：</p>
<picture><source type="image/webp" srcset="https://i.gslin.com/s/1790015890-96ff3a73.webp" /><img decoding="async" src="https://i.gslin.com/s/1790015890-96ff3a73.png" alt="" /></picture>
<p>不過看了一下用戶數，看起來還是有慢慢的在上升，現在到 75k 左右了？算了一下 YoY 大約 45%～50%？</p>
<picture><source type="image/webp" srcset="https://i.gslin.com/s/1790016053-1f673cc2.png" /><img decoding="async" src="https://i.gslin.com/s/1790016053-1f673cc2.png" alt="" /></picture>
<p>可以參考去年六月底寫的「<a href="https://blog.gslin.org/archives/2025/06/09/12445/kagi-%e8%a8%82%e9%96%b1%e8%b6%85%e9%81%8e-50k-%e7%94%a8%e6%88%b6/">Kagi 訂閱超過 50k 用戶</a>」，huh？</p>
]]></content:encoded>
					
					<wfw:commentRss>https://blog.gslin.org/archives/2026/09/22/13218/kagi-%e7%9a%84%e6%90%9c%e5%b0%8b%e6%95%b8%e9%87%8f%e4%b9%9f%e5%a4%a7%e5%b9%85%e9%99%8d%e4%bd%8e%e4%ba%86/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
		<post-id xmlns="com-wordpress:feed-additions:1">13218</post-id>	</item>
		<item>
		<title>Grok 4.7 出了</title>
		<link>https://blog.gslin.org/archives/2026/09/22/13217/grok-4-7-%e5%87%ba%e4%ba%86/</link>
					<comments>https://blog.gslin.org/archives/2026/09/22/13217/grok-4-7-%e5%87%ba%e4%ba%86/#respond</comments>
		
		<dc:creator><![CDATA[Gea-Suan Lin]]></dc:creator>
		<pubDate>Mon, 21 Sep 2026 18:02:10 +0000</pubDate>
				<category><![CDATA[API]]></category>
		<category><![CDATA[Computer]]></category>
		<category><![CDATA[Murmuring]]></category>
		<category><![CDATA[Network]]></category>
		<category><![CDATA[Service]]></category>
		<category><![CDATA[ai]]></category>
		<category><![CDATA[grok]]></category>
		<category><![CDATA[language]]></category>
		<category><![CDATA[large]]></category>
		<category><![CDATA[llm]]></category>
		<category><![CDATA[model]]></category>
		<category><![CDATA[spacexai]]></category>
		<category><![CDATA[xai]]></category>
		<guid isPermaLink="false">https://blog.gslin.org/?p=13217</guid>

					<description><![CDATA[在 grok 的 cli 裡面看到 Grok 的消息：「4.7 Introducing Grok 4.7」，另外翻了一下 Hacker News 上也有些討論：「Grok 4.7 (x.ai)」。 跟自家的 Grok 4.6 比可以預期有進步，但光是從官方的 benchmark 看起來，應該是不會到太大，首先是這張表格拿 Grok 4.7 的 xhigh 對 Grok 4.6 的 high 比，但 Grok 4.6 是有 xhigh 的： 所以就當作個簡單的升級吧，價錢還是一樣。不過 Hacker News 上面的 comment 提到一些有趣的小道消息，在 id=49789234 這邊提到不少有趣的小道消息： Apparently Grok 4.7 has 40% more weights than Grok 4.6, but &#8230; <a href="https://blog.gslin.org/archives/2026/09/22/13217/grok-4-7-%e5%87%ba%e4%ba%86/" class="more-link">Continue reading<span class="screen-reader-text"> "Grok 4.7 出了"</span></a>]]></description>
										<content:encoded><![CDATA[<p>在 grok 的 cli 裡面看到 <a href="https://en.wikipedia.org/wiki/Grok_(chatbot)">Grok</a> 的消息：「4.7 <a href="https://x.ai/news/grok-4-7">Introducing Grok 4.7</a>」，另外翻了一下 <a href="https://en.wikipedia.org/wiki/Hacker_News">Hacker News</a> 上也有些討論：「<a href="https://news.ycombinator.com/item?id=49788838">Grok 4.7 (x.ai)</a>」。</p>
<p>跟自家的 Grok 4.6 比可以預期有進步，但光是從官方的 benchmark 看起來，應該是不會到太大，首先是這張表格拿 Grok 4.7 的 xhigh 對 Grok 4.6 的 high 比，但 Grok 4.6 是有 xhigh 的：</p>
<picture><source type="image/webp" srcset="https://i.gslin.com/s/1790012951-0216e830.webp" /><img decoding="async" src="https://i.gslin.com/s/1790012951-0216e830.png" alt="" /></picture>
<p>所以就當作個簡單的升級吧，價錢還是一樣。不過 Hacker News 上面的 comment 提到一些有趣的小道消息，在 <a href="https://news.ycombinator.com/item?id=49789234">id=49789234</a> 這邊提到不少有趣的小道消息：</p>
<blockquote><p>Apparently Grok 4.7 has 40% more weights than Grok 4.6, but the price ($6 output token, $2 input) is the same.</p>
<p>Given that the decrease in their margin and the fact they delayed the release of Grok 4.7 almost two weeks past the original date, XAI must not have been happy with the results for 4.7. And XAI also waited the day before Opus 5.5 is rumored to launch. I imagine Opus 5.5 will blow Grok 4.7 out of the water benchmark wise.</p></blockquote>
<p>首先是 Grok 4.7 的參數量比 Grok 4.6 大 40%？看起來是為了競爭力決定維持原來的價錢 (以及 Grok 4.7 的名字，沒有用 Grok 5)，以遲到兩個禮拜才釋出的情況來看，應該是很不滿意的情況下硬推的...</p>
<p>另外一個有趣的是提到 Opus 5.5 的傳言，會超過 Fable 5 嗎？還是小幅度升級？目前看起來也是蠻值得期待的... 提到 <a href="https://en.wikipedia.org/wiki/OpenAI">OpenAI</a> 這邊，話說 tibo 好久沒放 reset 了 XD</p>
<picture><source type="image/webp" srcset="https://i.gslin.com/s/1790013661-8f17636b.webp" /><img decoding="async" src="https://i.gslin.com/s/1790013661-8f17636b.png" alt="" /></picture>
]]></content:encoded>
					
					<wfw:commentRss>https://blog.gslin.org/archives/2026/09/22/13217/grok-4-7-%e5%87%ba%e4%ba%86/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
		<post-id xmlns="com-wordpress:feed-additions:1">13217</post-id>	</item>
		<item>
		<title>Cloudflare 的免註冊 reverse tunnel</title>
		<link>https://blog.gslin.org/archives/2026/09/21/13216/cloudflare-%e7%9a%84%e5%85%8d%e8%a8%bb%e5%86%8a-reverse-tunnel/</link>
					<comments>https://blog.gslin.org/archives/2026/09/21/13216/cloudflare-%e7%9a%84%e5%85%8d%e8%a8%bb%e5%86%8a-reverse-tunnel/#respond</comments>
		
		<dc:creator><![CDATA[Gea-Suan Lin]]></dc:creator>
		<pubDate>Sun, 20 Sep 2026 22:42:00 +0000</pubDate>
				<category><![CDATA[Computer]]></category>
		<category><![CDATA[Murmuring]]></category>
		<category><![CDATA[Network]]></category>
		<category><![CDATA[Service]]></category>
		<category><![CDATA[WWW]]></category>
		<category><![CDATA[cloudflare]]></category>
		<category><![CDATA[cloudflared]]></category>
		<category><![CDATA[reverse]]></category>
		<category><![CDATA[try]]></category>
		<category><![CDATA[tunnel]]></category>
		<guid isPermaLink="false">https://blog.gslin.org/?p=13216</guid>

					<description><![CDATA[Hacker News 上看到「Cloudflare Quick Tunnels (cloudflare.com)」這個討論，據說 Cloudflare 提供 https://try.cloudflare.com/ 這個服務已經有至少五年了，不過好像沒寫過... 軟體可以透過 mise 裝： mise use -g cloudflared@latest 或是透過 extrepo 裝： sudo extrepo enable cloudflare; sudo apt update; sudo apt install -y cloudflared; sudo apt clean 裝完後就很簡單的跑： cloudflared tunnel --url http://127.0.0.1:8080/ 接下來畫面上就會出現 .trycloudflare.com 的網址讓你可以提供給同事或是朋友，可以注意到不需要 Cloudflare 的帳號。 這東西類似 ngrok 的服務，但功能少了些，不過如果是要丟些東西給同事或是朋友看的話，算是夠用...]]></description>
										<content:encoded><![CDATA[<p><a href="https://en.wikipedia.org/wiki/Hacker_News">Hacker News</a> 上看到「<a href="https://news.ycombinator.com/item?id=49754785">Cloudflare Quick Tunnels (cloudflare.com)</a>」這個討論，據說 <a href="https://en.wikipedia.org/wiki/Cloudflare">Cloudflare</a> 提供 <a href="https://try.cloudflare.com/">https://try.cloudflare.com/</a> 這個服務已經有至少五年了，不過好像沒寫過...</p>
<p>軟體可以透過 <a href="https://mise.jdx.dev/">mise</a> 裝：</p>
<pre>mise use -g cloudflared@latest</pre>
<p>或是透過 <a href="https://salsa.debian.org/extrepo-team/extrepo">extrepo</a> 裝：</p>
<pre>sudo extrepo enable cloudflare; sudo apt update; sudo apt install -y cloudflared; sudo apt clean</pre>
<p>裝完後就很簡單的跑：</p>
<pre>cloudflared tunnel --url http://127.0.0.1:8080/</pre>
<p>接下來畫面上就會出現 <code>.trycloudflare.com</code> 的網址讓你可以提供給同事或是朋友，可以注意到不需要 Cloudflare 的帳號。</p>
<p>這東西類似 <a href="https://ngrok.com/">ngrok</a> 的服務，但功能少了些，不過如果是要丟些東西給同事或是朋友看的話，算是夠用...</p>
]]></content:encoded>
					
					<wfw:commentRss>https://blog.gslin.org/archives/2026/09/21/13216/cloudflare-%e7%9a%84%e5%85%8d%e8%a8%bb%e5%86%8a-reverse-tunnel/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
		<post-id xmlns="com-wordpress:feed-additions:1">13216</post-id>	</item>
		<item>
		<title>HEIF 的 RCE</title>
		<link>https://blog.gslin.org/archives/2026/09/20/13215/heif-%e7%9a%84-rce/</link>
					<comments>https://blog.gslin.org/archives/2026/09/20/13215/heif-%e7%9a%84-rce/#respond</comments>
		
		<dc:creator><![CDATA[Gea-Suan Lin]]></dc:creator>
		<pubDate>Sun, 20 Sep 2026 13:49:51 +0000</pubDate>
				<category><![CDATA[Computer]]></category>
		<category><![CDATA[Murmuring]]></category>
		<category><![CDATA[Security]]></category>
		<category><![CDATA[Software]]></category>
		<category><![CDATA[code]]></category>
		<category><![CDATA[execution]]></category>
		<category><![CDATA[heif]]></category>
		<category><![CDATA[image]]></category>
		<category><![CDATA[photo]]></category>
		<category><![CDATA[rce]]></category>
		<category><![CDATA[remote]]></category>
		<category><![CDATA[security]]></category>
		<category><![CDATA[vulnerability]]></category>
		<guid isPermaLink="false">https://blog.gslin.org/?p=13215</guid>

					<description><![CDATA[前幾天看到 Mastodon 的 4.7.2 版直接關掉 HEIF 的 changelog：(https://github.com/mastodon/mastodon/releases/tag/v4.7.2) Temporarily disable HEIF support 就大概知道 HEIF 這邊有什麼是情要發生了... 然後今天就看到「HEIF Heist」這個了，HEIF 上的 RCE： By exploiting underlying native libraries, these vulnerabilities allow an attacker to bypass application-level defenses and trigger memory corruption, data exposure, or remote code execution (RCE). 另外「Hacking OpenAI」這篇也是相關的，看起來用了這個洞打進 OpenAI (話說... 居然沒有 sandbox 掉？)。 這樣就解釋的過去了，但特地註冊一個 HEIF Heist 的網域就... &#8230; <a href="https://blog.gslin.org/archives/2026/09/20/13215/heif-%e7%9a%84-rce/" class="more-link">Continue reading<span class="screen-reader-text"> "HEIF 的 RCE"</span></a>]]></description>
										<content:encoded><![CDATA[<p>前幾天看到 <a href="https://en.wikipedia.org/wiki/Mastodon_(social_network)">Mastodon</a> 的 4.7.2 版直接關掉 <a href="https://en.wikipedia.org/wiki/High_Efficiency_Image_File_Format">HEIF</a> 的 changelog：(<a href="https://github.com/mastodon/mastodon/releases/tag/v4.7.2">https://github.com/mastodon/mastodon/releases/tag/v4.7.2</a>)</p>
<blockquote><p>Temporarily disable HEIF support</p></blockquote>
<p>就大概知道 HEIF 這邊有什麼是情要發生了... 然後今天就看到「<a href="https://heif-heist.com/">HEIF Heist</a>」這個了，HEIF 上的 <a href="https://en.wikipedia.org/wiki/Remote_code_execution">RCE</a>：</p>
<blockquote><p>By exploiting underlying native libraries, these vulnerabilities allow an attacker to bypass application-level defenses and trigger memory corruption, data exposure, or remote code execution (RCE).</p></blockquote>
<p>另外「<a href="https://www.hacktron.ai/blog/hacking-openai">Hacking OpenAI</a>」這篇也是相關的，看起來用了這個洞打進 <a href="https://en.wikipedia.org/wiki/OpenAI">OpenAI</a> (話說... 居然沒有 sandbox 掉？)。</p>
<p>這樣就解釋的過去了，但特地註冊一個 HEIF Heist 的網域就... 不予置評 (?)</p>
]]></content:encoded>
					
					<wfw:commentRss>https://blog.gslin.org/archives/2026/09/20/13215/heif-%e7%9a%84-rce/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
		<post-id xmlns="com-wordpress:feed-additions:1">13215</post-id>	</item>
		<item>
		<title>AWS 中東兩區的戰損情況</title>
		<link>https://blog.gslin.org/archives/2026/09/20/13214/aws-%e4%b8%ad%e6%9d%b1%e5%85%a9%e5%8d%80%e7%9a%84%e6%88%b0%e6%90%8d%e6%83%85%e6%b3%81/</link>
					<comments>https://blog.gslin.org/archives/2026/09/20/13214/aws-%e4%b8%ad%e6%9d%b1%e5%85%a9%e5%8d%80%e7%9a%84%e6%88%b0%e6%90%8d%e6%83%85%e6%b3%81/#respond</comments>
		
		<dc:creator><![CDATA[Gea-Suan Lin]]></dc:creator>
		<pubDate>Sun, 20 Sep 2026 11:44:48 +0000</pubDate>
				<category><![CDATA[AWS]]></category>
		<category><![CDATA[Cloud]]></category>
		<category><![CDATA[Computer]]></category>
		<category><![CDATA[Murmuring]]></category>
		<category><![CDATA[Network]]></category>
		<category><![CDATA[Service]]></category>
		<category><![CDATA[amazon]]></category>
		<category><![CDATA[aws]]></category>
		<category><![CDATA[bahrain]]></category>
		<category><![CDATA[cloud]]></category>
		<category><![CDATA[damage]]></category>
		<category><![CDATA[east]]></category>
		<category><![CDATA[middle]]></category>
		<category><![CDATA[service]]></category>
		<category><![CDATA[uae]]></category>
		<category><![CDATA[war]]></category>
		<guid isPermaLink="false">https://blog.gslin.org/?p=13214</guid>

					<description><![CDATA[前幾天的新資訊，在「AWS says wartime damage means some Middle East cloud resources are gone for good」這邊看到的，主要是來自 https://health.aws.amazon.com/health/status 這邊的： Sep 15 3:27 AM PDT We are providing an update on the disruption affecting the Middle East (UAE) (me-central-1) Region. After a thorough assessment, we have determined that we are unable to restore access to the resources and data &#8230; <a href="https://blog.gslin.org/archives/2026/09/20/13214/aws-%e4%b8%ad%e6%9d%b1%e5%85%a9%e5%8d%80%e7%9a%84%e6%88%b0%e6%90%8d%e6%83%85%e6%b3%81/" class="more-link">Continue reading<span class="screen-reader-text"> "AWS 中東兩區的戰損情況"</span></a>]]></description>
										<content:encoded><![CDATA[<p>前幾天的新資訊，在「<a href="https://www.theregister.com/off-prem/2026/09/16/aws-says-wartime-damage-means-some-middle-east-cloud-resources-are-gone-for-good/5296830">AWS says wartime damage means some Middle East cloud resources are gone for good</a>」這邊看到的，主要是來自 <a href="https://health.aws.amazon.com/health/status">https://health.aws.amazon.com/health/status</a> 這邊的：</p>
<blockquote><p>Sep 15 3:27 AM PDT We are providing an update on the disruption affecting the Middle East (UAE) (me-central-1) Region. After a thorough assessment, we have determined that we are unable to restore access to the resources and data hosted exclusively in the mec1-az2 Availability Zone. We continue to work on recovering regional resources, as well as zonal resources hosted in the other affected Availability Zones (mec1-az1 and mec1-az3).</p></blockquote>
<p><code>mec1-az2</code> 被判定滅了，其他兩區繼續救資料...  另外因為沒找到 permalink，所以抓張圖起來記錄：</p>
<picture><source type="image/webp" srcset="https://i.gslin.com/s/1789904388-41fab371.webp" /><img decoding="async" src="https://i.gslin.com/s/1789904388-41fab371.png" alt="" /></picture>
<p>以及：</p>
<blockquote><p>Sep 15 3:22 AM PDT We are providing an update on the disruption affecting the Middle East (Bahrain) (me-south-1) Region. The damage to our infrastructure spanned multiple Availability Zones and exceeded what our regional and multi-AZ services are designed to withstand. After a thorough assessment, we have determined that we are unable to restore access to the resources and data hosted exclusively in this Region.</p></blockquote>
<p>巴林則是 region 全滅...</p>
<p>對應的圖：</p>
<picture><source type="image/webp" srcset="https://i.gslin.com/s/1789904429-a8c71efe.webp" /><img decoding="async" src="https://i.gslin.com/s/1789904429-a8c71efe.png" alt="" /></picture>
<p>應該是目前主要雲端廠商第一個 region 全滅的案例...</p>
]]></content:encoded>
					
					<wfw:commentRss>https://blog.gslin.org/archives/2026/09/20/13214/aws-%e4%b8%ad%e6%9d%b1%e5%85%a9%e5%8d%80%e7%9a%84%e6%88%b0%e6%90%8d%e6%83%85%e6%b3%81/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
		<post-id xmlns="com-wordpress:feed-additions:1">13214</post-id>	</item>
		<item>
		<title>把更多套件改用 extrepo 裝</title>
		<link>https://blog.gslin.org/archives/2026/09/20/13213/%e6%8a%8a%e6%9b%b4%e5%a4%9a%e5%a5%97%e4%bb%b6%e6%94%b9%e7%94%a8-extrepo-%e8%a3%9d/</link>
					<comments>https://blog.gslin.org/archives/2026/09/20/13213/%e6%8a%8a%e6%9b%b4%e5%a4%9a%e5%a5%97%e4%bb%b6%e6%94%b9%e7%94%a8-extrepo-%e8%a3%9d/#respond</comments>
		
		<dc:creator><![CDATA[Gea-Suan Lin]]></dc:creator>
		<pubDate>Sun, 20 Sep 2026 10:45:59 +0000</pubDate>
				<category><![CDATA[Computer]]></category>
		<category><![CDATA[Linux]]></category>
		<category><![CDATA[Murmuring]]></category>
		<category><![CDATA[Network]]></category>
		<category><![CDATA[OS]]></category>
		<category><![CDATA[Software]]></category>
		<category><![CDATA[apt]]></category>
		<category><![CDATA[debian]]></category>
		<category><![CDATA[extrepo]]></category>
		<category><![CDATA[linux]]></category>
		<category><![CDATA[repo]]></category>
		<category><![CDATA[repository]]></category>
		<category><![CDATA[software]]></category>
		<guid isPermaLink="false">https://blog.gslin.org/?p=13213</guid>

					<description><![CDATA[去年有寫「把一些 repository 改用 extrepo 加入」提到用 extrepo 來管理 APT repository，剛剛發現裡面的東西又變多了... 有些軟體我比較偏好透過 mise 裝在 user 目錄下，不過有些我就偏好用 APT repository 了。 從 https://salsa.debian.org/extrepo-team/extrepo-data/-/tree/master/repos/debian 這邊可以看到支援的 APT repository，這次看到的是 Caddy 與 Tailscale。 裡面其實還看到不少可以切過去的，看起來 Debian 這邊的 external repository 之後要有習慣來這邊找...]]></description>
										<content:encoded><![CDATA[<p>去年有寫「<a href="https://blog.gslin.org/archives/2025/03/12/12301/%e6%8a%8a%e4%b8%80%e4%ba%9b-repository-%e6%94%b9%e7%94%a8-extrepo-%e5%8a%a0%e5%85%a5/">把一些 repository 改用 extrepo 加入</a>」提到用 <a href="https://salsa.debian.org/extrepo-team/extrepo">extrepo</a> 來管理 <a href="https://en.wikipedia.org/wiki/APT_(software)">APT</a> repository，剛剛發現裡面的東西又變多了...</p>
<p>有些軟體我比較偏好透過 <a href="https://mise.jdx.dev/">mise</a> 裝在 user 目錄下，不過有些我就偏好用 APT repository 了。</p>
<p>從 <a href="https://salsa.debian.org/extrepo-team/extrepo-data/-/tree/master/repos/debian">https://salsa.debian.org/extrepo-team/extrepo-data/-/tree/master/repos/debian</a> 這邊可以看到支援的 APT repository，這次看到的是 <a href="https://en.wikipedia.org/wiki/Caddy_(web_server)">Caddy</a> 與 <a href="https://en.wikipedia.org/wiki/Tailscale">Tailscale</a>。</p>
<p>裡面其實還看到不少可以切過去的，看起來 <a href="https://en.wikipedia.org/wiki/Debian">Debian</a> 這邊的 external repository 之後要有習慣來這邊找...</p>
]]></content:encoded>
					
					<wfw:commentRss>https://blog.gslin.org/archives/2026/09/20/13213/%e6%8a%8a%e6%9b%b4%e5%a4%9a%e5%a5%97%e4%bb%b6%e6%94%b9%e7%94%a8-extrepo-%e8%a3%9d/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
		<post-id xmlns="com-wordpress:feed-additions:1">13213</post-id>	</item>
	</channel>
</rss>

<!--
Performance optimized by W3 Total Cache. Learn more: https://www.boldgrid.com/w3-total-cache/?utm_source=w3tc&utm_medium=footer_comment&utm_campaign=free_plugin

Object Caching 40/52 objects using APC
Page Caching using APC (Page is feed) 
Minified using APC
Database Caching using APC (Request-wide modification query)

Served from: blog.gslin.org @ 2026-09-23 03:56:42 by W3 Total Cache
-->