<p>small stdlib is like small government</p><p>it doesn't work.</p><p>cargo will truly be rust's demise.</p>

<p><span class="h-card" translate="no"><a href="https://mastodon.social/@dabeaz" class="u-url mention">@<span>dabeaz</span></a></span> &quot;of_thing&quot;, said ocaml</p>

<blockquote><p>be me<br>rewrite a silly go irc bot in rust and target matrix this time<br><code>2.7G ./target</code><br>looks at <code>cargo tree</code><br><a href="https://gist.github.com/arachnist/b93c36ed513177588095ab61283d1bca" target="_blank" rel="nofollow noopener" translate="no"><span class="invisible">https://</span><span class="ellipsis">gist.github.com/arachnist/b93c</span><span class="invisible">36ed513177588095ab61283d1bca</span></a></p></blockquote><p>truly a "what is jezu and kurwa" moment</p>

<p><span class="h-card" translate="no"><a href="https://mastodon.social/@dabeaz" class="u-url mention">@<span>dabeaz</span></a></span> <span class="h-card" translate="no"><a href="https://hachyderm.io/@glitzersachen" class="u-url mention">@<span>glitzersachen</span></a></span> I think there’s huge value in having native idioms used in a language, and that includes between the various versions of python. </p><p>I prefer Rust’s error handling to Go’s, but I’m not going to try to write all my Go code like Rust because it’ll be awkward and confuse others.</p>

<p><span class="h-card" translate="no"><a href="https://hachyderm.io/@glitzersachen" class="u-url mention">@<span>glitzersachen</span></a></span> <span class="h-card" translate="no"><a href="https://linuxlab.sh/@palendae" class="u-url mention">@<span>palendae</span></a></span> Honestly, this is a big reason why we wrote everything from scratch. Sure, we could have taken some crap code originally written for Python 1.5.2, run it through 2to3, and patched it up, but why bother? That would have sucked.</p>

<p><span class="h-card" translate="no"><a href="https://mastodon.social/@dabeaz" class="u-url mention">@<span>dabeaz</span></a></span> I remember not long ago that people were looking for reasons to have less code existing, not generating more of it faster.</p>

<p><span class="h-card" translate="no"><a href="https://mastodon.social/@dabeaz" class="u-url mention">@<span>dabeaz</span></a></span> I’ve had similar thoughts along these lines - LLMs can generate code quickly that looks like code that’s already been written.</p><p>Arguably, that code then is not novel, could be used more easily as a dependency, or some combination of the two.</p>

<p>This got me thinking about LLMs and AI assistants. If you&#39;re relying on something like that, you&#39;re basically coding in the past. </p><p>Whether that&#39;s good or bad, I&#39;m not here to say. I just know that it&#39;s a wildly different thing than what we tried to do in that book 12 years ago.</p><p>I&#39;m not even sure if it would be practical to write a book like that now. Or if anyone even reads programming books for that matter. </p><p>Anyways, just an idle thought for the day. 2/2.</p>

<p>Had a recent conversation with someone who made a comment about the Python Cookbook saying &quot;they had never seen any other Python code like it.&quot;</p><p>I then recounted how the 3rd edition was written from a spreadsheet. Brian Jones and I made a spreadsheet of all planned topics in advance. We then worked to write each topic completely from scratch in the most non-hacky forward-looking Python 3 way that we could think do it. I don&#39;t really remember looking at much already existing code. 1/2</p>