<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0">
    <channel>
        <title><![CDATA[honnibal.dev]]></title>
        <description><![CDATA[Matthew Honnibal is a computational linguist from Sydney based in Berlin, Germany. He's the author of the spaCy Natural Language Processing library and the co-founder of Explosion.]]></description>
        <link>https://honnibal.dev</link>
        <image>
            <url>https://honnibal.dev/icon.png</url>
            <title>honnibal.dev</title>
            <link>https://honnibal.dev</link>
        </image>
        <generator>RSS for Node</generator>
        <lastBuildDate>Wed, 18 Nov 2026 19:40:51 GMT</lastBuildDate>
        <atom:link href="https://honnibal.dev/feed.xml" rel="self" type="application/rss+xml"/>
        <pubDate>Wed, 18 Nov 2026 19:40:51 GMT</pubDate>
        <copyright><![CDATA[All rights reserved 2026, Matthew Honnibal]]></copyright>
        <language><![CDATA[en]]></language>
        <managingEditor><![CDATA[matt@explosion.ai (Matthew Honnibal)]]></managingEditor>
        <webMaster><![CDATA[matt@explosion.ai (Matthew Honnibal)]]></webMaster>
        <item>
            <title><![CDATA[The 'Astra can read and write weights to improve Qwen' post is misleading]]></title>
            <description><![CDATA[The experiments that read raw weights and the ones that fix Qwen use completely different techniques.]]></description>
            <link>https://honnibal.dev/blog/astra-read-write-misleading</link>
            <guid isPermaLink="false">https://honnibal.dev/blog/astra-read-write-misleading</guid>
            <dc:creator><![CDATA[Matthew Honnibal]]></dc:creator>
            <pubDate>Tue, 03 Nov 2026 00:00:00 GMT</pubDate>
        </item>
        <item>
            <title><![CDATA[Restricting Coding Agents' CLI GitHub Write Access]]></title>
            <description><![CDATA[Recently I was using Claude Code and the agent decided to close two issues I was referencing on one of our public repos, as it had access to the gh command, which I hadn't installed previously. The specific action was fine, but I realised the agent actually had full access to my Github, including…]]></description>
            <link>https://honnibal.dev/blog/locking-down-gh</link>
            <guid isPermaLink="false">https://honnibal.dev/blog/locking-down-gh</guid>
            <dc:creator><![CDATA[Matthew Honnibal]]></dc:creator>
            <pubDate>Sat, 21 Mar 2026 00:00:00 GMT</pubDate>
        </item>
        <item>
            <title><![CDATA[The looming AI clownpocalypse]]></title>
            <description><![CDATA[Exploits will soon be cheaper to develop autonomously than they earn. What then?]]></description>
            <link>https://honnibal.dev/blog/clownpocalypse</link>
            <guid isPermaLink="false">https://honnibal.dev/blog/clownpocalypse</guid>
            <dc:creator><![CDATA[Matthew Honnibal]]></dc:creator>
            <pubDate>Sun, 01 Mar 2026 00:00:00 GMT</pubDate>
        </item>
        <item>
            <title><![CDATA[LLMs don't suffer]]></title>
            <description><![CDATA[The "LLM welfare" question provokes a lot of uncertainty in people, but it really shouldn't. The analogy between LLM computations and what we view as morally relevant experiences is very loose.]]></description>
            <link>https://honnibal.dev/blog/suffering</link>
            <guid isPermaLink="false">https://honnibal.dev/blog/suffering</guid>
            <dc:creator><![CDATA[Matthew Honnibal]]></dc:creator>
            <pubDate>Fri, 27 Feb 2026 00:00:00 GMT</pubDate>
        </item>
        <item>
            <title><![CDATA[If I ever meet a flat earther]]></title>
            <description><![CDATA[I've never met a flat earther, and probably this post will never be read by one. I turn 44 this year and when I was growing up this wasn't something anyone believed. For a long time I thought it was a meme, and honestly I still wonder. Until I actually meet someone who makes the case I think I'll be…]]></description>
            <link>https://honnibal.dev/blog/flat-earth</link>
            <guid isPermaLink="false">https://honnibal.dev/blog/flat-earth</guid>
            <dc:creator><![CDATA[Matthew Honnibal]]></dc:creator>
            <pubDate>Fri, 20 Feb 2026 00:00:00 GMT</pubDate>
        </item>
        <item>
            <title><![CDATA[Why I don't think AI is a bubble]]></title>
            <description><![CDATA[Most of the people I like think AI is a bubble. This is a tricky topic to discuss, because the "bubble" framing couples financial and technical issues. It's like a sports fan debating "Is this player overrated?". The answer depends on how good you think that player is, and how good you think other…]]></description>
            <link>https://honnibal.dev/blog/ai-bubble</link>
            <guid isPermaLink="false">https://honnibal.dev/blog/ai-bubble</guid>
            <dc:creator><![CDATA[Matthew Honnibal]]></dc:creator>
            <pubDate>Mon, 16 Feb 2026 00:00:00 GMT</pubDate>
        </item>
        <item>
            <title><![CDATA[Sneaky quokka: Testing and debugging with LLMs]]></title>
            <description><![CDATA[If you write your tests well, at some point they'll identify an important error. If you're using an LLM assistant such as Claude Code or OpenAI Codex, you'll then enter a minigame we can call sneaky quokka. In this minigame the LLM will attempt to cope, rationalise or lie egregiously to get the…]]></description>
            <link>https://honnibal.dev/blog/sneaky-quokka</link>
            <guid isPermaLink="false">https://honnibal.dev/blog/sneaky-quokka</guid>
            <dc:creator><![CDATA[Matthew Honnibal]]></dc:creator>
            <pubDate>Sun, 08 Feb 2026 00:00:00 GMT</pubDate>
        </item>
        <item>
            <title><![CDATA[Style tips for less experienced developers coding with AI]]></title>
            <description><![CDATA[If you're not careful about how you write software, the complexity of the system will grow exponentially with its functionality. Most of the accumulated art of software engineering is about stopping that from happening. The lessons are mostly the same no matter how the code was authored. If you're…]]></description>
            <link>https://honnibal.dev/blog/llm-style-tips</link>
            <guid isPermaLink="false">https://honnibal.dev/blog/llm-style-tips</guid>
            <dc:creator><![CDATA[Matthew Honnibal]]></dc:creator>
            <pubDate>Thu, 05 Feb 2026 00:00:00 GMT</pubDate>
        </item>
        <item>
            <title><![CDATA[Exception handlers are not general-purpose conditionals]]></title>
            <description><![CDATA[There's some bad advice about Python style that's been circling in the community for a long time, and it shows up in LLM-generated code now. If you're using LLMs to help you program, it's extremely important to have good habits dialled in and to recognise problematic patterns. It's therefore…]]></description>
            <link>https://honnibal.dev/blog/try-except</link>
            <guid isPermaLink="false">https://honnibal.dev/blog/try-except</guid>
            <dc:creator><![CDATA[Matthew Honnibal]]></dc:creator>
            <pubDate>Tue, 03 Feb 2026 00:00:00 GMT</pubDate>
        </item>
        <item>
            <title><![CDATA[LLM thoughts emailed to a friend]]></title>
            <description><![CDATA[One time when we went camping you bought your skim board, and I was trying to do it. I had trouble committing so I kept under-shooting, and you observed that since I kept failing in the same direction I should try harder and harder to over-commit and get the opposite failure.]]></description>
            <link>https://honnibal.dev/blog/llm-thoughts</link>
            <guid isPermaLink="false">https://honnibal.dev/blog/llm-thoughts</guid>
            <dc:creator><![CDATA[Matthew Honnibal]]></dc:creator>
            <pubDate>Sat, 01 Nov 2025 00:00:00 GMT</pubDate>
        </item>
        <item>
            <title><![CDATA[Back to our roots]]></title>
            <description><![CDATA[For most of Explosion's life we've been a very small company, running off revenues. In 2021 that changed, and we became a slightly less small company running off venture capital. We've been unable to make that configuration work, so we're back to running Explosion as an independent-minded…]]></description>
            <link>https://honnibal.dev/blog/back-to-our-roots</link>
            <guid isPermaLink="false">https://honnibal.dev/blog/back-to-our-roots</guid>
            <dc:creator><![CDATA[Matthew Honnibal]]></dc:creator>
            <pubDate>Wed, 17 Jul 2024 00:00:00 GMT</pubDate>
        </item>
        <item>
            <title><![CDATA[Deriving the gradient of the cosine loss with Tangent and function composition]]></title>
            <description><![CDATA[Over the last couple of months, I've found myself working out the gradient of the cosine loss twice. I seem to have lost the work the first time. This post makes sure I don't have to work it all out a third time. Plus, if it turns out I've screwed this up, hopefully someone will point it out.]]></description>
            <link>https://honnibal.dev/blog/cosine-derivation</link>
            <guid isPermaLink="false">https://honnibal.dev/blog/cosine-derivation</guid>
            <dc:creator><![CDATA[Matthew Honnibal]]></dc:creator>
            <pubDate>Wed, 20 Mar 2019 00:00:00 GMT</pubDate>
        </item>
        <item>
            <title><![CDATA[Neural program synthesis is a dangerous technology]]></title>
            <description><![CDATA[Today's computer viruses can copy themselves between computers, but they can't mutate. That might change very soon, leading to a serious crisis.]]></description>
            <link>https://honnibal.dev/blog/neural-program-synthesis</link>
            <guid isPermaLink="false">https://honnibal.dev/blog/neural-program-synthesis</guid>
            <dc:creator><![CDATA[Matthew Honnibal]]></dc:creator>
            <pubDate>Fri, 12 Jan 2018 00:00:00 GMT</pubDate>
        </item>
    </channel>
</rss>