2025/12/11
\nAt the front of my website stands a file, robots.txt, and you'll find one in the root directory of most websites. Its job is to tell bots (automated scripts and browsers) where they can and can't go within the website, and it's as useless as a dead guard dog.
It's very much like the Do Not Track (DNT) header: it's a law with no enforcement — a worthless gesture that will only be adhered to by voluntarily honorable and trustworthy parties, exactly the type of people who wouldn't want to track you in the first place. In fact, there's nothing stopping a morally bankrupt organization from deciding to track you more scrutinously when you enable DNT headers, perhaps under the premise that they think you have something to hide. I deeply appreciate an unobtrusive notification when I first enter a website saying my DNT header will be honored; I feel more comfortable sticking around.
\nMy robots.txt just has one instruction, and it's identical to the one reddit swapped to in 2024. It tells bots to fuck off from the entire site.
The reason reddit does it is obvious: it allowed them to take all of the data contributed by redditors over decades and auction it off to different AI firms, data brokers, and search engines under private (profitable) agreements. Instantaneously disabling dozens of third-party reddit reading apps, upon which some independent developers had built their livelihoods, was a welcome side effect, as all of the value they added to the reddit experience over the years, in good faith, was clawed back to shareholders who could now negotiate how to sell our intellectual property.
\nThe reason I do it is out of an indiscernable mix of spite and principle. It doesn't work, and it elicits a benign but annoying side effect of receiving automated emails about my trash SEO and how I can pay someone to improve it.
\nI don't want code that I wrote, the product of years of expensive education and arduous trial and error, to be used to improve the quality of software that is openly threatening my job security.
\nI don't have a solution here. Anyone who hosts a website will understand the plight: even if your dead guard dog adamantly declares the entire house off-limits, the most dishonorable, unwanted scrapers you cared most about deterring will walk past unobstructed, scraping to their iron hearts' content and filling your forms with PHP debugging code.
\nThis state of affairs leaves the deterrence of unwanted bots up to you. A modern website owner must understand and account for the inevitable fact that bots will crawl their site, and they'll likely pay for that traffic just the same as legitimate visits by humans. Techniques like Cloudflare's "AI Labyrinths" are intriguing, as are putting sneaky, invisible form inputs to trick a bot into ousting itself, but these are cat and mouse games. An author of a crawler can account for these tactics once enough is known about them, and then it's on to the next mitigation technique, ad infinitum.
\nAt the time being, the onus is on the site owner to erect a fence of bot-bouncing safeguards, a hurdle that further impedes the pursuit of an open web.
\n", "url": "https://johntorsten.com/posts/my-dead-guard-dog", "title": "My Dead Guard Dog", "date_modified": "2025-12-11T00:00:00.000Z" }, { "id": "https://johntorsten.com/posts/the-cardlendar", "content_html": "2026/01/01
\nThere are 52 cards in a conventional playing card deck and 52 weeks in a year.
\nEach time you thoroughly shuffle a deck of cards, the order of the cards is very likely the first arrangement of its kind in human history.
\nThe Cardlendar is a simple practice of associating a playing card with each week:
\nThe card is an arbitrary label, but it's entirely unique to you. Even if everyone on earth participated, the saga of the cards would unfold differently to each of us.
\nI started doing this in November of 2025 with a sliver of the deck remaining. My last two cards for the year were the nine of hearts followed by the nine of diamonds. It's not a practice I think will change your life, but maybe it will help if you find the weeks blending together like I did.
\n", "url": "https://johntorsten.com/posts/the-cardlendar", "title": "The Cardlendar", "date_modified": "2026-01-01T00:00:00.000Z" }, { "id": "https://johntorsten.com/posts/own-your-feed", "content_html": "2026/01/21
\nThis is a post imploring you to use a feed reader.
\nI use Feeder, but you can use whatever you like. Stringer looks neat, as does Miniflux, and Tiny Tiny RSS looks cute.
\nA feed reader is an elegantly simple way of keeping up with stuff online. Give it a bunch of links to web feeds for your favorite people and organizations and it shows you their latest posts.
\nFeed readers operate on open standards, meaning there are many to choose from, and you can switch between them at any time by saving a file of your feeds that can be exported and imported. Because it's open, any website can provide a feed, from your favorite record label to your local police department.
\n\n\nFeeds use XML format, so they'll end in ".xml". If you don't know where to start, copy and paste this ScienceDaily "all" feed link (no affiliation). Right click (or tap and hold) that link, copy it to your clipboard, and paste it into the field.
\n
\n\nFeeds can be tricky to find, but they can be lurking in places you don't expect. Look for a WiFi-looking wavy symbol, or try going to the homepage of what you're trying to follow and adding "/feed.xml" or "/feed" to the end of the URL in your browser. Browsers often download the XML file instead of displaying its text to you, which can be a bit confusing.
\n
Most applications will then ask you to categorize and configure your feeds. I recommend at least quarantining news to its own category.
\nI envision a world in which my friends' feeds are nestled nicely in their own "Friends" category in my Feeder, and I could check up with them on my phone every once in a while, but I must concede that most people my age prefer Instagram, and getting them off is like pulling teeth.
\nI believe the biggest hurdles can be overcome with good tools and services that abstract the process of blogging, as well as effective language (the word "blog" might be a hurdle to widespread adoption; it has a bit of a dated context, right?). "Feed" has become a ubiquitous term; it has passed the public taste test. Let's reclaim it to mean what it's intended to mean: a source of links to new stuff on the web. It's stuff you care about, and chose to follow.
\nThe most obvious problem is how difficult it is to set up your own web feed.
\nRight now, to set up a website and feed whose content you truly own and isn't locked into a proprietary ecosystem, you have to:
\nThat's prohibitive! Doing all that is time-consuming, costly, and requires technical background, but anyone should be able to have their own website. The web has been alive for thirty years! Such a fundamental process shouldn't be so complicated and barrier-ridden.
\nPlatforms like Substack or Medium abstract this process, and at the time of writing, allow you to export all your posts and uploaded content. However, these are companies in the tech sector, and thus are at the whims of their investors. They must make money, and continue to make more money. They retain the capacity, at any time, to censor content they find disagreeable, change any part of the infrastructure that delivers your posts to users, make deals to analyze and sell accumulated data to other firms, and update their terms of service. This makes them unresilient long-term solutions.
\nUsing a feed reader is liberating. There are no likes. You can follow and unfollow anyone at anytime, and it's an open protocol. It can be used for everything from public service announcements to podcasts. Your local library might have a feed for their local events. If an update happens to your feed reader that you don't like, or you want to switch to another, you can export your feeds in OPML and import that file into any other reader.
\nCompared to social networks, feed readers don't lock you in. They don't intersperse ads between your friends' posts. They can't change the user experience or the terms of use whenever they please. They can't spy on you, analyze you, sell your data, nor train their machine learning models on the duration of every swipe, systematically deducing your habits, interests, and weaknesses. It's a night and day difference that accomplishes the same goal. You keep up to date with your friends, but you don't get abused by a multinational corporation in the process.
\nThat being said, it's not perfect, and ideally the process for subscribing to someone's feed shouldn't be as convoluted as how I documented above. Emails have a "mailto" URI scheme which allows tapping an email address link to automatically open your preferred email client and pre-populate the "to" field. Why can't we have the same for feeds? Megacorporations have demonstrated the effectiveness of removing friction at every corner, and the open web needs to catch up.
\nStill, just try it!
\n2026/06/02
\nThis is a rant about "AI": not the technology, nor its politics, but the term, and how we're better off without it.
\nColloquially, "AI" is used to describe:
\nA lot of these innovations have to do with recent, notable leaps in machine learning (which are fascinating, historical, and deserving of being disseminated as public knowledge), but certainly not all. With each miscellaneous technology that has this label tacked onto it, the term grows further diluted, and less clearly tied to this domain.
\nAdd the perverse economic incentive in which adding "AI" to your product offerings can result in 600% stock price surges and the semantic umbrella for this term is set up to expand continually.
\nTechnology impacts all of us. Increasingly, it impacts those who never opt in to its use.
\nhttps://apnews.com/surveillance-digital-cage
\nIn order to have a conversation about any of the above technologies with anyone, be it a stranger, a family member, or your elected representative, we have to agree on what words mean. If we don't, we necessarily get caught up in semantic incompatibility, and ideas aren't successfully communicated.
\nIn a sense, adopting this language makes us dumber, because we can't effectively communicate with each other.
\nTo no one's surprise, 65% of US residents believe they are smarter than average, according to a somewhat recent survey.
\nWhat does this mean in practice? It means when the average American conceives the idea of the "average American", they likely think of someone significantly dumber than reality would reflect.
\nEvery writer or speaker that opts to use "AI" to cater to this imagined average American when a more specific term would be apt contributes to this cycle of industrial condescension: people get less technical, more vague information than what they could feasibly comprehend, and so are deprived of helpful knowledge they could have integrated. Repeated day after day, year after year, doesn't this become a self-fulfilling prophecy? Don't you feel the belittling and debilitating tone of communication around AI?
\nIn conclusion:
\nLet's reject this term entirely. Engage your linguistic gag reflex. It doesn't serve the public.
\nMy intent isn't to police language, but to hopefully have convinced you that this term is unhelpful. It's used to disorient and disable us from comprehending or discussing the impact of new technology that heavily affects our lives.
\nLanguage is so crucial. It's how ideas, thoughts, concepts, and feelings live or die. If you want to talk about ChatGPT, don't call it "AI"; call it an "LLM". That's what it is. If you want to talk about machine learning in general, just use that term! It's specific and communicative! The words we use matter, and using this uncommunicative acronym for everything hinders our ability to engage in discourse and action to control the technologies it alludes to, effectively forfeiting our role in determining their integration, and our future.
\n", "url": "https://johntorsten.com/posts/bad-words", "title": "Bad Words", "date_modified": "2026-06-02T00:00:00.000Z" } ] }