<?xml version="1.0" encoding="utf-8"?><feed xmlns="http://www.w3.org/2005/Atom" ><generator uri="https://jekyllrb.com/" version="4.4.1">Jekyll</generator><link href="https://wsvincent.com/feed.xml" rel="self" type="application/atom+xml" /><link href="https://wsvincent.com/" rel="alternate" type="text/html" /><updated>2026-11-10T20:25:30-05:00</updated><id>https://wsvincent.com/feed.xml</id><title type="html">Will Vincent</title><subtitle>Python Software Foundation Fellow, Developer Advocate, and writer on AI, Django, and software development.</subtitle><entry><title type="html">Claude’s Wager: Why you don’t have to believe in AI to bet on it</title><link href="https://wsvincent.com/claudes-wager/" rel="alternate" type="text/html" title="Claude’s Wager: Why you don’t have to believe in AI to bet on it" /><published>2026-11-09T01:57:06-05:00</published><updated>2026-11-09T01:57:06-05:00</updated><id>https://wsvincent.com/claudes-wager</id><content type="html" xml:base="https://wsvincent.com/claudes-wager/"><![CDATA[<p><a href="https://en.wikipedia.org/wiki/Pascal%27s_wager">Pascal’s Wager</a> is a famous philosophical argument that a rational person should act as if God exists, whether or not God <em>does</em> exist. As a software developer, I think it offers a useful way to think about AI coding agents. The question isn’t whether they’ll replace us; it’s what we should do while we find out.</p>

<h2 id="how-we-got-here">How We Got Here</h2>
<p>I’ve been a programmer for years and have experimented with “AI” (really Large Language Models) since November 2022 when ChatGPT was first launched. In the beginning, way back four years ago now, it was truly jaw-dropping that a program could output complete sentences and paragraphs on almost any subject. Coding didn’t seem like the focus at first, but in retrospect it should: text-based, somewhat verifiable, and with customers willing to pay.</p>

<p>The early chatbots and autocomplete tools in 2023–think Cursor here–were impressive but unreliable. I remember trying Cursor in 2023 and being amazed by what it could attempt, then frustrated by how often it got things wrong. For a long time, AI coding felt like more trouble than it was worth. That has changed dramatically.</p>

<p>By 2024, AI tools could perform meaningful tasks on their own, not just suggest code improvements. In 2025, Claude Code, Codex, and more tools took a huge leap by coupling language models with harnesses and agentic patterns, iteratively working through problems. And in 2026, what AI can do is truly jaw-dropping. Humans still need to point AI at a task and provide some taste and judgment, but a growing number of programmers not only don’t write code by hand anymore, but have also become comfortable not reading it.</p>

<h2 id="today">Today</h2>
<p>AI-generated code is typically ugly to my eyes: overly verbose in all the wrong ways. But it’s become the norm in many sectors. Humans in the loop feel overwhelmed by the pace of development, with ever-increasing lines of code and pull requests. Are humans slowing progress or the only things keeping modern software development on the rails?</p>

<p>AI has advanced so quickly that many of us are shell-shocked, to be honest. Stack Overflow is fading, traditional search is increasingly dominated by AI-generated answers, and the internet is filling with AI slop. Even programmers who can manage to avoid the AI wave are still engulfed by it.</p>

<p>The real divide now, in late 2026, is what’s left for humans to do. Some believe that direction, taste, and product focus will remain for software developers. We’ll all become more like product managers rather than coders. Others believe that ship has sailed too and “code is free.”</p>

<h2 id="whats-next">What’s Next?</h2>
<p>I don’t want to spend my days writing code that AI can generate in seconds. But I also don’t want to become a programmer who can no longer understand what they’re building.</p>

<p>My bet—let’s call it Claude’s Wager—is to work alongside AI while continuing to develop the programming skills that let us understand, evaluate, and challenge its output. Maybe coding agents eventually become so capable that they don’t need our guidance at all. Or maybe they become another part of our existing toolkit, alongside linters, syntax checkers, unit tests, and everything else we use to build better software.</p>

<p>You don’t have to believe AI will replace programmers to think it’s worth learning. If AI continues improving at its current pace, programmers who have learned to work with it will be better positioned. If AI hits a wall tomorrow, those programmers will still know how to code.</p>

<p>I don’t know which future we’ll end up in. But the cost of experimenting thoughtfully seems relatively small, while refusing to engage could be enormous.</p>

<p>And refusing to make a bet is a bet in itself.</p>]]></content><author><name></name></author><category term="ai" /><category term="programming" /><summary type="html"><![CDATA[What is left for humans-in-the-loop.]]></summary></entry><entry><title type="html">Artisanal Podcasting in the Age of AI</title><link href="https://wsvincent.com/artisanal-podcasts/" rel="alternate" type="text/html" title="Artisanal Podcasting in the Age of AI" /><published>2026-11-03T01:57:06-05:00</published><updated>2026-11-03T01:57:06-05:00</updated><id>https://wsvincent.com/artisanal-podcasts</id><content type="html" xml:base="https://wsvincent.com/artisanal-podcasts/"><![CDATA[<p><img src="https://wsvincent.com/assets/images/djangochat/garageband_full.png" alt="GarageBand" /></p>

<p>That image is what I see while hand-editing each episode of <a href="https://djangochat.com/">Django Chat</a>, a long-running podcast on the Django web framework that I’ve done alongside <a href="https://noumenal.es/">Carlton Gibson</a> for the better part of seven years. We are releasing episode 209 later this week.</p>

<p>We get a few thousand downloads in the first week of each episode. Nothing enormous, but a respectable audience for an independent podcast that Carlton and I produce alongside our day jobs.</p>

<p>We do this podcast because we enjoy it. So if anything cried out for AI automation it would seem to be this. And there are plenty of AI tools promising to handle the entire production process. In fact, we use one of these tools, Riverside.fm, to record the video, though try as I might, the AI version of the show is never quite good enough.</p>

<p>Instead, I’ve maintained roughly the same production format since 2019. We record the show and have separate audio tracks for each speaker (usually three). Then I run the audio through <a href="https://auphonic.com/engine/">Auphonic</a> to improve it, and drag the processed audio files into GarageBand and edit the show. I also add two additional tracks for the intro and sponsors.</p>

<p>That’s the flow. That’s what I do. And that’s how I’ve spent an inordinate number of hours in my life. I don’t know if I do it better now than I did the first several dozen times, but I certainly am faster at editing. However, it still all goes through my ears/head at least once more after recording before it is released.</p>

<p>So why bother? Well, I think it sounds better. But there are other, less obvious benefits. I experience the conversation a second time, with a different perspective, and without the pressure to think of the next question. Editing also forces me to listen closely, notice moments where things lagged, or needed more emphasis, and so on. And the show notes for the show are also, you guessed it, done by hand.</p>

<p>Now <em>could</em> I dump a single track of the episode into AI, ask for a title and description, audio editing, show notes, and the rest? You bet. And if we were producing more episodes, if we still maintained the weekly pace we started out with, I’d likely be trying to pull the ripcord as well. But we’re not like that. We do about twenty shows a year. It’s possible to do it manually.</p>

<p>As a programmer, I can’t resist feeling parallels between editing a podcast and coding, which <em>is</em> my day job. AI can do an increasingly passable job at the latter. In many organizations, writing and reading code all with AI is the default. And the world still turns, for now.</p>

<p>But what is lost? One thing is deep focus. I can get lost in a flow state where the 1-2 hours it takes to produce a show just goes by in a flash and I lose track of time. I feel the same way writing: for example, that happened with this essay. What felt like fifteen minutes turned out to be an hour. When I’m using AI to code though? I don’t feel any of that. It’s spinning up five different AI agents and waiting for them to finish. My attention bounces between whichever agent finishes first. A cursory review, a follow-up prompt, and on to the next one. It’s exhausting.</p>

<p>I’m undoubtedly producing more code this way. But I’m not sure I’m enjoying it more. Instead of getting lost in the work, I’m managing the work. Those are two very different experiences.</p>

<p>If we started Django Chat today, I doubt I’d choose the same production process. And maybe someday AI will edit the show so well that I’ll happily hand everything over. But for now, I enjoy getting lost in the work. I enjoy listening to our guests again, shaping the conversation, and making something that feels like ours.</p>

<p>I haven’t found that same feeling managing AI agents. Maybe I will. I hope I do.</p>

<p>Until then, do not go gentle into that good AI night.</p>]]></content><author><name></name></author><category term="ai" /><category term="programming" /><summary type="html"><![CDATA[Why I still edit podcasts by hand.]]></summary></entry><entry><title type="html">Questions to Ask a Company Using Django</title><link href="https://wsvincent.com/questions-for-django-employer/" rel="alternate" type="text/html" title="Questions to Ask a Company Using Django" /><published>2026-10-07T02:57:06-04:00</published><updated>2026-10-07T02:57:06-04:00</updated><id>https://wsvincent.com/questions-for-django-employer</id><content type="html" xml:base="https://wsvincent.com/questions-for-django-employer/"><![CDATA[<p>A local friend of mine just joined a very successful Django company. They’ve been around for 10 or so years, so they’re not really a startup, but they are still a fairly small team and yet financially do very, very well. It got me thinking about questions I would ask any future employer for an engineering role.</p>

<p>I want to know what it is actually like to work in their codebase every day.</p>

<p>For a mature Django company, the answers to the following questions reveal a lot about engineering culture, technical debt, developer experience, and how seriously the company treats software development.</p>

<h2 id="1-what-version-of-django-are-you-on">1. What version of Django are you on?</h2>

<p>This is perhaps the most revealing question. I don’t care how big your codebase is; you should have a strategy in place for upgrades.</p>

<p>The easiest security and performance wins revolve around staying current with Django and Python versions. Do you have to be on the absolute latest version? No. That’s what LTS releases are for, and why Django supports older versions for years. But you should have a clear upgrade strategy.</p>

<p>If a company is on a very old version of Django, that’s a big smell test. The next question is: why?</p>

<p>Is the code so broken no one dares touch anything? Did engineers get too clever in the past and now you can’t upgrade third-party packages or core infrastructure without breaking everything else? Is there one mysterious internal package written by someone who left six years ago that prevents the entire application from moving forward?</p>

<p>Those are very different problems from “We’re on the current LTS and plan to upgrade next quarter.”</p>

<p>In a good engineering culture, I’d expect the code to be relatively close to the latest version and for there to be a clear, documented strategy for upgrades and testing.</p>

<p>A useful follow-up question is: How long after a new Django release do you typically upgrade?</p>

<p>The exact answer matters less than whether you have one.</p>

<h2 id="2-how-good-is-your-test-suite">2. How good is your test suite?</h2>

<p>Of course a company has tests. Or at least I hope they do. The more important question is whether they trust them. Can an engineer make a substantial change, run the test suite, and feel reasonably confident that they haven’t broken production? Can the company upgrade Django without months of manual testing? Does CI reliably catch regressions?</p>

<p>Big mature codebases will have test suites that take a while to run, but there are always ways to improve them. Bonus points if they know of and have read Adam Johnson’s excellent <a href="https://adamchainz.gumroad.com/l/suydt">Speed Up Your Django Tests</a> book on this subject. I’d ask how long the test suite takes to run, whether tests run in parallel, how you handle flaky tests, and whether failed tests actually block merges.</p>

<p>A test suite that takes 45 minutes, fails randomly, and is routinely ignored is technically a test suite. It is not a good one.</p>

<h2 id="3-how-do-you-deploy">3. How do you deploy?</h2>

<p>I want to hear something like: boring, automated, frequent, reversible. It should be a non-issue to push code daily, especially with a strong CI/CD pipeline in place. Engineers shouldn’t be afraid to deploy code.</p>

<p>How often do they deploy? Can any engineer deploy? Does merging to the main branch trigger a deployment? Are there staging environments? Feature flags? Automated database migrations? How quickly can a bad release be rolled back? A related question is: What happens when a deployment goes wrong?</p>

<p>Surely they have horror stories to tell. Every company that’s been around long enough does. Listen closely not just to what happened, but to how they fixed it and what changed afterward. Was the response, “Alice made a mistake, so now only Bob is allowed to deploy”? Or was it, “Our process allowed this failure, so we added a check to make sure it couldn’t happen the same way again”?</p>

<p>Those answers reveal a lot about whether an engineering organization is learning or blaming.</p>

<h2 id="4-what-does-your-architecture-look-like">4. What does your architecture look like?</h2>

<p>Is it a monolith? Microservices? Something in between? How has it changed over time?</p>

<p>Personally, if a successful ten-year-old Django company told me most of the product still lived in one large Django application, I would probably consider that a good sign. Software engineering tends to assume that architectural complexity represents maturity. It often represents the opposite. A boring Django monolith backed by Postgres, Redis, and a task queue can go an awfully long way.</p>

<p>What I want to understand is why the architecture looks the way it does. Did they split things into services because there was a clear operational or organizational reason? Or because microservices were fashionable five years ago?</p>

<p>Do they have one database or many? How do background tasks work? What runs outside Django? How much JavaScript is involved? Where are the actual performance bottlenecks?</p>

<p>And perhaps most importantly: If they were starting again today, what would they do differently? Any mature engineering team should have an interesting answer.</p>

<h2 id="5-how-do-you-decide-when-to-adopt-a-new-technology">5. How do you decide when to adopt a new technology?</h2>

<p>This is a good way to understand the engineering philosophy actually in place. One current example is Rust. These days, the answer to every performance issue seems to be “rewrite it in Rust,” preferably with the help of AI. That’s fine if you actually need to do it and it’s a small, contained, performance-critical piece of code. But what is the thought process that gets you there?</p>

<p>I would hope, expect, and perhaps demand that an employer can walk through the reasoning process. What problem are we trying to solve? What are the alternatives? What are the costs of introducing another programming language? Who will maintain it three years from now?</p>

<p>The same question applies to AI tools, frontend frameworks, databases, task queues, type checkers, and whatever comes next. I don’t want an engineering culture that refuses to adopt anything new. But I also don’t want one where every conference talk generates a new infrastructure project.</p>

<p>Good engineering teams should be able to evaluate new technology, weigh the tradeoffs, decide, and then commit. How they answer the Rust question probably shows how they’ll handle future technical changes, too. Make sure their philosophy aligns reasonably well with your own.</p>

<h2 id="6-how-do-you-handle-technical-debt">6. How do you handle technical debt?</h2>

<p>Every company has technical debt. The important question is what they do about it.</p>

<p>Can engineers fix problematic code when they encounter it? Do they set aside dedicated time for maintenance work? Do dependency upgrades count as legitimate engineering work, or does everything need to be tied to a customer-facing feature? One of the fastest ways to make a codebase miserable is to insist that every engineer constantly add new functionality without ever cleaning anything up.</p>

<p>Technical debt also compounds. I’d ask for recent examples of how they balanced resolving technical debt while continuing to push new features.</p>

<h2 id="7-do-you-support-django-and-open-source">7. Do you support Django and open source?</h2>

<p>Many companies do not. They assume quality open source software comes out of thin air. They literally run their businesses on software they don’t give back to. That works up to a point. But it also does <em>not</em> attract the kind of engineers I want to work with.</p>

<p>Sponsoring open source is cheap. So, so cheap. At a minimum, you should <a href="https://www.djangoproject.com/sponsor/">sponsor Django</a> and receive recognition on the main Django website. You can also loudly trumpet this on your repos and hiring pages. Are you using various Django third-party packages? I bet you are. Sponsor them too. Even a small amount goes a long way. And what’s the risk that the maintainer just stops for one reason or another? It’s bigger than you think.</p>

<p>The best Django engineers are involved in the community one way or another. It doesn’t have to be financial. But they are involved.</p>

<h2 id="8-what-is-engineering-life-like-when-something-goes-wrong">8. What is engineering life like when something goes wrong?</h2>

<p>Production will break. A database query will suddenly slow. What happens next? Who gets paged? Is there an on-call rotation? How often does it happen? Are there logs, traces, error reports? And what happens after the incident?</p>

<p>I’d also ask how often engineers work outside normal hours because of production issues. Occasional emergencies happen everywhere. If midnight Slack messages and all-hands weekend fixes are a routine part of engineering life, I’d want to know that upfront before accepting the job.</p>

<h2 id="final-thoughts">Final Thoughts</h2>

<p>There isn’t a “right” answer to any of these questions. What I’m looking for is a clear understanding of the current technical challenges and how the company thinks about things.</p>

<p>They don’t have to be on the latest version of Django or have 100% test coverage. Instead, I’m after a process for deliberation and decision-making. Do they learn from mistakes? What are they planning to work on? Good engineering teams should have thoughtful answers to these questions.</p>

<p>Bonus: If you need help with Django technical questions to ask someone, I recently made a list: <a href="https://learndjango.com/tutorials/django-technical-interview-questions">Django Technical Interview Questions</a>.</p>]]></content><author><name></name></author><category term="django" /><category term="python" /><summary type="html"><![CDATA[Interview questions for your next job.]]></summary></entry><entry><title type="html">Modern Django Deployments in 2026: My DjangoCon US 2026 Conference Talk</title><link href="https://wsvincent.com/modern-django-deployments-in-2026/" rel="alternate" type="text/html" title="Modern Django Deployments in 2026: My DjangoCon US 2026 Conference Talk" /><published>2026-08-26T02:57:06-04:00</published><updated>2026-08-26T02:57:06-04:00</updated><id>https://wsvincent.com/modern-django-deployments-in-2026</id><content type="html" xml:base="https://wsvincent.com/modern-django-deployments-in-2026/"><![CDATA[<p>This is the text version of my DjangoCon US talk, <a href="https://2026.djangocon.us/talks/modern-django-deployments-in-2026/">Modern Django Deployments in 2026</a>, which I recently presented in Chicago.</p>

<p>The video version should be up on YouTube shortly and I will add it here then!</p>

<h2 id="modern-django-deployments-in-2026">Modern Django Deployments in 2026</h2>

<p><img src="https://wsvincent.com/assets/images/djangoconus-2026-talk/slide-01.png" alt="Slide 1" /></p>

<p>I’m going to talk to you today about deployment.</p>

<p>The best Django app in the world is useless if it’s stuck on your local laptop. That gap from “works on my machine” to deployed to anyone on the internet, is what I’m talking about today. And it’s a big one.</p>

<p>I am going to resist the urge to mention tools and instead zoom out and focus on the parts of Django deployments that are the same regardless of how big your application is or what hosting provider you are using.</p>

<p>Who am I? I’m a Developer Advocate at PyCharm/JetBrains and have been in the Django community for a while. Amongst other things, I host the Django Chat podcast with Carlton Gibson and write the Django News newsletter with Jeff Triplett.</p>

<p>I attend quite a few conferences in my current role, but without question DjangoCon US and Europe are my favorite. So I’m very happy to be here with all of you today and this week.</p>

<p><img src="https://wsvincent.com/assets/images/djangoconus-2026-talk/slide-02.png" alt="Slide 2" /></p>

<p>Why is deployment hard? Well, we ask a lot of web developers.</p>

<p>Be an expert in deployment, databases, security, authentication, also be polyglot and comfortable with not just Python but also JavaScript, TypeScript, CSS, HTML, and so on.</p>

<p>It’s a lot. It’s what makes web development fun that we get our hands dirty in so many areas, but it can feel like we have no deep expertise because the surface level of the job is so big.</p>

<p><img src="https://wsvincent.com/assets/images/djangoconus-2026-talk/slide-03.png" alt="Slide 3" /></p>

<p>If you look for advice online on deployment it seems everyone does deployment differently. Best practices are hard to find. As are recommended tooling and vendors.</p>

<p>On Reddit everyone will tell you you’re wrong. By contrast, if you ask AI it will tell you you’re right. Who’s to know?</p>

<p>Everyone thinks their choices are the one true way. And it just leads to confusion.</p>

<p><img src="https://wsvincent.com/assets/images/djangoconus-2026-talk/slide-04.png" alt="Slide 4" /></p>

<p>This is just a short list of available providers out there. Not only do they all have different names and logos, but different approaches. Some are HyperScalers like AWS that offer literally everything. Others offer Virtual Private Servers, hosted options like Platforms as a Service, and many cross those broad boundaries I just sketched out.</p>

<p><img src="https://wsvincent.com/assets/images/djangoconus-2026-talk/slide-05.png" alt="Slide 5" /></p>

<p>But here’s the thing: 90% of Django deployments are the same (not a made-up statistic).</p>

<p>It’s easy to become lost in long deployment checklists and so focused on the trees that we miss the forest. We’re all susceptible to this as engineers.</p>

<p>I want to zoom out today and show how Django deployments are basically the same, regardless of your hosting provider. It’s just the specific details that vary a bit.</p>

<p><img src="https://wsvincent.com/assets/images/djangoconus-2026-talk/slide-06.png" alt="Slide 6" /></p>

<p>Ideally we could just drag our local Django site–which works great by the way–into the cloud and everything would “just work.”</p>

<p>I’m going to walk through in 10 steps why it’s not quite that simple.</p>

<p><img src="https://wsvincent.com/assets/images/djangoconus-2026-talk/slide-07.png" alt="Slide 7" /></p>

<p>Step 1. It can be hard to visualize exactly what the request/response cycle is like in a Django application. But it’s essentially this. A request from the web browser hits your web server that then hits your app server and then your code and data.</p>

<p>This purple box can be displayed as a circle really, since as each request comes in it goes through middleware, URLs, views, models, templates, and more. That however is a separate talk. Let’s just accept that the purple box is the “Django part” here.</p>

<p>The important thing is that locally you don’t have a web server. That magic <code class="language-plaintext highlighter-rouge">runserver</code> command does both these things for us. But in production they must be split apart.</p>

<p>20 years ago there weren’t specs for how web servers and app servers communicated–anyone remember cgi or mod_python? WSGI is that standard: a contract between your app server and your code, so either side can be swapped without the other noticing. ASGI is the same idea, for async. And ASGI came out of Django Channels (thank you Andrew Godwin), which is still maintained by the community, most noticeably Carlton Gibson.</p>

<p><img src="https://wsvincent.com/assets/images/djangoconus-2026-talk/slide-08.png" alt="Slide 8" /></p>

<p>Step 2: The environment. We need a way to toggle between local and production setups. The code is mostly the same but the configuration has changed, hence this dashed box.</p>

<p>Back in the day we did this with multiple <code class="language-plaintext highlighter-rouge">settings.py</code> files, but now we use environment variables in a single settings.py file.</p>

<p>There are a handful of configurations that need to change when readying for production: <code class="language-plaintext highlighter-rouge">DEBUG</code>, <code class="language-plaintext highlighter-rouge">SECRET_KEY</code>, <code class="language-plaintext highlighter-rouge">ALLOWED_HOSTS</code>, the Database URL, etc.</p>

<p>The elegant part of this is we can toggle the configs, but largely not the code. That’s a blessing.</p>

<p><img src="https://wsvincent.com/assets/images/djangoconus-2026-talk/slide-09.png" alt="Slide 9" /></p>

<p>Step 3: The database. Locally, you’re almost certainly on SQLite because it is Django’s default: file-based and fast. You can use it in production–that’s another talk that would be fun to give–but most of the time you will switch to a server-based database like Postgres. Django supports 5 databases out of the box: Sqlite, Postgres, MariaDB, MySQL, and OracleDB. There’s also an officially supported third-party package for MongoDB.</p>

<p>You can absolutely set up a server-based database for local use, too, and it’s pretty straightforward these days, especially with Docker. But you can get quite far with just SQLite locally and then switching to something else in production.</p>

<p>Note that Django 6.1 requires Postgres 15 or newer, so if you’re upgrading Django you might need to bring your database along as well.</p>

<p>And if you are using, say Postgres, you will need to install the Postgres driver as a third-party package, <code class="language-plaintext highlighter-rouge">psycopg</code>.</p>

<p><img src="https://wsvincent.com/assets/images/djangoconus-2026-talk/slide-10.png" alt="Slide 10" /></p>

<p>Step 4. Static files. These are the files that ship alongside your code but don’t change per user: CSS, JavaScript, images, fonts.</p>

<p>The very first PyCon US I ever attended, I saw Jacob Kaplan-Moss give a talk titled “Assets in Django without losing your hair.” If a creator of Django has trouble with assets too, you know it’s “a thing.”</p>

<p>Locally the dev server handles them. It’s only in production when they don’t show up that you go, what?!</p>

<p>Django stops serving static files once DEBUG is off, so now something has to serve them.</p>

<p>3 options in production:</p>

<ol>
  <li>WhiteNoise, a powerful 3rd party package that is the default for many PaaS’s</li>
  <li>Web server directly, likely nginx</li>
  <li>CDN if you have real traffic and want those files near your users</li>
</ol>

<p>In your code these files probably live in a <code class="language-plaintext highlighter-rouge">static</code> directory either project-level or within an app. The <code class="language-plaintext highlighter-rouge">collectstatic</code> command compiles them all into one location and should be a standard build step.</p>

<p>Note that media files are something different. That’s when a user uploads a file, like an avatar or an attachment. A user equals, can’t trust, so they should not be stored on the server’s own disk. Use object storage like S3 for that and you can configure it with the <code class="language-plaintext highlighter-rouge">django-storages</code> package.</p>

<p><img src="https://wsvincent.com/assets/images/djangoconus-2026-talk/slide-11.png" alt="Slide 11" /></p>

<p>Step 5. Authentication — how users sign up and log in.</p>

<p>Locally we can fake it. Use <code class="language-plaintext highlighter-rouge">createsuperuser</code> from the command line and you’re good. In production, a mistake here ends up in the news.</p>

<p>You can do this yourself or use a 3rd party package, but you will need another service for verification, either email or SMS, passkeys. After all, that’s how password reset works. The user has to receive and verify a message from somewhere.</p>

<p>The email service is another API key read at startup (thank you, Environment box). And Django 6.1 just gave this a proper shape: MAILERS, a dict that looks exactly like DATABASES and CACHES, replacing the pile of flat EMAIL_* settings. Those still work, but they’re deprecated — gone in 7.0.</p>

<p>You’ve got public auth now. Great. But now you have another problem, which is where we go next.</p>

<p><img src="https://wsvincent.com/assets/images/djangoconus-2026-talk/slide-12.png" alt="Slide 12" /></p>

<p>Step 6. Security. Notice something different: no new box. For the first time we’re not adding to the diagram, we’re hardening the lines already on it. Security in deployment isn’t a thing you install — it’s defending the connections between boxes you already have.</p>

<p>Locally, none of those lines need defending; the only person on your network is you. In production every one of them is on the public internet.</p>

<p>First, everything is HTTPS. But that’s not enough because it doesn’t protect the session cookie, which remembers who you are, so it must be hardened directly.</p>

<p>Second, once the page is loaded we’re trusting it not to run somebody’s else script. That’s what Content Security Policy (CSP) is for, and as of Django 6.0 it’s built-in via <code class="language-plaintext highlighter-rouge">ContentSecurityPolicyMiddleware</code>.</p>

<p>Third, set <code class="language-plaintext highlighter-rouge">DEBUG=False</code> to not show stack traces on error pages, set <code class="language-plaintext highlighter-rouge">ALLOWED_HOSTS</code> as well.</p>

<p>There’s a genuinely useful command that most people don’t still know about: <code class="language-plaintext highlighter-rouge">manage.py check --deploy</code>, which audits most of this for you and tells you what you missed. Run it in CI, not just once.</p>

<p>That’s the request hardened. But hardening doesn’t make anything faster — and some of the work we cram into a request is simply too slow to belong there.</p>

<p><img src="https://wsvincent.com/assets/images/djangoconus-2026-talk/slide-13.png" alt="Slide 13" /></p>

<p>Step 7. Background workers. You’ve never needed these locally, for a simple reason: when something’s slow on your laptop, you just… wait. But in production milliseconds matters.</p>

<p>Slow tasks can include sending email, cleaning up after a Stripe charge, and many others.</p>

<p>The fix is a task queue: hand the job off, respond immediately, do the work in the background. Django 6.0 added Tasks for exactly this — but note what it gives you. An interface, not a worker.</p>

<p>Matias gave a talk earlier on “Building a Celery backend for Django Tasks” and there already are backends published and more to come, that hook into this interface.</p>

<p>The broker (probably Redis) is your mailbox: your app drops a note in and moves on. The worker pulls notes out.That’s a gross simplification but is roughly accurate.</p>

<p><img src="https://wsvincent.com/assets/images/djangoconus-2026-talk/slide-14.png" alt="Slide 14" /></p>

<p>Step 8. Performance. Like security, no new boxes — just problems on the ones we have.</p>

<p>Where do apps get slow? Almost never Python, almost never Django. Almost always the database. Three moves, in order.</p>

<ol>
  <li>First, check for N+1 issues. Jacob Walls gave a talk on this earlier so I won’t also give a full explanation. But historically this is an issue with ORMs. The classic way to address is with <code class="language-plaintext highlighter-rouge">select_related</code> and <code class="language-plaintext highlighter-rouge">prefetch_related</code></li>
</ol>

<p>Django 6.1 has modelfield fetch modes that, again, Jacob gave a full talk on earlier.</p>

<ol>
  <li>
    <p>Second, indexes. Django already indexes your foreign keys; add the rest once production reveals which queries matter.</p>
  </li>
  <li>
    <p>Third, caching. Compute once, park it in Redis, serve from memory. The most powerful and dangerous option: two copies of the truth and you have to know when to throw away the stale one.</p>
  </li>
</ol>

<p><img src="https://wsvincent.com/assets/images/djangoconus-2026-talk/slide-15.png" alt="Slide 15" /></p>

<p>Step 9. Monitoring. Anything that can break will, but how do you know? There’s many moving pieces in production that can go wrong, not to mention users will eventually find all the pathways that break. Will you be notified when that happens?</p>

<p>That’s what this box fills. Notice it connects with dashed lines, not solid arrows, because it’s not part of the request path. It sits off to the side and watches.</p>

<p>Think about how this is different from local. On your laptop, when something breaks you know instantly–the traceback is right there in your terminal, or you see a yellow error page in the case of Django. In production, if you see it at all, it’s just a polite blank 500 because we hid the details on purpose, for security.</p>

<p>This monitoring box does a few things. A tool like Sentry (or DataDog, New Relic, GlitchTip) catches every exception, tells you which pages are slow.</p>

<p>There’s only one thing remaining…</p>

<p><img src="https://wsvincent.com/assets/images/djangoconus-2026-talk/slide-16.png" alt="Slide 16" /></p>

<p>Step 10, deployment. Locally, deploying is just hitting save: the dev server reloads, your change is live.</p>

<p>In production, deploying is this strip along the bottom. 4 steps from left to right.</p>

<ol>
  <li>You <code class="language-plaintext highlighter-rouge">git push</code> the code.</li>
  <li>Then tests. You have tests, right? That could be pre-commit, CI. Something.</li>
  <li>Then build, where your source code becomes a runnable artifact, usually a container image, whether you wrote the Dockerfile yourself or a buildpack inferred one for you. This is where <code class="language-plaintext highlighter-rouge">collectstatic</code> for static files runs, if you remember way back to step 4.</li>
  <li>The release: before your new code serves a single request, the database schema has to match it. So your migrations run here.</li>
</ol>

<p>And with that last arrow… we’re done.</p>

<p>This is the entire talk, on one slide. It’s the mental model I think you need for Django deployments. Every box up there is something <code class="language-plaintext highlighter-rouge">runserver</code> was quietly doing for you before. That gap - the distance between one process running on localhost and this whole diagram - is what “Django deployment” means to me.</p>

<p><img src="https://wsvincent.com/assets/images/djangoconus-2026-talk/slide-17.png" alt="Slide 17" /></p>

<p>We’re back to where we started and back to the question: why can’t we just drag and drop local to production?</p>

<p>I’ve given you 10 reasons why not and there are many more once you start going down the infrastructure rabbit hole.</p>

<p>I hope this overview is helpful in demystifying what’s happening in deployment. Because if you have this mental model, then diving into whatever deployment checklist you come up with for yourself or use at your company, should make a lot more sense.</p>

<p><img src="https://wsvincent.com/assets/images/djangoconus-2026-talk/slide-18.png" alt="Slide 18" /></p>

<p>Thank you for your attention. You can find me online and the slides will be up there as well.</p>]]></content><author><name></name></author><category term="django" /><category term="python" /><category term="conferences" /><summary type="html"><![CDATA[A written guide to my talk on deploying Django and why 90% of it is the same.]]></summary></entry><entry><title type="html">Deploying Web Apps in 2026: My EuroPython 2026 Conference Talk</title><link href="https://wsvincent.com/deploying-web-apps-in-2026/" rel="alternate" type="text/html" title="Deploying Web Apps in 2026: My EuroPython 2026 Conference Talk" /><published>2026-07-21T02:57:06-04:00</published><updated>2026-07-21T02:57:06-04:00</updated><id>https://wsvincent.com/deploying-web-apps-in-2026</id><content type="html" xml:base="https://wsvincent.com/deploying-web-apps-in-2026/"><![CDATA[<p>This is the text version of my EuroPython talk, <a href="https://ep2026.europython.eu/session/deploying-python-web-apps-in-2026">Deploying Python Web Apps in 2026</a>, which I recently presented in Krakow, Poland.</p>

<div class="bookshelf">
<iframe width="650" height="365" src="https://www.youtube.com/embed/QhhnMwLEXMA?si=UYfKAF5WTB75DHoD" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" referrerpolicy="strict-origin-when-cross-origin" allowfullscreen=""></iframe>
</div>

<h2 id="deploying-web-apps-in-2026">Deploying Web Apps in 2026</h2>

<p><img src="https://wsvincent.com/assets/images/europython-2026-talk/slide-01.png" alt="Slide 1" /></p>

<p>Hi everyone, Thank you for coming. I’m going to talk to you today about deployment. I’m a Developer Advocate at PyCharm/JetBrains and have been involved in the Python and Django communities for a while.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026-talk/slide-02.png" alt="Slide 2" /></p>

<p>We ask a lot of modern web developers. Be an expert programmer in multiple programming languages: Python, JavaScript, TypeScript, and increasingly be dangerous with Rust, too. Know HTML and CSS. The former is learnable quickly, but CSS can become complicated fast and talented programmers and developers can spend an entire career on just it. That’s why there are CSS frameworks out there like Tailwind, Bootstrap, and others, but that’s yet another level of complexity we are expected to manage.</p>

<p>Web developers should also understand databases, write performant SQL, and know how to manage databases so they’re secure and not slow. It’s also important to implement user authentication so people can use your site, but don’t mess up the security here–keep out the bad actors! And so on. It’s kinda a lot.</p>

<p>Deployment fits into this bucket list of skills for web developers, too. Not our core job, but critical. Not something we implement or even focus on all the time, but a tool we use everyday, since most people are deploying from local to production regularly.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026-talk/slide-03.png" alt="Slide 3" /></p>

<p>If you look for advice online, it seems everyone does deployment differently. There are few best practices–everyone has their favorite PaaS or VPS or custom approach. Even an LLM can’t make up its mind.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026-talk/slide-04.png" alt="Slide 4" /></p>

<p>This slide is just a short list of available providers out there. You might have a choice on which one to use for work, or maybe not.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026-talk/slide-05.png" alt="Slide 5" /></p>

<p>If you look closely there are different categories of hosting options within this collage.</p>

<ul>
  <li><strong>Hyperscalers</strong> like Google, Azure, and AWS give you a catalog of services: hosting, databases, media, queues, etc.</li>
  <li><strong>VPS (Virtual Private Server)</strong> options sit slightly down the stack, offering you a virtual private server with as much size as you want.</li>
  <li><strong>Self-hosted PaaS (Platforms as a Service)</strong> options are a hybrid category that is growing, whereby you can operate your own PaaS on your VPS of choice.</li>
  <li><strong>Traditional PaaS</strong> options that handle most of this for you, for a fee, are there</li>
  <li><strong>Framework-specific</strong> hosting options, most notably FastAPI cloud, are the final option</li>
</ul>

<p>The general point is that there are a lot of options along the spectrum. The question is how much do you want to do yourself.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026-talk/slide-06.png" alt="Slide 6" /></p>

<p>But here’s the thing: 90% of Python app deployments are the same. (This is definitely not a made up number). Without a mental model of the entire process you will be entirely lost. But almost all of it is independent of a specific provider. You just need to know where to focus. Otherwise you’ll be blindly following online tutorials or AI text, following a checklist you don’t understand leading to an outcome you can’t define.</p>

<p>I’ve been there. I deployed dozens and dozens of websites without, to be honest, a clear mental model of what I was doing.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026-talk/slide-07.png" alt="Slide 7" /></p>

<p>My goal is to get you to this. Some feeling of peace and zen. Knowing how the big pieces of deployment work so that you can zoom in and out as needed.</p>

<p>We’re not infrastructure engineers whose job is 24/7 to do deployments and manage up time. Most of us are simple web developers, but we are forced to act as infrastructure engineers, too.</p>

<p>In the real world, you need a mental model of deployment because otherwise you’re hunting for online tutorials, or asking AI for tips, and just blindly following a checklist you don’t understand.</p>

<p>This talk is the one I wanted to see when learning how to be a web developer. Overwhelmed by all these new tools. And lacking a mental model of how the pieces fit together. A lot of this is web app deployment 101 type of information that applies to any programming language or framework. But as you’ll see, there are some Python-specific things developed over the years to make things even nicer for us in the Python community.</p>

<p>So whether you have never deployed a website before, have done it in one framework only, just used a PaaS and not a VPS, I think you will take something away from this presentation.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026-talk/slide-08.png" alt="Slide 8" /></p>

<p>We all know this is Steve Jobs, right? There is an apocryphal tale of how the Mac’s drag and drop functionality for CD burning came to be. The short version is Apple engineers and designers and marketers spent months on this new feature, leading up to a big meeting with Jobs. They’re eager to present their research, implementation plans, all of it. They’re all seated around a table. Jobs walks in, goes the whiteboard, and draws a music folder with an arrow to a CD.</p>

<p>That’s it. That’s the functionality. Do it. So this tale goes.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026-talk/slide-09.png" alt="Slide 9" /></p>

<p>Deployment is NOT this simple unfortunately. But we will use this story as a guiding metaphor for the talk. What we want is to just drag our local website onto a server somewhere and be done with it. Things aren’t quite that simple, but we can get close.</p>

<p>The problem is that there is quite a lot going on locally in our beloved frameworks of choice to make a seamless local development environment experience happen.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026-talk/slide-10.png" alt="Slide 10" /></p>

<p>In fact, a lot of what is happening locally feels like magic at first. I know we all hate magic as developers–we want things explicit, especially as Python developers–but software is built on abstractions so some degree is good. And you can go in and fiddle with these things if you want, but then you are largely fighting the framework, which is bad.</p>

<p>I have a list of things that differ and my goal is that there’s nothing magical about all this at the end. In fact, maybe you’ll go up to some of the maintainers–I think Sebastian from FastAPI is here–and thank them for doing all this work for us.</p>

<p>We’re going to build up the complexity step-by-step (10 total) so you can see what’s involved step-by-step. Remember, these 10 steps are “all” that’s happening in any Python web deployment. So it’s worth internalizing them.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026-talk/slide-11.png" alt="Slide 11" /></p>

<p>Step 1. There’s a lot going on locally to serve you a Python site–and the key thing to notice is two different jobs are happening, even though locally it only looks like one.</p>

<p>Locally, a single process does everything. For Django that’s <code class="language-plaintext highlighter-rouge">runserver</code>, for Flask <code class="language-plaintext highlighter-rouge">flask run</code>, and for FastAPI <code class="language-plaintext highlighter-rouge">fastapi dev</code>. You type one command that one server accepts the connection, parses the HTTP, runs your code, and even hands back your static files (more on that soon). One command, one process. It feels seamless and that’s exactly the point. It’s hiding a split.</p>

<p>In production, those two jobs come apart into two separate servers. The first job is the <strong>web server</strong>–nginx or Caddy. Think of it as the front door. It speaks HTTP to the browser, handles TLS, and serves static files directly. It’s built to juggle thousands of connections fast, and it doesn’t run a single line of your Python web app code.</p>

<p>The second job is the <strong>app server</strong>–Gunicorn or Uvicorn. This is the one that actually runs your code: Django, Flask, FastAPI. The web server takes every request at the door and hands the dynamic ones off to the app server, which calls your application and passes a response back.</p>

<p>This is the first thing installed purely for production. Locally you never need it; the dev server was your app server. It’s the first of a small handful of production-only packages, and we’ll pick up more as we go.</p>

<p>And that handoff–the line between the two boxes–is WSGI or ASGI: a standard contract for how a web server talks to a Python app server, so any of them can mix and match. WSGI is the original synchronous one; ASGI is the newer async version for things like websockets.</p>

<p>So: one process locally, two servers in production.</p>

<p><strong>A brief history lesson</strong></p>

<p>PEP 333 from 2003 is for WSGI (Web Server Gateway Interface), which standardized how this works “for Python apps.” If you’re not using Python, you’re on your own. Or rather need to use something else.</p>

<p>This was later updated to PEP 3333 (the Python 3 update) in 2010. And ASGI (Asynchronous Server Gateway Interface) emerged in the mid 2010s from Django Channels.</p>

<p>Django ships with both ASGI/WSGI. FastAPI is built on Starlette: there is no WSGI mode; it’s async all the way up. Flask is WSGI-only, but Quart is Flask with ASGI.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026-talk/slide-12.png" alt="Slide 12" /></p>

<p>Step 2: Here’s the surprising part: between local and production your code doesn’t really change. The same files roughly ship to both. What changes is the configuration fed into them–and that’s this new dashed box, the environment.</p>

<p>It wasn’t always this way. The old approach was separate settings files, one for local, one for staging, one for production, and you’d swap them depending on where you were. The problem is your secrets end up sitting in your codebase. The modern best practice is to pull all that out into environment variables, read in at startup.</p>

<p>So what actually lives here? The database URL. The secret key. Whether DEBUG is on or off in a Django app. API keys for email, payments. The handful of values that differ between your laptop and the real thing–the ones you never want committed to git.</p>

<p>This comes straight from the twelve-factor app, written by Heroku about fifteen years ago, and it does two things at once: it keeps secrets out of your code and it lets the exact same code behave differently in each environment just by changing what you feed it. Locally you keep a <code class="language-plaintext highlighter-rouge">.env</code> file that you gitignore; in production you set the real values on the platform.</p>

<p>And here’s why it comes so early, as Step 2: every box we add for the rest of the talk wires in the same way–a URL or a key, sitting in the environment, read once at startup. The code stays the same. We’re just going to keep handling it more connections.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026-talk/slide-13.png" alt="Slide 13" /></p>

<p>Step 3: The database. Locally, you’re almost certainly on SQLite–it’s Django’s default, and the usual starting point for SQLAlchemy or SQLModel. And for good reason: it’s just a file, fast, and there’s no separate server to install or run. You get a real database for free.</p>

<p>But this is the first box where what’s perfect locally is often the wrong tool in production–and it’s worth being precise about why. SQLite isn’t a toy; it’s the most deployed database on Earth. In every cell phone, appliance, car, and so on. The catch is it’s almost entirely embedded–one file, one user, one device. Web production is the opposite: many users writing at the same time, which is historically where SQLite struggles.</p>

<p>There’s a real SQLite-on-the-web movement now, but it mostly solves replication and read performance. The single-writer limit is still there. I’ll leave the rest for the hallway track.</p>

<p>So the production default is a server-based database, Postgres is particularly popular, that runs as its own independent service. Your app connects to it over the network. A PaaS will often hand you one and wire it up; otherwise you point at a managed database, or run your own.</p>

<p>You can run a server-based database like Postgres locally–it’s not too hard, especially with Docker containers–but that’s a more advanced local set up. Most developers start, at least, with SQLite.</p>

<p>If you’re keeping track, this is the second production-only package we need: the Postgres driver, <code class="language-plaintext highlighter-rouge">psycopg</code>.</p>

<p>Here’s the first payoff of that environment box from a minute ago: the PaaS hands you the database as a single value–a <code class="language-plaintext highlighter-rouge">DATABASE_URL</code>–and a line or two of config points your framework at it.</p>

<p>One gotcha I see all the time newcomers deploy a new site and then go: where’s the info? Because they have added, say, blog posts, and created users locally. But in production it’s a different database. Seems simple, but I bet everyone does this a few times when learning web development.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026-talk/slide-14.png" alt="Slide 14" /></p>

<p>Step 4. Static files. If you’ve deployed before, you probably just groaned a little–this is the box that bites everyone. So what are they? Everything that isn’t code and doesn’t change per user: your CSS, your JavaScript, images, fonts. The files that just ship alongside your app.</p>

<p>Locally, you have never thought about these, and that’s the tell — it’s the clearest example yet of the dev server quietly doing a job for you. Django and Flask serve your static folder automatically; you drop a file in, it shows up. (FastAPI makes you mount it explicitly, but FastAPI apps are usually JSON APIs, so there the static problem moves into a JavaScript bundler instead.)</p>

<p>The rite of passage here is a Django one: you set <code class="language-plaintext highlighter-rouge">DEBUG=False</code> like the docs say, reload, and your CSS is gone–because Django stops serving static once debug is off. But the underlying lesson is universal: in production, the dev server’s free static-serving goes away in every framework. Flask won’t serve it at scale, FastAPI never offered to. Somebody else has to serve those bytes now.</p>

<p>In production, there are three options, from simplest to beefiest:</p>

<ol>
  <li>WhiteNoise–another of those production-only installs–where your Python process serves the files itself, properly, with compression and cache headers and no extra infrastructure, which is why it’s the PaaS default.</li>
  <li>the web server, nginx or Caddy serving them directly, which is what’s drawn on the diagram</li>
  <li>a CDN (Content Delivery Network), once you have real traffic and want those files sitting close to your users.</li>
</ol>

<p>Two things to bank before we move on. First, there’s a step hiding in here: in Django, a command called <code class="language-plaintext highlighter-rouge">collectstatic</code> gathers every static file into one place so it can be served — and that runs in what’s called the build step. Hold onto that phrase; it’s the second thing we’ll cash in at the very end. Second, a quick distinction people miss: files a user uploads–avatars, attachments–are media files, not static files, and those belong in off-server object storage like S3, never on the server’s own disk.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026-talk/slide-15.png" alt="Slide 15" /></p>

<p>Step 5. Still with me? There’s only 10 of these. That’s the talk. So hang in there. I know this information is a little dense. I’ll post the slides online later, so you can refer to them as needed.</p>

<p>Step 5 is Authentication–how users sign up and login. Locally, this is a solved problem, because you can just fake it: spin up a user straight from the terminal (<code class="language-plaintext highlighter-rouge">createsuperuser</code> in Django) and you’re in. In production it suddenly gets real, and getting it wrong is the kind of mistake that ends up in the news. (You can dodge a lot of this with social login–”login with Google”–but assume you’re rolling your own.)</p>

<p>Here’s the part that sneaks up on people. Say you let users create their own accounts without email confirmation. Fine. But even if you YOLO that, what about password reset? Now you’re sending mail. Authentication, it turns out, secretly depends on email–and that’s the box on the diagram.</p>

<p>Yes, I know you don’t have to use email. Could do SMS, 2FA, passkey, other options. But take this box as a proxy for some service to power authentication.</p>

<p>Continuing with our email example, you’ll need a transactional email service like Postmark, SES, SendGrid, Mailgun, and then you have to prove to the receiving inboxes that you’re actually you, by setting SPF and DKIM records on your domain. None of this is Python, but all of it has to be in place before a single password-reset-link reaches anyone.</p>

<p>The good news, if there is one, is that the environment box pops up again. The email service is another API key sitting in your config, read at startup.</p>

<p>Also a couple of things change the instant your auth is public. Your login form is now on the open internet, which means bots will find it and hammer it, so you want rate limiting on your auth endpoints.</p>

<p>But notice that we’re quietly assuming the login session itself is trustworthy. That little cookie traveling back and forth between the browser and your server can’t be read or forged. And that assumption is exactly where we go next.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026-talk/slide-16.png" alt="Slide 16" /></p>

<p>Step 6. Security. And notice something different about this slide: no new box. For the first time we’re not adding a diagram, we’re hardening the lines already on it. Security in deployment isn’t a thing you install; it’s defending the connections between boxes you already have.</p>

<p>Locally, none of those lines need defending–the only person on your network is you. In production, every one of them is out on the public internet, where anyone can reach them.</p>

<p>So first, everything is HTTPS, which is just HTTP wrapped in TLS encryption, and the good news is your platform almost certainly handles the certificates for you automatically. But HTTPS alone isn’t enough, because it doesn’t specifically protect the session cookie–and that cookie is the thing that remembers who you are, since HTTP itself has no memory. So you harden the cookie directly: The <code class="language-plaintext highlighter-rouge">Secure</code> flag, so the browser never sends it over plain HTTP; <code class="language-plaintext highlighter-rouge">HttpOnly</code> so JavaScript can’t read it; and HSTS so the browser refuses to speak plain HTTP at all.</p>

<p>Then there’s the code side. In Django, there’s a <code class="language-plaintext highlighter-rouge">DEBUG=False</code> config that stops showing stacktraces on error pages, forces a proper <code class="language-plaintext highlighter-rouge">SECRET_KEY</code>, forces updating <code class="language-plaintext highlighter-rouge">ALLOWED_HOSTS</code>, and there’s a genuinely useful command most people don’t know about–<code class="language-plaintext highlighter-rouge">manage.py check --deploy</code>–that audits most of this for you and tells you what you missed.</p>

<p>The other two frameworks have their own version of this checklist. Flask leans on the Flask-Talisman extension to force HTTPS and set security headers. FastAPI often sidesteps the session-cookie question entirely, because it’s usually an API handing out tokens instead–and when it is the backend for, say, a Next.js frontend, the cookie lives on the frontend server, so the lock travels with it there.</p>

<p>That’s the request hardened. But hardening doesn’t make anything faster–and some of the work we cram into a request is simply too slow to belong there. Fixing things brings new boxes back to the diagram.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026-talk/slide-17.png" alt="Slide 17" /></p>

<p>Step 7. Background workers. You’ve never needed these locally, for a simple reason: when something’s slow on your laptop, you just… wait. Nobody’s timing you. In production, a web request has a budget measured in milliseconds — and some of the work we ask it to do blows right through that.</p>

<p>What kind of work? Sending a few thousand emails. Resizing the photo someone just uploaded. The cleanup after a Stripe charge — the charge itself is quick, but the webhooks, receipts, and reconciliation that follow are not. Or, increasingly, an LLM call that needs ten seconds to think. None of these belong inside a request, with a user watching a spinner while they run.</p>

<p>The fix is a task queue: hand the slow job off, return a response immediately, and let the work happen in the background. This is common enough that it’s becoming built-in — Django 6.0 ships a <code class="language-plaintext highlighter-rouge">Tasks</code> interface for exactly this, and FastAPI has <code class="language-plaintext highlighter-rouge">BackgroundTasks</code> for smaller jobs — but the shape underneath is worth seeing, because it’s two new boxes.</p>

<p>The first is the message broker — usually Redis. Think of it as a mailbox: a fast little datastore whose only job is holding the list of work waiting to be done. Your app drops a note in and moves on; it doesn’t wait around for the result.</p>

<p>The second is the task worker — and here’s the nice part: it’s a second copy of your application. The same code you’ve been deploying all along, except with no web server bolted to the front of it. Remember the app server from Step 1? This is that same Python, running on its own, pulling notes out of the mailbox and doing the work. Celery and RQ are the usual tools.</p>

<p>And notice the wiring: your web process never talks to the worker directly — everything goes through the broker. That decoupling is the entire point. Either side can crash, restart, or scale up without the other even noticing. (It also means that when a worker fails, there’s no user sitting there to complain about it — hold onto that, because it comes back at the monitoring box.)</p>

<p>So we’ve moved the slow work out of the request. The next question is what makes the work that’s left slow — and the answer is almost always one box we already have. That Redis mailbox, by the way, is about to do more than hold mail.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026-talk/slide-18.png" alt="Slide 18" /></p>

<p>Step 8. Performance. And like security, this slide adds no new boxes–just problems to solve on the boxes we already have. So where do production apps actually get slow? Here’s the thing: it’s almost never Python, and it’s almost never your framework. It’s almost always <strong>the database</strong>. You’ve got three moves, and the order matters.</p>

<p>First, ask better questions. The classic is the <strong>N+1 problem</strong>–invisible locally, brutal in production. Picture a homepage listing a hundred blog posts, each showing its author. Your ORM runs one query for the posts… and then, as the template renders each one and reaches for its author, one more query per post. That’s a hundred and one queries for a page that needed two. The fix is telling the ORM to fetch it all up front — <code class="language-plaintext highlighter-rouge">select_related</code> and <code class="language-plaintext highlighter-rouge">prefetch_related</code> in Django, joined loads in SQLAlchemy.</p>

<p>Second, help the database answer faster–that’s <strong>indexes</strong>. Django already indexes your foreign keys for you; the rest you add once production shows you which queries actually matter. And don’t reflexively index everything: every index you add is one more thing to update on every single write.</p>

<p>Third, <strong>caching</strong>. Compute the answer once, park it in Redis, serve it from memory. It’s the most powerful move and the most dangerous, because now you have two copies of the truth and you have to know when to throw the stale one away. That’s why it’s last: reach for it after you’ve fixed the queries, not instead of fixing them.</p>

<p>Now, two quiet promotions on the diagram. First, that Redis box–it arrived as your message broker, but it’s just as happy holding your cache, and your session store too. One service, three jobs, one <code class="language-plaintext highlighter-rouge">REDIS_URL</code>. In practice you’ll usually split this into separate Redis databases or instances once you have real cache traffic. Second, the app server: you scale it by running more copies–more workers, more machines, sitting behind a load balancer. On a PaaS, that’s often a literal slider you drag.</p>

<p>And here’s the question worth ending on: why does that slider work? Why can you scale your most important box just by dragging a dial? Because nothing important lives inside it anymore. Your sessions are in Redis. Your data is in Postgres. Your uploads are in object storage. Your slow jobs are in the queue. We have spent this entire talk quietly carrying state out of that purple box–and that’s the whole reason scaling it is a dial, and not a project.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026-talk/slide-19.png" alt="Slide 19" /></p>

<p>Step 9. Monitoring. Anything that can break will, but how do you know? There’s many moving pieces in production that can go wrong, not to mention users will eventually find all the pathways that break. Will you be notified when that happens? That’s what this box fills. Notice it connects with dashed lines, not solid arrows, because it’s not part of the request path. It sits off to the side and watches.</p>

<p>Think about how this is different from local. On your laptop, when something breaks you know instantly–the traceback is right there in your terminal, or you see a yellow error page in the case of Django. In production, if you see it at all, it’s just a polite blank 500 because we hid the details on purpose, for security.</p>

<p>This monitoring box does a few things. A tool like Sentry (or DataDog, New Relic, GlitchTip) catches every exception, tells you which pages are slow.</p>

<p>Whatever monitoring tool you use is the last of those production-only packages we started collecting back in Step 1: Gunicorn for a WSGI server, psycopg for the Postgres, database driver, WhiteNoise for static files, and this. Like every single box before it, the wiring can go through Environment variables.</p>

<p>There’s only one thing remaining…</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026-talk/slide-20.png" alt="Slide 20" /></p>

<p>Step 10, the final one, is the deploy itself. Locally, deploying is just hitting save: the dev server reloads, your change is live. In production, deploying is this strip along the bottom. 4 steps from left to right.</p>

<ul>
  <li>You <code class="language-plaintext highlighter-rouge">git push</code> the code.</li>
  <li>Then tests. There are talks and entire books on testing, so I’m going to largely wave my hands here, but you want automated tests, especially in production, to run before allowing any changes.</li>
  <li>Then build, where your source code becomes a runnable artifact, usually a container image, whether you wrote the Dockerfile yourself or a buildpack inferred one for you. This is where <code class="language-plaintext highlighter-rouge">collectstatic</code> for static files runs, in Django anyway, if you remember way back to step 4.</li>
  <li>The release: before your new code serves a single request, the database schema has to match it. So your migrations run here, as a deliberate release command.</li>
</ul>

<p>And with that last arrow… we’re done. This is the entire talk, on one slide. Every box up there is something your dev server was quietly doing for you, or quietly hiding from you, on your laptop. The gap–the distance between one process running on localhost and this whole diagram–is what the word “deployment” actually means.</p>

<p>The good news is you now have the map. A decent PaaS will handle most of these boxes already assembled and wired together. Or you can do it yourself if you need to.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026-talk/slide-21.png" alt="Slide 21" /></p>

<p>We’re back to where we started and back to the question: why can’t we just drag and drop a website like burning a CD?</p>

<p>I’ve given you 10 reasons why not and there are many more once you start going down the infrastructure rabbit hole.</p>

<p>I hope this overview is helpful in demystifying what’s happening in deployment. Because if you have this mental model, then diving into whatever deployment checklist you come up with for yourself or use at your company, should make a lot more sense.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026-talk/slide-22.png" alt="Slide 22" /></p>

<p>Thank you for your attention.</p>]]></content><author><name></name></author><category term="django" /><category term="python" /><category term="conferences" /><summary type="html"><![CDATA[A written guide to my recent EuroPython talk on modern Python web deployments.]]></summary></entry><entry><title type="html">EuroPython 2026 Recap</title><link href="https://wsvincent.com/europython-2026/" rel="alternate" type="text/html" title="EuroPython 2026 Recap" /><published>2026-07-21T02:56:06-04:00</published><updated>2026-07-21T02:56:06-04:00</updated><id>https://wsvincent.com/europython-2026</id><content type="html" xml:base="https://wsvincent.com/europython-2026/"><![CDATA[<p><a href="https://ep2026.europython.eu/">EuroPython</a> just concluded in Krakow, Poland, and I was there as a member of the <a href="https://www.jetbrains.com/pycharm/">PyCharm</a> team, manning a booth for three days, giving a <a href="https://programme.europython.eu/europython-2026/talk/AGX8D9/">talk on deployment</a>, and more. This post is an image-filled recap of the full week.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026/badge.JPG" alt="Badge" /></p>

<h2 id="tuesday-july-14">Tuesday, July 14</h2>
<p>I departed from Boston Monday evening and arrived, via Munich, in Krakow on Tuesday afternoon. Krakow and indeed Poland itself was a first for me, and I can report that the location was lovely. The conference venue (seen here on the right in the photo) was along the river, and there was even a small amusement park and hot air balloon nearby. The <a href="https://ep2026.europython.eu/speakers-dinner/">Speakers’ Dinner</a> on Tuesday evening was just a short walk from the venue.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026/balloon.JPG" alt="Balloon" /></p>

<p>Across the street from the venue is the Wawel Royal Castle, which is mostly open, meaning you can walk around it freely.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026/castle.JPG" alt="Castle" /></p>

<p>The Speakers’ Dinner was on top of a local building along the river, and downstairs on the grass there was a huge screen setup to show the France vs Spain World Cup semifinal game. I didn’t manage a picture of the screen itself, but here is the view looking down from the venue.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026/speakers_dinner.JPG" alt="Speakers Dinner" /></p>

<p>Despite my jet lag, it was a lot of fun to see familiar and new faces at the event. By accident, I walked over from a nearby hotel with none other than <a href="https://en.wikipedia.org/wiki/Guido_van_Rossum">Guido van Rossum</a> himself. <em>Both of us</em> were required to show a speakers’ dinner QR code at the door, which I found amusing.</p>

<p>At the venue, I ended up having a good conversation on the balcony with <a href="https://programme.europython.eu/europython-2026/talk/NWLBJC/">Katie Bickford</a> and <a href="https://blog.europython.eu/tag/ep2026/">Imogen Wright</a>, who didn’t mention they were giving the opening keynote the next morning!</p>

<p>I ended up leaving after two hours, sooner than I would have liked, but having been up for well over 24 hours at this point, my bed was calling.</p>

<h2 id="wednesday-july-15">Wednesday, July 15</h2>

<p>PyCharm had a full team at the conference. In this picture are four Developer Advocates (myself included), our Head of Python Ecosystem, PyCharm’s Product Manager, and a Technical Content Creator who is, at this writing, now an AI Developer Advocate.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026/pycharm_team.JPG" alt="PyCharm Team" /></p>

<p>We had demos to give, swag to give away–sunglasses, tote bags, and a raffle for a longboard–as well as a beading station that proved quite popular. PyCharm continues to add new features, especially around typing, and ever-more AI options for agentic workflows, supporting basically every model out there.</p>

<p>Also, <a href="https://blog.jetbrains.com/pycharm/2026/07/what-s-new-in-pycharm-2026-2/">PyCharm 2026.2 came out today</a>, so you can read all about it on the official blog post and video if you are interested.</p>

<p>Back to the conference, things started off with an opening discussion, ably led in part by my colleague Jodie Burchell (seen here).</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026/beginners_day.JPG" alt="Beginners' Day" /></p>

<p>I didn’t manage to take a picture of Imogen’s incredible keynote, but I <em>highly</em> recommend you give it a watch once all videos are up on the <a href="https://www.youtube.com/@EuroPythonConference">EuroPython YouTube channel</a>.</p>

<p>During the day, traffic to our booth and those of other sponsors was robust.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026/kristel_booth.JPG" alt="Kristel at the Booth" /></p>

<p>I was pleased that Django had its own community booth downstairs, showcasing the new 6.1 release in a few weeks.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026/django_booth.JPG" alt="Django Booth" /></p>

<p>And Paolo Melchiorre, one of the organizers of <a href="https://djangomed.eu/">Django on the Med</a>, development sprints for Django, was tirelessly promoting the event.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026/djangoonthemed.JPG" alt="Django on the Med" /></p>

<p>There were talks all day long, although aside from the opening keynote, I was mostly at the PyCharm booth.</p>

<h2 id="thursday-july-16">Thursday, July 16</h2>

<p>Thursday morning was another day of talks. You can see just one of the beautiful rooms here, in this photo.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026/room.JPG" alt="Talk Room" /></p>

<p>My own talk was in the late morning, on Python Deployment, and I was very pleased–and somewhat surprised–that the room was completely full well before I started.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026/talk.PNG" alt="Will's Talk" /></p>

<p>There was also a lively q&amp;a afterward, as it turns out deployment is still an interesting topic to everyone. The official video of the talk will be up in a few weeks.</p>

<p>EuroPython serves as a large umbrella for many other Python-related groups to meet and host events, most notably PyLadies, which had <a href="https://ep2026.europython.eu/pyladies/#pyladies-lunch">a lunch</a> right after my talk. I was invited to attend. My colleague Jodie was MCing the event.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026/pyladies_lunch.JPG" alt="PyLadies Lunch" /></p>

<p>And Kristel from the PyCharm team also stood up and spoke.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026/pyladies_kristel.JPG" alt="PyLadies Kristel" /></p>

<p>At the end of the day, there was a <a href="https://ep2025.europython.eu/social-event/">EuroPython Social Event</a> that was well-attended, but I had had a full day and opted instead for a quiet dinner and stroll along the river. It is light until after 9pm in the summer, and the riverbank is full of people sitting, eating, having a drink, and generally enjoying the pleasant weather.</p>

<h2 id="friday-july-17">Friday, July 17</h2>

<p>Friday morning I managed a 6am run with my ultramarathon colleague, Galina. Then it was off to the PyCharm booth again at 8am and a final day of speaking with attendees and sneaking off to a few more talks, when possible.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026/strava.png" alt="Strava" /></p>

<p>Paul Everitt from PyCharm gave the closing keynote, which I, again, highly recommend watching when it is available online. It was about the Python community, AI, and really all the current things shaking up our ecosystem.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026/closing_keynote.JPG" alt="Closing Keynote" /></p>

<p>In the evening, I had dinner with Paolo and Galina, finding a really nice steak restaurant in the old town.</p>

<p><img src="https://wsvincent.com/assets/images/europython-2026/PaoloGalina_Krakow2026.jpg" alt="Paolo, Galina, Will" /></p>

<p>After that, it was drinks with many people and recapping the full event.</p>

<h2 id="saturday-july-18">Saturday, July 18</h2>

<p>My flight was in the early afternoon, which meant I could wake up when I felt like it (turns out that’s 7am), go for a run, have breakfast, pack up, and share a cab with a colleague to the airport. It was a <em>long</em> way back flying against the jet stream via Frankfurt, but I made it home in the evening Boston time.</p>

<h2 id="takeaways">Takeaways</h2>

<p>My quick takeaway from the event was that whereas a year ago the discussions were around AI, which models to use, and whether it was “a thing,” this year it is clear to everyone that AI is here to stay in one form or another. Agentic workflows are all the rage and the focus is less on a specific model and more on “the harness” that feeds information to it and prompts it to try again and again until some measure of success is achieved.</p>

<p>There are built-in agents available like Claude Code and Codex, and more complex options via OpenClaw, but developers at larger companies all reported quite complex internal tooling. The gulf I see emerging is between developers using off-the-shelf tools and those in enterprises who are fully committed to agentic workflows and, presumably, are seeing better results.</p>

<p>The tricky thing with AI is that there are <em>still</em> no truly objective ways to measure performance. Costs can be measured and are increasing. But actual performance is much more difficult. Yes, there are benchmarks, and everyone tries to craft their own evals–though good luck finding a common definition of what “evals” really means–but really, it’s all based on feel and vibes, which are unsettling metrics for most programmers.</p>

<p>PyCharm continues its mission to support as many workflows as possible. That means virtually any model or agent you want to use is supported. The IDE itself is, to me, more important than ever if a developer wishes to add value as a human-in-the-loop here, but it remains to be seen how much developers will even read most code, let alone write it.</p>

<p>I have my own reservations about a full shift towards agentic-all-the-things while relying on integration tests, but it is clear that is where things are headed at the moment in many companies. At the same time, senior Python engineers are urging caution, pointing out that fundamentals and human judgment still do matter, maybe more than ever, despite this shift towards handing more and more of the workload over to AI.</p>

<p>I’m curious to see what kind of conversations happen next year at this event, which will most likely again be in Krakow.</p>]]></content><author><name></name></author><category term="python" /><category term="ai" /><category term="conferences" /><category term="django" /><summary type="html"><![CDATA[Seven days of sponsor booth, talks, sprints, and hallway chats.]]></summary></entry><entry><title type="html">LLM-Inspired Development</title><link href="https://wsvincent.com/llm-inspired-development/" rel="alternate" type="text/html" title="LLM-Inspired Development" /><published>2026-06-15T06:28:42-04:00</published><updated>2026-06-15T06:28:42-04:00</updated><id>https://wsvincent.com/llm-inspired-development</id><content type="html" xml:base="https://wsvincent.com/llm-inspired-development/"><![CDATA[<p>One of my favorite takeaways from the Django Chat podcast last fall was when Adrian Holovaty, one of the original creators of Django, discussed the recent case of <a href="https://djangochat.com/episodes/django-20-years-later-adrian-holovaty-g7z78kc0">LLM-Inspired Development</a>.</p>

<p>You can see a short clip of it here:</p>

<div class="short-video">
  <iframe src="https://www.youtube.com/embed/IbzCV-D79To" title="Adrian Holovaty on LLM-Inspired Development" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" referrerpolicy="strict-origin-when-cross-origin" allowfullscreen="" loading="lazy"></iframe>
</div>

<p>Basically, users of his service, <a href="https://www.soundslice.com/">SoundSlice</a>, kept asking for a new feature because ChatGPT hallucinated that it existed. And so, eventually, Adrian and his team built it!</p>

<p>I just had my own version of that with my personal site. I’m giving a talk at <a href="https://ep2026.europython.eu/session/deploying-python-web-apps-in-2026">EuroPython</a> next month, and after creating all my slides–from scratch, by the way–I asked Claude to review them for feedback. One thing it mentioned was that the closing slide didn’t exist and it volunteered a suggestion for what it should look like.</p>

<p><img src="https://wsvincent.com/assets/images/llm-inspired/claude_slides.png" alt="Flight Map" /></p>

<p>Now there’s a lot that’s mediocre about this slide. But the interesting part was the suggested URL at the bottom: <code class="language-plaintext highlighter-rouge">wsvincent.com/talks</code>.</p>

<p>I have long taken the time to post my conference talks over the years on the <a href="https://wsvincent.com/about/">About page</a>, but admittedly it’s at the bottom and is probably hard to find. So, inspired by Claude, I’ve now added a <a href="https://wsvincent.com/talks/">Talks page</a> that features them. And while I’m not going to use this slide to conclude my talk, I will link to the new talks page in it.</p>

<p>To me, this episode highlights a positive use of LLMs: generating ideas. Notice I didn’t just paste my talk prospectus into Claude and use that. Granted, I was curious and tried, and it was not anything I’d be willing to present. But if you already put in some work, maybe use Claude or another LLM for research and actively guide the process; there is real value to be had if you are otherwise doing all the work yourself.</p>

<p>In this case, Claude giving me boilerplate advice to just have the talks at a <code class="language-plaintext highlighter-rouge">/talks</code> page was just the push I needed to do the obvious thing.</p>]]></content><author><name></name></author><category term="ai" /><category term="python" /><category term="llms" /><category term="django" /><category term="jekyll" /><summary type="html"><![CDATA[How Claude inadvertently suggested new features for my personal site.]]></summary></entry><entry><title type="html">Please add an RSS Feed to Your Site</title><link href="https://wsvincent.com/please-add-feed-to-django-site/" rel="alternate" type="text/html" title="Please add an RSS Feed to Your Site" /><published>2026-05-27T12:57:06-04:00</published><updated>2026-05-27T12:57:06-04:00</updated><id>https://wsvincent.com/please-add-feed-to-django-site</id><content type="html" xml:base="https://wsvincent.com/please-add-feed-to-django-site/"><![CDATA[<p>I’ll start with the main take here: <strong>please, for the love of God, add an RSS feed to your personal site</strong>.</p>

<p>If you are using Django to host your site, you can do it in minutes.  Here is a <a href="https://learndjango.com/tutorials/django-rss-feed-tutorial">simple and up-to-date tutorial</a> on how to do it. Django makes it very easy.</p>

<p><img src="https://wsvincent.com/assets/images/use-rss/yellsatclouds.jpg" alt="" /></p>

<p>Why does RSS matter more than ever in 2026? As a co-writer of the <a href="https://django-news.com/">Django News Newsletter</a> for over 7 years, I am constantly looking for high-quality posts from people in the Django and broader Python communities. And yet, these days, it is harder than ever to find new content or even keep up with the efforts of people we’ve featured in the past. Social media is–how to say this kindly–rough in terms of uncovering content. These days, it feels like I spend more time sniffing out AI-written content than just exploring. It is exhausting.</p>

<p>My favorite way to consume the internet is via RSS. It’s a way to keep up with what friends and people I respect are saying. There are still multiple good free or low-cost options out there. Personally, I use <a href="https://betamagic.nl/products/newsexplorer.html">News Explorer</a> based on Carlton Gibson’s recommendation.</p>

<p>I should note that it is fine and indeed often helpful to use AI/LLMs when you write, provided you are still doing the work! They are wonderful tools to copyedit a post, maybe even suggest ideas that are worth exploring, but if you just enter some short prompts and copy-paste the result as your own work, then no. Discerning readers can tell, but more importantly, you are depriving yourself of the main benefit of writing, which is that it focuses your own thoughts. Sheena O’Connell had a good <a href="https://preludetraining.substack.com/p/how-im-using-llms-to-share-my-ideas">recent post on using LLMs to share ideas</a> that is worth reading if you want more context.</p>

<p>So how else do <a href="https://jefftriplett.com/">Jeff Triplett</a> and I curate our newsletter? The official <a href="https://www.djangoproject.com/weblog/">Django blog</a> has good content, and there are active efforts to add even more. The <a href="https://www.djangoproject.com/community/blogs/">Django Community RSS</a> feed is a little buggy, but still unearths good articles. I always check there, too, in case I’ve missed something.</p>

<p>But to go back to my opening point, without name-shaming anyone, there are more than a few Django contributors whose content I might miss if I don’t see it on Mastodon, or maybe on LinkedIn, or maybe someone mentions it to me.</p>]]></content><author><name></name></author><category term="django" /><category term="education" /><category term="python" /><category term="internet" /><category term="web" /><summary type="html"><![CDATA[Why syndication feeds are having a moment in 2026.]]></summary></entry><entry><title type="html">PyCon US 2026 Recap</title><link href="https://wsvincent.com/pyconus-2026-recap/" rel="alternate" type="text/html" title="PyCon US 2026 Recap" /><published>2026-05-20T02:56:06-04:00</published><updated>2026-05-20T02:56:06-04:00</updated><id>https://wsvincent.com/pyconus-2026-recap</id><content type="html" xml:base="https://wsvincent.com/pyconus-2026-recap/"><![CDATA[<p>I just spent seven days in Long Beach, California, for <a href="https://us.pycon.org/2026/">PyCon US 2026</a>, the biggest Python conference in the United States and one of the largest in the world. Many of the people behind the language (including the creator, Guido van Rossum), packages, frameworks, and the community were there, enjoying two days of tutorials, three days of talks, two days of sprints, and lots in between.</p>

<p><img src="https://wsvincent.com/assets/images/pyconus-2026/django_booth.jpeg" alt="Django Booth" /></p>

<p><em>This picture is from the Django community booth, and if you look closely, you’ll see Guido himself, two Django Fellows, a Django President, Django Board members, DjangoCon organizers, EuroPython organizers, and more.</em></p>

<p>Long Beach is a fantastic venue for a conference: good weather, on the water, excellent food, and a large nearby airport. I had a six-hour flight from Boston on Thursday.</p>

<p><img src="https://wsvincent.com/assets/images/pyconus-2026/plane.png" alt="Flight Map" /></p>

<p>Here are some more photos to get a feel for the place.</p>

<p>The Queen Mary is an old ocean liner that is now a hotel and venue. A good friend was married on it a number of years ago. It now sits permanently in the harbor.</p>

<p><img src="https://wsvincent.com/assets/images/pyconus-2026/boat.JPG" alt="Queen Mary" /></p>

<p>There is also a cool bridge downtown that lets you walk from the beach area towards the convention center and main streets.</p>

<p><img src="https://wsvincent.com/assets/images/pyconus-2026/bridge.JPG" alt="Bridge" /></p>

<p>And even a sign with sister cities.</p>

<p><img src="https://wsvincent.com/assets/images/pyconus-2026/sign.JPG" alt="Sign" /></p>

<h2 id="thursday">Thursday</h2>
<p>I arrived on Thursday in time for the evening opening reception, though I wish I had arranged my schedule to also attend the <a href="https://us.pycon.org/2026/events/education-summit/">Education Summit</a> earlier in the day.</p>

<p>At the check-in booth, I picked up my official badge. Custom ribbons are quite the thing at these events; I only had one for “Sponsor” initially, though.</p>

<p><img src="https://wsvincent.com/assets/images/pyconus-2026/badge.JPG" alt="Badge" /></p>

<p>We had a full team at the <a href="https://www.jetbrains.com/pycharm/">PyCharm/JetBrains</a> booth, including Developer Advocates (like me), product managers, marketing, and engineers. For years, PyCharm has been a big sponsor of Python: attending PyCon US and EuroPython, running the <a href="https://blog.jetbrains.com/pycharm/2025/08/the-state-of-python-2025/">annual Python survey</a>, supporting frameworks like Django, and so on.</p>

<p>A shoutout to my colleague Paul Everitt, who was presented with the <a href="https://pyfound.blogspot.com/2026/05/announcing-psf-community-service-award.html">PSF Community Service Reward</a>, a well-deserved honor given his decades of work to advance Python.</p>

<p><img src="https://wsvincent.com/assets/images/pyconus-2026/pycharm_team.jpeg" alt="PyCharm Team" /></p>

<p>There is always a lot of excitement at the opening reception, and this year was no exception. Our booth had several demos around new AI features and an AI agent sticker leaderboard (more on that shortly).</p>

<p>Here is my colleague, Kristel, demoing a Python game to showcase some PyCharm features.</p>

<p><img src="https://wsvincent.com/assets/images/pyconus-2026/demo.JPG" alt="Demo" /></p>

<p>Following the reception, PyCharm hosted a small gathering at the nearby <a href="https://maps.app.goo.gl/Kkn7f3XHAKqQTrEv7">Breakers hotel</a> and I had the chance to chat with engineers from Spotify, Pydantic, and other companies about all things, especially AI and how it is being used and misused in professional settings.</p>

<h2 id="friday">Friday</h2>
<p>The next morning, it was up early to work at the sponsor booth starting at 7:30am. We had two book signings with both <a href="https://stefaniemolin.com/about/">Stefanie Molin</a> and <a href="https://ehmatthes.com/">Eric Matthes</a>.</p>

<p>One of our goals this year was to take as many videos as possible of attendees discussing their work, whether or not they use PyCharm. Alex from the JetBrains team was busy the whole time taking photos and filming interviews. Here he is with me and Paolo Melchiorre after a short video session. <em>(Photo courtesy of Paolo.)</em></p>

<p><img src="https://wsvincent.com/assets/images/pyconus-2026/alex.JPG" alt="Photo with Paolo and Alex" /></p>

<p>I spent most of the day talking to attendees and, in spare moments, visiting other sponsor booths. There are financial firms, energy companies, developer tooling companies, aerospace companies, education companies, and much more. If you are at a conference, please do stop by the sponsor booths. They are happy to chat, and I promise they won’t hard-sell you on anything.</p>

<p>After a full day on my feet and talking with people, I had a quiet night and rested for the next day.</p>

<h2 id="saturday">Saturday</h2>
<p>More of the same: sponsor booth, recording interviews, and learning from attendees. My colleague, Galina, was demoing a new PyCharm AI feature in beta, which was very interesting to see how people reacted. The UI of AI models is still in its infancy, and PyCharm is one of many companies experimenting with how to make them easier and more powerful to use. I hope this new interface makes its way into a future PyCharm update.</p>

<p>Near the back of the booth, we had a sticker board asking visitors about the use of AI agents. As you can see, there is a lot of Claude Code, followed by Copilot, Gemini, Codex, and Junie. Cursor didn’t rate very high, and Cline was practically absent. Interesting anecdotal evidence for us to consider, as all of them are easily available within PyCharm and JetBrains IDEs.</p>

<p><img src="https://wsvincent.com/assets/images/pyconus-2026/agent_board.png" alt="Agent Board" /></p>

<p>In the evening was the annual PyLadies auction and dinner. It is quite the event. I attended with Jon Gould from Foxley Talent and my colleague, Mark Smith. Many items are auctioned off to raise funds for PyLadies, including a Django sweatshirt (Guido not included).</p>

<p><img src="https://wsvincent.com/assets/images/pyconus-2026/pyladies.jpg" alt="Sweatshirt" /></p>

<p>JetBrains donated several items as well: embroidery from my colleague Jodie, a gift bag with items from PyTV earlier in the year, and a custom skateboard to fit with the Southern California theme.</p>

<p><img src="https://wsvincent.com/assets/images/pyconus-2026/skateboard.jpg" alt="Skateboard" /></p>

<p>The event carries on late into the evening, though I retired after several hours to rest up for the next day and a 6am run with colleagues.</p>

<h2 id="sunday">Sunday</h2>
<p>If you head to the water, you can turn left and run along Alamitos Beach for a good five miles, or turn right and head up the Los Angeles River. With Mark and Galina, we decided to head right; after snaking our way around the fake lighthouse and other buildings, we landed on the bike path that goes quite a way along the river. For Mark and me, this was a 10k run; Galina continued on for at least double that, but she is a competitive ultra-runner, so in a whole other bracket of “runner.”</p>

<p><img src="https://wsvincent.com/assets/images/pyconus-2026/run.JPG" alt="Run Group" /></p>

<p>After a quick breakfast, it was time for coffee at <a href="https://www.recreationalcoffee.com/">Recreational Coffee</a>—a place I managed to visit every day of my trip!</p>

<p><img src="https://wsvincent.com/assets/images/pyconus-2026/rec.JPG" alt="Rec Coffee" /></p>

<p>Then I raced over to catch the morning keynote from <a href="https://research.google/people/107947/?&amp;">Amanda Casari</a>, who works at Google Open Source. She gave a very inspiring and thoughtful talk about open source, AI, and how the two can meet. In particular, she highlighted the Google Summer of Code, which has long supported Django and directly led to many new features.</p>

<p><img src="https://wsvincent.com/assets/images/pyconus-2026/amanda.JPG" alt="Amanda Casari" /></p>

<p>I attended talks all day, as well as an open space on the Python survey. Open spaces are where groups come together to discuss CPython internals, web frameworks, or even the PSF official survey.</p>

<p>The evening highlight was a big party thrown by <a href="https://www.revsys.com/">RevSys</a>, a Django and AI consultancy. There were a lot of people there and it was a ton of fun, marking the end of the talks portion of the conference.</p>

<p><img src="https://wsvincent.com/assets/images/pyconus-2026/revsys.JPG" alt="Revsys Party" /></p>

<h2 id="mondaytuesday">Monday/Tuesday</h2>
<p>Monday was the first of two sprint days, during which attendees self-organized into groups to work on projects. It’s an incredibly welcoming and exciting time. You’ll find the long-time maintainer of Flask at a table helping both first-time contributors and experienced members with tickets. There will be impromptu discussions around security, as happened between the Flask and Django teams. And the PSF Board of Directors even finds time to meet during all of this.</p>

<p>Here’s my colleague Cheuk, who was the Sprints chair, standing by the large board of topics available for sprints.</p>

<p><img src="https://wsvincent.com/assets/images/pyconus-2026/cheuk.JPG" alt="Cheuk" /></p>

<p>And it’s a chance for a slightly slower pace than the talks themselves. For example, I finally had a chance to chat briefly with <a href="https://blog.pamelafox.org/2026/05/pycon-2026-reflections.html">Pamela Fox from Microsoft</a> and <a href="https://djangochat.com/episodes/pycon-us-2026-elaine-wong-jon-banafato">Elaine Wong</a>, who was the Chair of PyCon this year.</p>

<p><img src="https://wsvincent.com/assets/images/pyconus-2026/sprints.JPG" alt="Sprints" /></p>

<h2 id="wednesday">Wednesday</h2>
<p>I headed to LAX for the flight home early Wednesday morning, accompanied in the cab by PyCharm’s Head of Python Ecosystem, Mark Smith. It was a good opportunity to recap the conference, share anecdotes, and plan for the next few months with PyCharm.</p>

<p>More than any prior PyCon I’ve attended, AI was the subtext of nearly every conversation — at the booth, in the hallways, in Amanda Casari’s keynote. The community is figuring out what it means, and there’s no better place to watch that happen in real time. Whereas last year there was palpable excitement about AI and uncertainty about how it would change programming, this year the vibes were closer to acceptance of it as a tool that’s not going anywhere, coupled with unease about both job cuts and the lack of knowledge amongst junior programmers. I expected more talk of agentic programming patterns, and while that happened, it was rarer than I anticipated. The bigger shift was people moving from “Is AI a thing?” to “How is AI going to change everything?”</p>]]></content><author><name></name></author><category term="python" /><category term="ai" /><category term="conferences" /><category term="django" /><summary type="html"><![CDATA[Seven days of sponsor booth, talks, sprints, and hallway chats.]]></summary></entry><entry><title type="html">Are Written Coding Tutorials Dead?</title><link href="https://wsvincent.com/future-written-tutorials/" rel="alternate" type="text/html" title="Are Written Coding Tutorials Dead?" /><published>2026-05-04T02:56:06-04:00</published><updated>2026-05-04T02:56:06-04:00</updated><id>https://wsvincent.com/future-written-tutorials</id><content type="html" xml:base="https://wsvincent.com/future-written-tutorials/"><![CDATA[<p>You can see the question of this essay in the title. It’s something my small support group of fellow programmers and writer/educators message together about regularly. What used to be the provenance of enviously looking at someone who had cracked the code and managed to attract real eyeballs and, therefore, a sales tunnel for a paid course or book, has now become fully existential.</p>

<p>More people than ever are trying to learn how to code, whether through traditional pathways, vibecoding, YouTube videos, or whatever. And yet, at least when it comes to written content, there is almost no SEO path left. Google is no longer interested in showing external links and defaults to AI-generated Gemini answers, while, increasingly, people don’t even use search at all but start with ChatGPT or Claude for such questions.</p>

<p>Look at what’s happened to Stack Overflow, the grandparent of question-and-answer programming sites.</p>

<p><img src="https://wsvincent.com/assets/images/coding-tutorials-dead/so_traffic.png" alt="StackOverflow Chart Traffic" /></p>

<p>The takeaways I see are a serious rise from 2008 to 2014, a leveling off until 2018, a small decline offset by a sharp jump around Covid in 2020, and, since 2022, a dramatic falling off to the point where the site is basically dead. We all know what happened in late 2022 (hello ChatGPT).</p>

<p>This pattern of growth and decline is mirrored in my own metrics and those of most of my content-creator friends. For a brief halcyon phase, roughly 2015-2020, you could write a high-quality, in-depth post on a programming topic and plausibly rank in the Google top 10 for it, receiving solid traffic to then divert elsewhere.</p>

<p>When I look at my <a href="https://wsvincent.com/year-in-review-for-2018/">2018 year in review</a>, what jumps out is that I wrote 72! new tutorials, largely because I was learning new things, but also because each one was a chance to rank highly and then be marketed for my books. In that same year, I went from 1,500 weekly unique visitors to 20,000 and was blindly thinking about what would happen if trends persisted.</p>

<h2 id="book-sales">Book Sales…</h2>
<p>These days are different. I still have book sales, mostly through <a href="https://learndjango.com">LearnDjango.com</a> and Amazon, though Amazon has undergone a remarkable enshittification process as well that is probably worth its own blog post. But tl;dr Amazon used to use search to show the best results; now it is just like Google, search is completely paid for, and results are about 8/10 ads versus organic links. Oh, and if you pay for ads with Amazon (wink, wink) your organic search results will do better. In short, it is a mess. I used to spend some money on Amazon ads in the early days. It was worth it. I completely stopped 2 years ago because prices spiked 5x and results dropped 3x over the same period.</p>

<p>How is it plausible for any company to buy $3-5 per click ad words on “django book” when they are selling a $40 book, meaning $20 profit, and when only a fraction of people who click buy? The math doesn’t math. At least for me.</p>

<p>And so I still have books on Amazon and provide links via <a href="https://learndjango.com">LearnDjango.com</a>, but I don’t focus on that area at all. Because it’s not worth it.</p>

<h2 id="llm-results">LLM Results</h2>
<p>Modern LLMs continue to improve across most benchmarks and in my everyday use. I’m not an AI-luddite at all; in fact, I’ve been using them daily for 3 years now, both ChatGPT and Claude. They have their place for sure. But I will admit I have deep reservations about how long this will persist.</p>

<p>All current models are trained on a free/stolen corpus of human data. Both written tutorials and code. Now, both are dominated by AI. We shall see how good the models really are now that they are consuming themselves.</p>

<p>Then there is the cost structure, where token prices are already rising and need to keep increasing a whole lot more for current frontier galaxy models that burn billions in training and inference to serve.</p>

<p>But for the moment, you can get great results. You don’t need me to tell you how to do something in Django. Just ask Claude to write a custom blog post in the style you prefer on the topic. You’ll get 70% of it right. Yes, some issues, no pretty screenshots, and no guarantee it is, you know, accurate. But better than searching around on the almost-dead internet for info.</p>

<h2 id="the-future">The Future</h2>
<p>What does all this mean? Maybe AI/LLMs really do figure out how to make software developers and technical writers obsolete. It’s possible. I’m biased against it for a number of reasons, but I would never bet against technological progress entirely.</p>

<p>The real question is: how much value does a human with a reputation and a voice actually add? If you read something from me, or hear me on a podcast, or see me in a video, I’m putting my reputation on the line each and every time. I’m also, hopefully, communicating based on many years of trying to explain topics clearly to people and on my many years as a programmer myself. The same goes for all the other technical content creators out there.</p>

<p>For the time being, I see no point in writing anything not behind a paywall that will just be consumed by an LLM with no attribution. So I’ll continue to write personal pieces, like this, but deep technical dives that take days or weeks? No. Those will be in books, private videos, or in person. It’s not worth it.</p>

<p>Many of my friends aren’t even that lucky. They are planning to just close up shop. No point in updating good materials or even continuing to host what they have. So there goes even more foundational knowledge that the internet and LLMs could have trained on.</p>

<p>I’ll end by making the point that fighting technological change is not a worthy mission. I’m not anti-technology at all. Even if all progress stopped today, LLMs are an incredibly valuable tool that has meaningfully improved my software development. But I harbor real doubts about whether it can continue to improve, suspect it might actually get worse given eroding training data, and I think a whole generation of newcomers is going to be more confused than ever because they are entrusting their educations to a next-token predictor rather than a proper teacher.</p>]]></content><author><name></name></author><category term="programming" /><category term="teaching" /><category term="python" /><category term="ai" /><summary type="html"><![CDATA[Have LLMs truly killed the point of human-written coding tutorials?]]></summary></entry></feed>