I will concede that while it brought upon us the end of civilized discourse online at a scale unseen and incredible pain and bad vibes, it also brought with it a simple piece of modern art: the meal picture.
I really did enjoy seeing what people were having for lunch because it was either really interesting or really boring and either way it was a kind of art.
So here’s my breakfast taken with the Kodak Charmera.
I needed to get honest with myself about how I really use my phone. Carrying it around is a pain and I’m not entirely sure it’s worth it.
A quick list of things that I do with my phone every day:
Things I do with my phone a couple of times a week:
The things I do daily (minus the time suck that is Balatro) can be done with another device. A device that is simpler, out of the way, and doesn’t invite distraction. It’s the Apple Watch.
The watch will pay for things, allow me to send/receive text messages, make phone calls, and read some notes. I can check email on it but the other weekly stuff kind of falls off of the table. Do I need to do those things while I’m riding my bike or at the library or at the grocery store? Nope. Does the phone beg for attention if I have it? Yeah. So ditch it.
I’ve moved to using the Apple Watch with some Bluetooth earbuds as my total phone replacement. I don’t have as much stuff to carry and it keeps me from mindlessly playing Balatro or scrolling nonsense. The only thing that I miss is the ability to take a picture and send it to someone, but that’s not a weekly use case when I’m out and about. The Kodak Charmera is doing a pretty good job as a camera for me when I’m riding and messing around. And it’s fun.
Less tech. Less engagement with online nonsense. Making the network and gadgets about doing what I need them to do rather than using them for entertainment or wasting time is the goal. It feels like a pretty significant step forward toward a world where the tech works for me and not the other way around.
]]>
There was a time in my memory when the web was a very, very useful place. There were advantages to really understanding search algorithms, keywords, and how to put your hands on the good stuff. Familiarity with solid sites and known-good authors was rewarded with truly useful information. It wasn’t just technical topics (though that was great!). It was how to reset the tire pressure warning on your car. How to replace the heating element in your dryer. How much sun that new plant you liberated from the big box store needed to thrive. You know, things you’d want to know.
But somewhere in the last decade, we lost all of that. Even looking up a simple recipe ends up with the life story of the poster’s maw-maw and as many popups to subscribe to newsletters and autoplay videos as one could handle. Scrolling down the page to get to the recipe ends up with floating ads. There’s more ad code in the page than content. Even Reader view in your browser of choice won’t save you most of the time. I won’t even start on paywalls.
It’s broken. Badly broken.
There’s a lot of “AI Slop” out there too. Most of the technical articles were clearly produced by a LLM. They have obvious errors, missed steps, and the kinds of things that are going to frustrate the kinds of people who really need to find that information. It is, in fact, misinformation. But given the latest GOOG case in Germany, no one believes LLMs anyway so why worry?
All of this crap is going to create a feedback loop that ends up training LLMs on LLM-created bad data. It’s an antipattern. But here we are. This loop only gets worse until it’s all more or less worthless because the effort required to sift through it versus the effort required to simple go out and learn something from scratch will be equivalent. It’s bad news.
But LLMs do adapt. They will have their training optimized. They may even have gasp ethical training sets that contain good information. Hey, anything can happen, right? Which brings me back to what’s bugging me.
The new GOOG paradigm of LLM “search” results first is going to be the only way to navigate it. The damage done to the web is severe. The are still lights in the dark and I’ll get to that in a second, but for searching for something, the only real interface will be a LLM. Yeah. You’ll have to ask one of the robots to pull together your searches and provide citations to get anything done because organic search and discovery won’t work.
It’s already where we are. The top of every search engine’s page is an AI summary. You can turn it off, sure, but what are you gaining when the top three search results are all AI generated bullshit anyway? I’m getting kinda fatalistic with all of this, but I’m not sure that I’m wrong.
The experiments that I’m doing with one of the LLMs is showing me that asking for a summary on a topic with N references and full citations with excerpts is actually playing to the strengths of the model. In its reformulation of the bullshit in the post of choice, it pulls me out of the newsletter popups and general noise and at least gets me to the point where I can see whether I’m closer to what I was looking for or not. When dealing with the corporate web of today, it’s ONE solution. I’m simply afraid that it will become the default because nothing else will produce usable results.
But the open web still exists! Blogs are coming back. RSS feeds are flowing. People are self-hosting and eschewing platforms. Or at least the ones who provide the kind of information and writing that I need are. I can see that growing. We may yet get back to an age of blogging before the platforms took over. I can also see a world where we start to treat social media like smoking in the 90s.
“Oh? You use Facebook? I quit years ago. Never been better. You should try it.”
Maybe I’m dreaming with that, but It’s something that I do hear in the wild.
The web itself is badly broken right now. Human usability is dropping like a stone. I don’t see a really happy ending to this from where I’m sitting today. I could be wrong. I’m happy to be wrong. I just don’t think I am. LLMs aren’t going to take anyone’s job. They’re just going to be the next piece of shit we’re sold to unbreak something that didn’t need to be broken. Seems like a thing that happens a lot these days: break something and then pull together an expensive and unnecessary “fix” for it.
Yeah. That sounds too familiar.
]]>I’m a huge Field Notes guy. I’ve been buying them since their debut because I was following really great blogs back then. Remember that? Blogs? Really great ones that shared really cool stuff? I remember it. It was awesome. Anyway, what carrying a little notebook taught me was that I can’t remember everything. Having kids really drove that home. I live in little lists because I can’t remember stuff. And this extends to data and all things computer.
If I have something administrative to do on a computer, I find a way to do it on the command line and then I script it. Why? Because remembering the name of my script is a lot easier than remembering 100 commands and their critical sequencing. So this blog? I write it with a text editor and then process it with Jekyll. Then I rsync it to a webserver. Do you think that I sit there and type all of this out every time? Do you think that I use a GUI to move files and…all of that? What? No. Not at all. I got it to work a couple of times and then banged out a script to do what I needed doing.
That’s the beauty of computers: if there’s something repetitious that needs doing, there’s a way to make that easier on a bit box.
Back in the day, I used to build dashboards. I wrote scripts that would assemble all of the bad-ass webcomics I used to read and list all of the most recent blog posts from the RSS feeds of the things that I followed. It was great! It got me in the habit of hitting my personal webserver for my daily stuff and things.
After the smartphone entered the world, a lot of that moved to widgets. The phone tried to be my dashboard. For things like my calendar and weather, it did a noble job. But for other stuff? Things that are fiddly or require some kind of parsing? Not so great.
Back to shell scripts. Lots of them. Scripts to grab the weather. Scripts to grab a single Bluesky status post from the local parks department. Scripts to check on RSS activities. All of those things run from the command line and made into neat little lists just for me. But I could only grab that when I was on my computer or could SSH to it.
I don’t know why I did it, but I started to poke at Nextcloud’s widgets. Maybe it was because I hated the default dashboard screen. I don’t care what files have changed. That’s not how I work. I sync a lot of stuff all day long so I don’t need that. Status? What? No one cares. Calendar? Well, now that I’m using Nextcloud as my calendar, maybe that’s not so bad. But what about my badass scripts that do cool stuff? Well…
It turns out that you can make your own widgets and it’s not that tough. If you have a grip on PHP and some javascript, you can knock one out really quickly. It’s not even awful to get them installed!
I built one out that runs my daily dashboard and now I can see it on any device localized as if I were sitting at home. That’s pretty powerful. Especially since it all sits inside of Nextcloud.
Look, the Nextcloud All-In-One docker container is updated by the community on a regular basis. It keeps bugs tamped down and the security up to snuff. If you run it locally and use wireguard to get to it, it’s a pretty waterproof solution. I really dig it. If you start building out your own widgets for it, it becomes more and more useful. It’s more than just cloud storage after that. It’s a place to keep contacts, calendar, and all of your notes, etc. It’s a place to grab RSS feeds and make a curated homepage like in the days of yore. If you want to stop scrolling perpetually, this is your way out.
It’s certainly my way out.
I’m going to build out a tutorial kinda thing for building widgets for Nextcloud. It’s going to take a minute, but when I get it up, I might actually publicize it for the first time. I really want to be a Nextcloud cheerleader. It’s great stuff and the more people who step away from the corporate cloud, the better.
]]>Everything we know about the LLMs that are out there right now grabbing all of the attention (Ha! Get it?! Attention!!!) points to them being trained on, well, The Internet. What’s interesting about that is that the internet leaves itself open in many ways to things like spiders and search engines that crawl sites and pick up patterns and keywords so that it makes it easier to find those things. That’s kind of an incentivized behavior. Most people who put things on the web want them to be found. But do they want their words to be taken and used to feed a language model? Do they want their words used to drive profit for those companies without any compensation to the author? Even in a Creative Commons license setting where there’s Share And Share Alike, there’s not really any callout that would grant permission for someone to aggregate the work for training a language model, right?
Doesn’t matter. OpenAI didn’t ask. They didn’t ask because it was easier not to ask. It was far easier to say that the entire internet was fair game for training because if people could read it, why not a machine?
So then, how did it get access to copyrighted literature? Sure are a lot of settlements out there that might lead one to believe that authors had their works stolen and used as a part of a training corpus. Sounds kind of like it runs against copyright law and a lot of intellectual property legislation that the folks at large corporations have worked hard to incentivize legislators to pass. That’s…awkward.
But set that aside.
They didn’t ask. They took. They extracted value from something that didn’t belong to them. That’s not something we generally accept in our culture. That’s the beginning of the pushback against “AI”. It was created unethically.
I’m not going to say with a straight face that “AI” data centers are The Problem. The truth is something closer to all of the data centers that have popped up in small communities and caused all kinds of havok for local residents by pushing up electricity prices because their demand requires infrastructure updates that, well, all users end up paying for because The Market Says So. Facebook, Amazon, Microsoft, and Google have been putting data centers in cornfields for a long time now. We’re just starting to pay attention. This is a problem with or without “AI” and it needs to be addressed. It’s yet another case of a business that is likely given a tax break to drop a data center into a small community and ends up abusing the neighbors because of its piggish energy consumption.
The worst part? Some county or city official pushed the zoning and tax breaks through on the promise of Jobs! I’ve been in a few data centers. There are usually about 8 people who work there. They make sure switches stay in the ON position and that the doors are locked and the floors swept. The highly paid technical positions? They sit on the other side of the world. There are no system administrators in the Facebook data center in Iowa. There’s a dude who locks and unlocks the door when servers come in or are retired. The person servicing those? That dude flies in from another state.
So to review: “AI” or not, data centers chew up more power than they should and don’t impact the local community in a positive fashion.
Maybe you weren’t in a meeting with muckity-mucks who put you on the spot and said, “How do we leverage the blockchain in our products?” Maybe you didn’t have to take a deep breath, look out your office window for a second, find that centering place inside of you and reply with a very dry, “I’m not sure that our audience and products need the blockchain. Did something occur to you that I’ve missed?”
Are there uses for a massive, overly-complicated audit trail based on an incredible amount of compute? Sure. Can 99.999% of all applications in the world survive and thrive without it? Also yes.
I was in a call where I was asked what our plan was for “Gen AI” which was very hip for the moment. I felt deja vu. Not all applications need a chatbot. Not all applications can make use of a LLM.
That isn’t to say that there aren’t places for it. There absolutely are! The speed at which these things work make the thought of configuring things by point-and-click seem silly. With the right training and application wrapped around it, LLMs can do some wild and crazy stuff that could radically improve performance of workflows and create powerful levers for smart people to do more smart stuff faster. But that’s not what they’re pimping, is it?
No. It’s not.
We’re being told that all knowledge worker jobs will be destroyed by robots and we’d better learn to work with them or we’re going to be left behind. OK. That’s the claim? I’m going to ask “How?” Five times. I bet we run out of answers before I get to three. It’s all claims and nothing to back it up. There are some very specialized people in highly specialized roles seeing ways that these things could do great work. Cool! It’s just a shame that it’s being over-sold as something that will touch every part of your life.
The Hype Cycle is exhausting and no one wants to play that anymore. The fact is, this stuff is so expensive that they have to go into overdrive before LLMs and The Blockchain end up in the same pile because frankly there’s too much money swinging around in a circle for them to “fail”.
Fear creates rage. As a tech worker from the 90s through today, I can say that there has never been a time that we weren’t under the impression that the next shift would cost us our jobs. Software got easier to develop. Microsoft created platforms that “anyone” could use to create software, systems, workflows, or content. That would surely be the end of all software developers! Not quite.
This latest round of “AI WILL EAT YOUR JOB” is just more of the same. The difference is that it’s hitting young people who don’t have anything to lose and are watching their planet burn around them. Their grandparents have long abandoned the path of reason. They’ve never known anything but social media and the abuses of capitalism. Watching the gap between the 1% and these kids grow by the second is crazy-making. But sure, let’s all sit down and learn how to use a chatbot because the next round of HR software will reject your resume if it can tell that a LLM didn’t generate it.
You can’t scare these younger people into submission. If you have nothing to lose, you have nothing to fear.
It really isn’t. ChatGPT? Claude? Nah. That’s not what this is about. It’s about the extractive class sitting behind them taking things for themselves (again) and crushing people on their way up the money mountain. This isn’t a Fuck You to Claude. It’s a Fuck You to billionaires who have enough cash in their couches to start their own personal space programs. It’s about burning down a system that has created inequity that hasn’t been seen in hundreds of years - if ever.
This is some serious shit.
Now? Me? I get out of the way. I’m a GenXer who clawed and scraped in the bit mines for a few decades and made a decent enough living at it. I will get out of the way and let these kids drive the bus. It’s their turn. The stakes are high and very real. If you have more years behind you than in front of you, then I suggest you do the same.
]]>I’m slowly becoming obsessed with the stuff I carry around with me all the time. I went from carrying a bulky laptop to a tablet to a phone and now a watch. Honestly? The move from trying to travel and use an iPad rather than a laptop was frustrating. I can read and watch things on an iPad but it’s far too clunky to use for my taste. Even a phone feels better. That makes no damned sense to me, but this is where I am with my relationship to tech.
Today, if I’m out at the library or wherever, I have my weird tablet situation running Debian and pretending to be a laptop (and succeeeding most of the time, if I’m honest), my iPhone, and my Apple Watch. Sure, I’m carrying around that weird mp3 player and my wacky little keychain camera too, but those things are always on me because they are the tech that I use.
The phone is the one I struggle with. I’ve got it on me all the time. I toss it into a weird pouch thing when I go mountain biking so that my watch can send an SOS if I fall and break something. I use my phone to pay for stuff. It’s always there. It’s this little brick I can’t do away with. But I’ve been leaving it at home or in the car more and more because I just don’t wanna anymore.
So is it the Apple Watch? My little Mortality Minder? Is that the one device I need to carry? I have it on all the time to track my steps and my vitals because I’m at an age where that data is actually useful to me. My workout data kind of matters now too. It also tells me the time. Plays music from The Onlines. Gets texts from my family. With a cellular connection, I could do calls on it. I can’t tether my “laptop” to it, but so what?
I’m trying to wrangle my computing and internet use such that it happens at my desk on my desktop machine(s) so that I’m not out there being a screenager all damned day. I need a lot less time online and when I’m at a computer I need to be getting shit done and not screwing around with all of the noise. Right now? I’m writing this at my desk in a terminal window because vim. That’s the kind of focused computing experience I need.
The Apple Watch fits this mode because I can pare down what it does for me to a very specific set of subfeatures. At some point, I will need to find a way to tear myself away from the Apple Ecosystem (probably?) because I don’t trust anything anymore. But for now, this could be my ticket to having one piece of tech on me 95% of the time.
The less I engage with the gadget, the better. Use it like the tool it is and then get it out of my sight.
I’m not getting away from my phone. I get that. This is just a step in the right direction. Probably.
]]>Sitting at my desktop computer and hacking away on a few mobile apps that will help me as I move toward a transition to a minimal tech lifestyle is teaching me a lot about what it might be like to start coding today. I think a lot of what I experienced early in my career is gone. Evaporated. No longer particularly useful. But what remains is definitely going to be important.
The truth is that power tools didn’t mean we needed fewer carpenters. It meant that the way the carpenter worked changed in a meaningful way. A nail gun means that a roof can go on in hours. When we put a new roof on our house it was a job that ran from after lunch until about 4 PM. It was crazy to see it all just…go. Nail guns, magnets on wheels, and a few specialized power saws made quick work of it. It’s going to be like that with application development from now on.
Talking to an LLM isn’t all that different from a civil sitdown with a senior developer. There is a list of requirements. The level of detail that goes into those early conversations often determines what you get in the end. Back of a napkin requirement gets a back of a napkin implementation. Talking in big blocks gets big blocks. Going deep and getting really specific? Yeah, that’s still the way to get the level of quality you’re looking for. The machine isn’t any different.
To see that I can go in and tell the machine that I want an app in this framework that does these three things and seeing it come back with a solution in under 10 minutes is still impressive to me. It won’t be for long. Expectations will change. But for now, I’m enjoying not typing up all of the nonsense that goes into a boilerplate level application. It’s like the Visual Studio Wizard finally delivered on its promise by being flexible enough and specific enough to be useful enough. All of the initial tasks are going to shift in a different direction.
The folks who sit down and pound out data models and designs are still going to do that work. But their work is going to go to an LLM and straight into the application. They won’t be putting together a design simply for a human developer and tester. They will be doing it for the machine and the human who is still very much in the loop. Yeah, I don’t see a LLM self-healing an application at 02:00 on a Saturday just yet. Someday? Probably. Sooner than I think? Yeah. But not today.
Today we still need to treat the design as though the machine and its handler are going to need to use it together. That will change how things are phrased, perhaps, and what tools are used to make it work. If the designer knows that a human developer and the machine both need to be able to read it (oh, and probably auditors and anyone who might drop in to buy an application or check its quality) then everything they do changes just a little. Fewer ambiguities. A touch more of How and not just What. That might be the biggest change: the designer is the first pass developer.
Management will see this and act accordingly. Responsibilities and job descriptions will change. Staffing levels will shift. The kinds of people who are hired will be different. The organization will adapt to the power of these tools. Dependence will lead to some unintended consequences that I’m still puzzling over. Most importantly? Labor will be valued differently. This is where the fear lives. This is why we are pushing back on this tech and its adoption.
Nothing is certain. There is no inevitable shift here. But there is benefit that cannot be ignored. Accuracy where it counts. Risk assessment. Finding patterns humans don’t see or can’t look for because we aren’t fast enough. Better software can, in some cases, save lives. No, I’m not talking about yet another social algorithm. I’m thinking about medication dosing. Looking for patterns during treatment. And in the case of LLMs, being able to communicate that in a conversational tone.
I’m not done thinking about this yet. All of the good I see is tempered by the very real challenges and abuses that we know will come as it has with all tech. Our legislatures and leadership will lag by 10 years. We still haven’t figured out what to do with social networks that doesn’t involve burning down privacy rights for users. I don’t expect it will be different with LLMs. That’s why I’m playing with them. By the time the law catches up, it will be too late and the expectations will be baked into the behavior of our organizations, economies, and personal lives.
]]>
I could get used to this simplicity. Quick shots. Interesting output. A bit of distortion and lack of all of the things that we thought we wanted in digital cameras. Oh yeah, and no geo-tagging. I love that part! This is tech I can use.
]]>I watched a video of a robot folding sheets. Not sheets of steel or uranium or something toxic. Sheets that go on a bed. Folding them. So that they can be put in a closet. Laundry.
This robot? $12k. Twelve Thousand Dollars to fold sheets. What in the holy hell is that?
I’m not stupid and I’m not being disingenuous here about what that means. The dexterity to fold a sheet implies the dexterity to perform complex tasks. Given that this is a machine, those tasks could be completed in environments that are not safe for humans or compatible with human life. That’s huge!
But please, please, please get the fuck out of here with the folding of sheets and the making of coffee and the Jettson’s level of tasks for rich people. That’s fucked up. We don’t need Rosie. We need R2-D2. We need robots that are Swiss Army Knives - even if they speak in a language that isn’t entirely human friendly.
For decades (since we’re in the Delay phase of Deny, Distract, Delay with Climate Change) politicians and technologists have talked about how we as a species can engineer our way out of a climate catastrophe. Right now the only things that I see in place for that are the micropayment structure for oxygen and water and surveillance to see who might be drinking or breathing illegally. If we’re even going to try to save what’s left of our burning ecosystem, we might want to spend less time with robots that fold sheets and more time on robots that maximize solar collection and water purification. How about a robot that collects microplastics from water supplies?
Or R2-D2.
]]>