<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Blog on Depth First</title><link>https://davidwright.ch/blog/</link><description>Recent content in Blog on Depth First</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Wed, 24 Jun 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://davidwright.ch/blog/index.xml" rel="self" type="application/rss+xml"/><item><title>Lucas Beyer on Transformers: A Talk at the Zurich AI Academy</title><link>https://davidwright.ch/blog/2026/2026-06-24-lucas-beyer-transformers/</link><pubDate>Wed, 24 Jun 2026 00:00:00 +0000</pubDate><guid>https://davidwright.ch/blog/2026/2026-06-24-lucas-beyer-transformers/</guid><description>&lt;p&gt;At Meta I led the Talent &amp;amp; Growth pillar&amp;rsquo;s IC Skills Growth track. As part of this I co-organised the Zurich AI Academy along with a brilliant colleague. In one of our most memorable events, we had the privilege of hosting &lt;a href="https://lucasb.eyer.be"&gt;Lucas Beyer&lt;/a&gt; for a talk on Transformers.&lt;/p&gt;
&lt;p&gt;Lucas co-authored the &lt;a href="https://arxiv.org/abs/2010.11929"&gt;Vision Transformer (ViT)&lt;/a&gt; paper, the &lt;a href="https://arxiv.org/abs/2105.01601"&gt;MLP-Mixer&lt;/a&gt;, &lt;a href="https://arxiv.org/abs/2303.15343"&gt;SigLIP&lt;/a&gt;, and &lt;a href="https://arxiv.org/abs/2407.07726"&gt;PaliGemma&lt;/a&gt;. Previously at Google Brain, where much of this research took place, he went on to co-found OpenAI&amp;rsquo;s Zurich office. At the time of the talk he was working at Meta.&lt;/p&gt;</description></item><item><title>nvim-dap-image: Viewing Image Variables at Breakpoints</title><link>https://davidwright.ch/blog/2026/2026-05-20-nvim-dap-image/</link><pubDate>Wed, 20 May 2026 00:00:00 +0000</pubDate><guid>https://davidwright.ch/blog/2026/2026-05-20-nvim-dap-image/</guid><description>&lt;p&gt;I&amp;rsquo;ve been wanting an image viewer for Neovim&amp;rsquo;s debugger for a long time, so I built &lt;a href="https://gitlab.com/david_wright/nvim-dap-image"&gt;nvim-dap-image&lt;/a&gt;. Place your cursor on an image variable at a breakpoint, run &lt;code&gt;:DapImageView&lt;/code&gt;, and the image appears in a floating window. Supports Python (OpenCV, PIL, NumPy, Matplotlib) and C++ (cv::Mat).&lt;/p&gt;
&lt;p&gt;&lt;video src="https://gitlab.com/david_wright/nvim-dap-image/-/raw/assets/assets/python_demo.mov" controls loop muted playsinline style="width:100%;border-radius:4px"&gt;&lt;/video&gt;&lt;/p&gt;
&lt;h2 id="background"&gt;Background&lt;/h2&gt;
&lt;p&gt;Debugging computer vision and machine learning code in a debugger has one persistent gap: you can inspect a variable&amp;rsquo;s shape, dtype, and values, but you cannot see the image.&lt;/p&gt;</description></item></channel></rss>