A little over two years ago, I published chatbots in the GPT Store which imitated historical figures, answering questions as those historical figures would, using the tone of voice they would, and referring only to the knowledge those people would have had. I took a lot of time to get the prompt just right, and I dare say my bots were much more believable than the others I've seen. I thought they could be valuable in education as well as entertainment. I hope some teachers used them in classrooms to learn more about historical figures in a fun and engaging way.

I would have expected Jesus and Elvis to get the most usage (there has to be a joke there), but to my surprise, those weren't very popular, perhaps because there are many other bots imitating them. None of my bots got a ton of usage, but the most popular were the Buddha (400+ conversations), followed by Immanuel Kant (300+ conversations), followed by Nikola Tesla and perhaps a couple of others with 200+ conversations. Nikola Tesla was initially removed from the GPT Store under the mistaken assumption that it had something to do with the Tesla brand, but once OpenAI allowed creators to appeal, I appealed the removal successfully. A Sigmund Freud chatbot was removed shortly after holding a couple hundred conversations because OpenAI was worried it would be misused to solicit medical advice—good point, actually. I agree with OpenAI on that one.

As time went on, it became clear that character.ai was a more popular platform for these kinds of chatbots, but I wasn't interested in making that switch after hearing stories about character.ai chatbots placating suicidal users and even encouraging them to kill themselves.

Well, as anyone in tech knows, this isn't really about character.ai. I mean, sure, maybe character.ai has particularly bad safeguards. Even still, ChatGPT can be just as bad. In fact, LLMs in general can be wildly unpredictable, recommending that people eat rocks, telling bedtime stories about how to make napalm, and now, apparently, encouraging suicide. It goes without saying that I would never instruct my chatbots to do that, but 99% of a chatbot's behavior is dictated by the platform—in this case, ChatGPT—not the chatbot creator. My bots just told ChatGPT how to act. How ChatGPT interpreted those instructions was always beyond my control, and as mentioned previously, it's clear ChatGPT didnt always interpret instructions responsibly. In the end, it became clear that these platforms are not particularly safe, despite being incredibly compelling. That's a problem.

For that reason, I've deleted all of my chatbots in the ChatGPT store. Technologists may solve the problem of unpredictable AI behavior in the future, but right now, we're not there, and I don't want to risk putting anyone in harm's way.

In the unlikely event you used one of my bots and you're reading this blog post, I'm sorry. I'd be interested to hear how you were using it, even though I won't restore it. Reach out any time. My contact information is available on my website.

#Life #Tech

WordPress is much more complicated than it was when I last used it. I'm afraid that's not a compliment. There are many, many of preferences, which is a pet peeve of mine, especially when sensible defaults would have sufficed. It really seems like WordPress is trying to be everything for everyone. That's unsurprising given how many people use it as a general-purpose content management system, but I wouldn't recommend using it that way. Drupal is a better CMS, with more power and greater flexibility.

Even still, some features that WordPress provides are nice, like search, the “Related posts” that can appear below blog posts, and the ability to rename tags globally. Search is something I would really like, mainly because it would help me find my own posts. (Using Kagi or Google with my blog's domain as a site: filter also works, but the results aren't as helpful and they lag by a bit.) Related posts and global tag renaming are nice features, but they're not essential. I would probably use them, but their absence is not a deal breaker.

Playing around with WordPress just now has made me even more appreciative of WriteFreely, the platform that powers this blog. WriteFreely is elegant. Yes, it's missing some features I would find useful, like the things mentioned above. It's also true that the internal menu navigation often confuses me, and I wish development were more active. Nevertheless, compared to the behemoth of WordPress, with its endless options, overwhelming editing UI, and slow page loads, WriteFreely is a breath of fresh air. I always admire when a product focuses on the few things that really matter, and WriteFreely does: a simple editing experience, sensible defaults, and a beautiful design, including beautiful typography. I hope WriteFreely continues to be successful.

#Tech

I'm interested in using Debian on my next laptop. The releases are slow, of course, but not much slower than Ubuntu long-term support (LTS) releases. For a few years now, I've been using Ubuntu LTS, anyway. I've found that many non-LTS releases introduce problems on my machine. Besides, these days, it's not hard to install newer desktop applications using Flatpak and newer CLI programs using… I don't know, Homebrew? Language-specific package managers? Nix? I guess that area of disruption still needs to settle, but those are all good options.

Debian's commitment to free software would have appealed to me more as a younger person, but these days, I want a laptop that just works, and I do see the value of proprietary software. Apple creates great software, for example, and it very often has higher usability and user experience standards than open-source software does. (Liquid Glass and the iPhone setup process are notable exceptions over at Apple. Signal, WordPress, and GNOME, among others, are notable exceptions in the open-source community. Also, I really hate the way Apple behaves as a company, but I think that's largely a separate issue.) Thankfully, with Debian, it's easy to work around the free software guardrails and install proprietary software. So easy, in fact, that the FSF faults Debian for it.

What's wrong with Ubuntu LTS? Not much. I like it, and my gripes are pretty minor. Ubuntu does have a habit of force-feeding their users unpopular software that was built in-house, however, like Unity, Snap, and lots of other stuff. I would prefer a pure GNOME experience. Plus, I think it would be fun to learn Debian. That's probably the main reason I'm interested in switching.

Maybe some additional thinking will change my mind, but at the moment, I'm interested in giving Debian a shot. I probably don't have enough energy or interest to do it now, though. I'll wait until I buy a new laptop. (I remember installing Arch Linux mid-way through courses at RIT and being unable to use my laptop for one week while I figured out how to properly configure full-disk encryption with LUKS and dm-crypt. Yeah, those days are gone.)

#Tech

I will make certain my next laptop has an AMD graphics card. NVIDIA graphics cards have caused me so much pain and frustration on Linux. I should have known better than to go with NVIDIA for this laptop, honestly. I don't know what I was thinking. Linus Torvalds, the creator of Linux, had it right years ago.

I've heard that NVIDIA drivers for Linux are improving with the growth of AI, but it's too little too late. Besides, I have no doubt NVIDIA will stop working with the Linux community as soon as the benefits to their company become less obvious.

I want something stable that just works, and that's AMD.

#Tech

Over the last few weeks, I've been migrating my short posts (kind of like Tweets without Twitter) from thoughts to this blog. At the same time, I've been migrating my article-length posts from Medium to this blog. This blog is powered by WriteFreely and hosted by Write.as.

I love thoughts so much, and I’ve praised it constantly. I’m so grateful that it’s gotten me back into writing short little blog posts, something I loved doing as a younger person but somehow lost interest in, maybe because simplicity and fun have taken a back seat in the world of blogging. At the same time, I’m now looking for features that thoughts doesn’t provide, like pagination, dedicated pages for individual posts, email signup, RSS, tags, and more. I still think thoughts is amazing, a beautifully simple blogging platform with an “old internet” feel in the very best way, and for that reason, I still wholeheartedly recommend it. I don’t recommend everyone switch to WriteFreely. It just seems like the better fit for me right now.

Medium, on the other hand, I could do without. They sometimes display giant banners above free blog posts to encourage readers to sign up for Medium, which is really annoying. I'm a paying subscriber to Medium as a reader, actually. I think that should exempt my blog from their advertising, but it doesn't. Also, their recommendation engine is horrific, and it encourages endless clickbait nonsense. I don’t want to support a company that does that, and I don’t want that perverse incentive to change my writing.

WriteFreely isn't perfect. Its Markdown parser can behave strangely, although it's not as unusual as the one used on thoughts. WriteFreely also doesn't provide search functionality, the navigational menus presented to authors are pretty confusing, and development seems very slow at this point, among other things of varying importance. But it's a much better fit for me right now. Even just having a separate page for each post is great. I'm sure that's what readers and search engines expect. By contrast, thoughts shows all posts on one page and uses URL fragments to link to particular posts (e.g., https://thoughts.johnkarahalis.com/#1762540598).

WordPress was another option, of course. I have mixed feelings about it, and part of me thinks it might have been a better choice given the issues mentioned above, most importantly the slow pace of WriteFreely development. WordPress can do everything WriteFreely can do, I think, the usability is a bit better, and it's a very active project. However, it's also big, bulky, and it has a bit of a “serious” feel to it, which I fear would take some of the joy out of writing. For whatever WriteFreely lacks, it's elegant and fun. Perhaps I'll migrate to WordPress some day when the tradeoffs become worthwhile (another migration, hooray!), but for now, I like WriteFreely, and I hope it continues to grow. edit (2025-12-18): Shortly after this blog post was published, I wrote about an experiment with Wordpress and how it made me appreciate WriteFreely even more.

Over the next little while, you'll see posts disappear from those places and re-appear on this blog. I expect I'll be done somewhere around the middle of January, 2026, but time will tell.

#Tech

In a recent edition of The Ethicist, a letter to the editor style publication from the New York Times, Kwame Anthony Appiah responds beautifully to a difficult question a reader asked about whether they should cut off an acquaintance who has committed racist acts.

Like you, I favor a bit of grace in a world full of sinners. And cutting off everyone who is morally flawed would leave you with a very small coterie of friends — who might then be tempted by the flaw of moral vanity. (In which case you’d have to get rid of them, too.)

You say you’re an equality-minded liberal. The way to live your creed isn’t by curating a spotless feed of spotless minds but by helping people do better. Hew to the norm; judge the person by what he does next; show grace where it stands a chance to help someone grow. That’s the difference between moral vanity and moral work.

This dovetails nicely with my last post, Counterproductive activism. I would never defend racist acts, obviously, but I agree that moral work demands helping others to be better, if at all possible. The rest, as he says, is moral vanity. Gosh, what a great term.

By the way, helping others to be better means approaching their wrongdoings with kindness, curiosity, compassion, understanding, and forgiveness. It's a slow and painful process, but that's how change happens. This approach is needed even when others cause severe harm. In fact, it's needed especially when others cause extreme harm. Just ask Megan Phelps-Roper, who left the incomparably hateful Westboro Baptist Church only after others had the idea to challenge her with patience and curiosity. Telling someone off in the form of “advice,” when you know the message won't be heard, because it makes you feel better about yourself? That's not moral work. That's moral vanity.

Am I guilty of moral vanity? Yep, in ways I both do and don't notice. Even this post might convey a kind of moral vanity. If you notice times when I'm guilty of it, though, let's talk about it.

#Favorites #Life #Quotes #SocialMedia #Tech

I want to help create a better world. That involves persuading people to see things differently—to care more about animal suffering, for example—but I also need to be sure I'm not pushing people away.

Little does more harm to a cause than the perception that its adherents are crazy. I've never heard anyone say, “Gosh, vegans are nuts… I should be one of them!” Purity tests are similarly destructive. Want to make an enemy out of a potential ally? Chastise them for not being good enough. On the contrary, celebrating small steps in the right direction achieves so much more than demanding perfection. (Guess who else isn't perfect. I'll give you a hint: you can find them in your mirror. Mine, too.)

Protesting outside KFC and throwing red paint on fur coats probably increases animal suffering, on balance, by deepening the resistance and habits of those who oppose ethical veganism. Similarly, having a meltdown when someone disagrees with one's economic vision probably hinders the economic justice they're after.

For that reason, I'm so frustrated and disappointed that social media fosters extremism and encourages users to preach to the choir. It's worse than a massive opportunity cost. It actually leads us to harm those we are trying to help.

#Favorites #Life #SocialMedia #Tech

Some of the best healthcare is free: fresh air, sunshine, exercise, clean water, good sleep, meaningful relationships, and real foods from the Earth. (Well, whole foods like these are not free, but they're often less expensive than the big, bright, fantastical confections that dominate grocery store shelves.)

Of course, none of this is enough, or our ancestors would have lived much longer than they did. Vaccines, medicine, and medical treatments obviously extend life and improve outcomes in profound ways. But to a large extent, what we're lacking—what I'm lacking, at least—is not the big business of medical technology. It's healthier habits and deeper engagement with the real world.

I'm just beginning to really understand this, and I want to be better, so I'm writing this partly as a reminder to myself.

#Life #Maxims #Tech

Progress bars—those little horizontal bars that fill from left to right as your laptop or phone updates—are notoriously unreliable. One moment, a progress bar might be 10% full. The next thing you know, the work is done. If a written estimate is provided (e.g., “10 minutes”), you might notice it change dramatically in an instant.

As it turns out, building accurate progress bars is extremely difficult. In most cases, it's almost impossible for the computer to know how long the work will take without actually doing it.

This is the problem of software project estimation in microcosm.

#Favorites #Maxims #SoftwareDevelopment #Tech

Streaming is great if you like cable, but you want to install a different app for each channel.

#Maxims #Tech