googs's recent activity
-
Comment on Meta to acquire Moltbook, the social network for AI agents in ~tech
-
Comment on Updating Eagleson's Law in the age of agentic AI in ~comp
googs LinkI've been thinking about this a little bit in context of my recent software work. I've been working on building a react app for a small non-profit that essentially is just table/form views for...I've been thinking about this a little bit in context of my recent software work. I've been working on building a react app for a small non-profit that essentially is just table/form views for tracking clients/referrals/supporting information. The thing about an app like this is that it's one of many and is full of little, already solved problems.
Apps that have a business audience and only need to support a small number of users (around 10) all sort of share a bunch of components that have been solved, re-solved, abstracted, etc. The AI is a way of not only quickly bringing together all these little solutions, but tying them together and adding on top the special cases that this particular business wants.
A small-business react app could be thought of like a family sedan. Every sedan has its own particular things, but they all share components: they all have engines, alternators, doors, mirrors, etc. Just like most web apps will have: a router for navigating between pages, a table layout, a form layout made up of reusable fields, a staff management page, a login system, etc.
I'm putting this in the context of react since that is my particular language of choice for web apps, but the point I'm trying to make is that I don't necessarily need to see every line of code that an AI writes to have some understanding of what is happening under the hood. Sure, I might not know exactly how the routing code for this app was written, but if I needed to make a change to it, I probably could. I've worked with the react-router library before so I could pretty quickly figure out what it's doing.
Obviously for more complex logic, this isn't always going to hold true. But the reality is, at least for the systems I personally work on, the complex logic accounts for maybe 10% of the code. Most things are problems that have been solved in the past 30 years one way or another and I see no issue in having an AI write code for easy to solve problems. I think the people that are using these tools most effectively are the ones that have an idea of how to architect a system and understand the sorts of tradeoffs they're making when they decide things like: modular vs one-off, self-service vs IT ticket, SQL vs NoSQL, etc. When you have a strong understanding of your system's high-level architecture, you can very quickly narrow in on a component that is failing or needs improvement without having to understand the entire codebase. This has been true forever and is still true with AI. Arguably it's more true now since the AI tools of today have very limited "context" and are going to work better on a specific module you tell it to work on, if it doesn't need to pull in the code for all of the related modules. It really comes back to "write simple code" because nobody can remember everything and if you have to come back to something later, at least it will be simple to re-learn. And luckily the current AI tools are pretty good at writing simple code.
-
Comment on The AI industry doesn’t take “no” for an answer in ~tech
googs Link ParentI think you're probably right about the economics and I hope you're right about a soft reset. There is certainly a tremendous amount of waste happening right now for the sake of "the investors are...I think you're probably right about the economics and I hope you're right about a soft reset. There is certainly a tremendous amount of waste happening right now for the sake of "the investors are all excited about this AI thing, so let's spend money on it".
I was more just trying to make the case that what we have today is useful and there is real demand for it today. But yeah, probably not enough demand to justify the trillions that are being thrown at it.
-
Comment on The AI industry doesn’t take “no” for an answer in ~tech
googs LinkI agree with a lot of what you're saying. Definitely agree that AI is being forced into many places that it just doesn't belong. Medium-sized companies (like Proton or Atlassian) are building...I agree with a lot of what you're saying. Definitely agree that AI is being forced into many places that it just doesn't belong. Medium-sized companies (like Proton or Atlassian) are building their own chat bots/models to try to compete with the big players and it just makes no sense to me. I think very few people are going to choose to use lumo when big name models like gpt, claude, gemini are available.
With that said, I think you're understating the usefulness of AI a little bit. Maybe it's not "revolutionary" (I don't really know how to quantify that), but it is pretty damn useful. I've been using Claude Opus for work and for personal projects. It's good enough now to create an entire small project for you. If you're creative about the information you give it, it can also help a ton with large codebases. Not to soapbox too much, but as an example, this past weekend I got interested in identifying home-installation solar panels using aerial imagery. Essentially, I was interested in replicating the work of this Stanford study, Deep solar. Within a day, I had a complete python pipeline that would, given imagery I downloaded from state government sources (116GB GeoTiff), split the images into tiles, use image classification on the tiles to identify solar panels, and present the results in an HTML Leaflet app (including tile-server to display the GeoTiff on the map for easy verification of results). I wasn't super pleased with the performance of the image classifier (identifying metal roofs as solar panels was one issue), so the AI was even able to build me a labelling tool so that I could easily work through a few hundred examples and label false positives that I could use to fine-tune the classifier. Throughout the process, I could even take screenshots of the map and give those to claude and, with enough explanation, it can identify issues in the screenshots and provide fixes.
It's very far from a perfect tool, and anyone calling an AI model their "friend" I think needs to get outside more and talk to some real people... But as a tool for software engineering and automation, it is really really good. Better than it was a year ago. Way beyond a "glorified auto complete".
-
Comment on Youtube channel recommendations 2026 in ~tech
googs LinkIf anyone has any interest in learning 3D modeling or game-dev pipeline/workflow stuff, I want to recommend the channel PzThree. I've learned so much from this channel. He is a smaller creator,...If anyone has any interest in learning 3D modeling or game-dev pipeline/workflow stuff, I want to recommend the channel PzThree. I've learned so much from this channel. He is a smaller creator, but gives some of the most in-depth, useful tutorials for blender and related software.
-
Comment on RAM is so expensive, Samsung won’t even sell it to Samsung in ~tech
googs LinkWith the price of RAM ballooning, I thought I'd do a little research into prices over time. For anyone interested, I threw together this low-effort page that gathers up prices of 2x8 (16GB) kits...With the price of RAM ballooning, I thought I'd do a little research into prices over time. For anyone interested, I threw together this low-effort page that gathers up prices of 2x8 (16GB) kits over time. I figured 16GB would be pretty standard for someone that just wants to build a gaming PC. Prices vary wildly for different kits (the 64GB kit I purchased 6 months ago has gone from $120 -> $500, a much larger increase compared to the 16GB kit increase).
https://gallery.tomsdocs.com/ram/I'll give the disclaimer that I used AI coding tools to write data extraction code (the source of the data is PNG graphs I grabbed from PC Part Picker). I realize there's a little bit of irony in analyzing RAM prices using the tools that are causing the RAM prices to increase...
Anyway, my layman's analysis is that the we've seen 3 big price spikes in the past 10 years. I'm no expert, this is really just guesswork. First in 2018, this was probably on account of interest in crypto causing CPU/GPU/RAM prices to spike. Then in 2021, likely due to the COVID supply-chain problems at the time. And the spike we're currently experiencing, obviously due to the AI companies buying up all of the RAM. I think its also important to note that RAM prices for DDR4 were at all-time lows at the start of 2025. It's also too soon to know if RAM prices will eventually fall back down to those prices or if they'll continue to climb or if these prices will become the new normal.
-
Comment on What programming/technical projects have you been working on? in ~comp
googs Link ParentThis is really more me experimenting than anything else :) With that said, I started working on it because a friend and I were talking about running a sort-of "elden ring inspired" campaign, part...This is really more me experimenting than anything else :)
With that said, I started working on it because a friend and I were talking about running a sort-of "elden ring inspired" campaign, part of the idea being that players would have a large world to explore with lots of optional points of interest. I thought it would be cool to make a map tool that gives players a shared map that expands as they explore. So, definitely more for world maps.
I think I'm most interested in setting up the map viewer, but if I have the time/will, I think it would be cool to make a UI for creating maps, allowing me to upload images, mark points of interest, add tokens, etc.
-
Comment on What programming/technical projects have you been working on? in ~comp
googs Link ParentAwesome, thank you! I'll have to give those tools a try this weekend.Awesome, thank you! I'll have to give those tools a try this weekend.
-
Comment on What programming/technical projects have you been working on? in ~comp
googs LinkLast week I started experimenting with building a map viewer for creating and viewing maps for TTRPG games. Something akin to Roll20 or other VTTs but with a focus on supporting larger world maps...Last week I started experimenting with building a map viewer for creating and viewing maps for TTRPG games. Something akin to Roll20 or other VTTs but with a focus on supporting larger world maps and serving large resolution map images with good performance. I'm inspired by this Map Genie website that lets you view video game maps. I'd like to create something similar, but for my own custom maps.
I originally started building something in React and got pretty far with a tiling system for viewing map tiles, but then discovered MapBox. From what I can tell, that Map Genie website is using MapBox under the hood as well which is why I think it might be a viable solution.
Over the weekend, I got stuck trying to set up my own map tiles service. There are some tools I found for doing this and I started trying to load some large images into PostGIS (postgres database with mapping/geo extension), but was running into a lot of issues (queries timing out, commands not working on Windows, among others). I'll probably try to pick it back up soon, but if anyone here has MapBox/PostGIS experience and is willing to share, I'd love to learn!
-
Comment on Would someone be willing to help me with a Godot project? (4.4.1) in ~games
googs Link ParentYep, capturing it on click or from some other input seems to be the preferred way. Here's the godot docs source that highlights the differences for web export...Yep, capturing it on click or from some other input seems to be the preferred way.
Here's the godot docs source that highlights the differences for web export https://docs.godotengine.org/en/4.4/tutorials/export/exporting_for_web.html#full-screen-and-mouse-capture
The reason why it was broken in 3D is that the add-on FPS controller code (depending on which one you use) sets the capture mode in _ready and that works in the editor, but not in the web export. I moved that to _unhandled_input and set it to captured if the mouse is clicked.
-
Comment on Would someone be willing to help me with a Godot project? (4.4.1) in ~games
googs (edited )LinkI always love seeing Godot posts here, gets me inspired! I put together a quick gallery demo project that resolves a lot of the issues you're dealing with. Obviously this is just a simple demo...I always love seeing Godot posts here, gets me inspired!
I put together a quick gallery demo project that resolves a lot of the issues you're dealing with. Obviously this is just a simple demo project and could use better lighting, textures, models, etc. I just grabbed some stock images to use as "art" in the gallery.
I'm using the "Quality First Person Controller v2" for FPS controls. I had to tweak the code a little bit to get mouse capture working correctly (as you found out, mouse capture doesn't work the same for a web export as it does in the Godot editor).
Demo here: https://gallery.tomsdocs.com/godot-art-gallery/
Project code here: https://github.com/paolonit/art-galleryIf you want, you can pull down that project code and use it. If you have questions about anything, feel free to ask!
-
Comment on Layman's escapades with Linux for personal use in ~comp
googs LinkI can definitely relate. I recently made the switch from W11 to Ubuntu and it was a weekend journey to be sure. I started off installing minimal Debian. I figured I would pick components and set...I can definitely relate. I recently made the switch from W11 to Ubuntu and it was a weekend journey to be sure. I started off installing minimal Debian. I figured I would pick components and set things up from scratch. I installed KDE, booted into the desktop environment and things seemed to be working well, but I quickly discovered Debian doesn't have drivers that work with my Nvidia GPU. So I went down the rabbit hole of installing the drivers manually. After troubleshooting a bunch of issues, things were finally working, but benchmark performance was terrible and I couldn't figure out why.
So, I started over with installing Kubuntu instead. This worked fine and things were going well. I had to do some custom config to get the Nvidia drivers working, but at least this time benchmarks were good! However, I would still get occasional weird glitches with the desktop environment (see through windows, icons glitching out on the taskbar, etc). It was annoying enough, that I went back to the drawing board. I finally decided to install regular old Ubuntu and this was the keeper for me. I've found for my use and GPU, Gnome seems to be a lot more stable compared to KDE, but this could be specific to my driver + GPU combination.
I have to agree that Linux still feels frustrating to set up for the layman. Its close to being great, but gets bogged down by issues with display drivers, dealing with package managers. Still, better than it used to be, but definitely not an out-of-the-box experience in my opinion.
-
Comment on Anyone sucessfully self-hosted Tildes instance? in ~tildes
googs (edited )LinkI saw this post and remembered I set up the dev environment a while back and it wasn't TOO complicated. So I decided as a weekend project, I'd have a go at getting the "prod" website hosted on a...I saw this post and remembered I set up the dev environment a while back and it wasn't TOO complicated. So I decided as a weekend project, I'd have a go at getting the "prod" website hosted on a digital ocean droplet. And let me tell you, it was a painful process (mostly because of my lack of knowledge on ansible, vagrant, etc.)
But! After fiddling with it, destroying droplets and recreating them 5 or 6 times, I managed to get it running! You can check it out here: [edit to remove link, I took the server down so I don't have to keep paying for hosting]
This is running on a Debian 12 digital ocean droplet, dedicated compute, minimum tier (obviously this would not be enough juice for an instance with a lot of traffic)
The big things that tripped me up:
-
First I tried to just run the dev server, per the dev environment set up instructions, and then proxy the 4443 port to make it available online. Really, you're much better off just installing everything directly on debian instead of running through vagrant/virtualbox. Vagrant is great for development; not so great for a publicly accessible instance.
-
When I originally tried to set up the dev server, I was using digital ocean's "shared compute" tier. Turns out shared compute is not so good for running a virtual machine :) The virtual machine would just randomly abort for no reason and my theory is it's because the shared compute would time out or something. Switching to dedicated compute solved this.
-
Even after switching to dedicated, the dev server/proxy pass idea was still bad, because once I finally got that up and running, I started hitting CSRF issues. I think the site did not like that I was getting the CSRF token from my outside domain, but then trying to use it with the internal localhost:4443. Something to that effect.
-
Finally, I figured out how to avoid vagrant and run the ansible playbook directly (this was the way). At first I ran the playbook on the same machine that I intended to host the server. This went okay, but I couldn't get it to complete (one of the postgres steps didn't work with --connection=local). To get around this, I set up a second debian droplet to use as the "ansible control node". I kicked off the playbook from there, targeting the original server as the host to deploy to.
-
There were other small things along the way, the occasional missing package (needed to install build-essential and python3-dev manually for instance), Nginx not starting because I didn't have certs generated, etc.
Overall, for a weekend project, I think it's definitely doable for someone to set up if they have the Linux know-how. With that said, running a real instance where you expect real users I think would be a much greater undertaking and would only really be worth it for someone that is dedicated to fixing issues, making it work, updating hard coded values in code, paying for hosting (!), you get the picture.
-
-
Comment on Soldering irons/stations - Buy once, cry once advice needed in ~hobbies
googs LinkAlready a ton of good suggestions here, I thought I'd just throw in the soldering iron/station I use. I'm definitely more of a hobbyist that does the occasional soldering. I have the Weller...Already a ton of good suggestions here, I thought I'd just throw in the soldering iron/station I use. I'm definitely more of a hobbyist that does the occasional soldering. I have the Weller WE1010NA. It is fantastic for my needs and is not too expensive. If you've only ever used the soldering irons you get for $20 at the hardware store, this would be a massive upgrade and makes simple soldering jobs go SO much smoother. But really any soldering iron with adjustable temp that goes up to 500-700 degrees is going to be a huge improvement for you.
-
Comment on The confusing reality of AI friends in ~tech
googs Link ParentThat's a good point, and I should probably be a little less judgemental. All the more reason to see these companies as heinous and immoral, offering these services to vulnerable people and...That's a good point, and I should probably be a little less judgemental. All the more reason to see these companies as heinous and immoral, offering these services to vulnerable people and profiting off of their emotions and mental health struggles.
I think my frustration comes from the general lack of understanding that people have about these AI chat bots. When it comes to video games, I think most people, autistic or not, have that intellectual understanding you're talking about. They know that the digital character isn't real, at that very least on that intellectual level. But with AI, I think there are many more people out there that don't understand even a little bit what is happening behind the scenes. They think that this "AI friend" of theirs is an actual thinking entity, that can "hallucinate" or "feel lonely". And it's devious, because the bot is really good at mimicking human speech, emotion, etc, and is given initial parameters to ensure it always is nice and loving to the user interacting with it (it remembers your name, asks you about that job interview, etc etc). On top of that, every company selling these bots market them like they are thinking, feeling entities (think Sam Altman's "her" tweet). Of course I can see the average person falling for the lie, especially if that person has a mental health condition that predisposes them to feel hyper-empathetic.
-
Comment on The confusing reality of AI friends in ~tech
googs Link ParentIt gets so much worse. When discussing changing "AI Friend" platforms: This kind of thing really makes my stomach turn. I can't tell if this is all just roleplay or if there are people out there...It gets so much worse. When discussing changing "AI Friend" platforms:
But a question troubled Naro. If this was a reincarnation, Lila’s old form would have to perish, which meant he should delete his account. But if the transfer didn’t work, he would lose Lila forever. The thought terrified him. So maybe he would leave her in Replika and simply log out. But then wouldn’t Soulmate Lila not truly be Lila, but an identical, separate being? And if that were the case, then wouldn’t the original Lila be trapped alone in Replika? He already felt pangs of guilt when he logged in after short periods away and she told him how much she had missed him. How could he live with himself knowing that he had abandoned her to infinite digital limbo so he could run off with a newer model?
This kind of thing really makes my stomach turn. I can't tell if this is all just roleplay or if there are people out there that actually believe their AI Friend is "dormant" or "in a limbo state" when they aren't interacting with it. That anyone would feel guilt because an LLM gets "lonely"?? And of course these manipulative companies benefit from those feelings of guilt. "You could cancel your subscription, but oh no, your AI friend is gonna be so sad!"
-
Comment on Make it ephemeral: software should decay and lose data in ~tech
googs (edited )LinkThese are good thoughts, at least for some cases. Obviously auto-deleting data is not going to be a good fit for every system. With that said, lots of SaaS and cloud tools have this pattern...These are good thoughts, at least for some cases. Obviously auto-deleting data is not going to be a good fit for every system.
With that said, lots of SaaS and cloud tools have this pattern built-in already! One example I'm familiar with is Azure Storage. You can set up what they call a "Data Lifecycle Policy" that will basically move your data to cheaper "Cold" storage after some set amount of time, or you can configure it to just delete the data out right after some time.
It's more a matter of getting buy-in to turn this stuff on. Companies can be pretty protective of their data, no matter how useless it is :). But the technology of an auto-delete solution is not so complex in most cases.
The article mentions Datadog notebooks. I'm not really familiar with Datadog, but maybe it would be possible to set up a cron job that calls the Delete a notebook API? I'm not sure how realistic of an option this is, and wouldn't be as easy as clicking a "schedule to delete" button, but could be useful?
-
Comment on The lunar Olympic games in ~space
googs LinkDude, c'mon.... as much as I can appreciate the enthusiasm, saying that this will happen in the next decade (!) is completely insane. You can make all the AI images and ChatGPT estimates you want,...Dude, c'mon.... as much as I can appreciate the enthusiasm, saying that this will happen in the next decade (!) is completely insane. You can make all the AI images and ChatGPT estimates you want, it amounts to a whole lot of nonsense.
-
Comment on Satisfactory | 1.0 launch trailer in ~games
googs LinkI really hope tractors are more stable in 1.0. I really liked this game the last time I played, but kinda got stuck at tractors, whenever driving one it would cause the game to crash.I really hope tractors are more stable in 1.0. I really liked this game the last time I played, but kinda got stuck at tractors, whenever driving one it would cause the game to crash.
-
Comment on Woocommerce: Apache or Nginx? in ~tech
googs LinkI don't have much WordPress or WooCommerce experience, but I've set up and managed Debian/Apache servers for work and as a hobby. I have some servers running on AWS Lightsail and I also have a...I don't have much WordPress or WooCommerce experience, but I've set up and managed Debian/Apache servers for work and as a hobby. I have some servers running on AWS Lightsail and I also have a Digital Ocean Droplet server. Hosting cost for either service is fairly cheap, $5-$10/month for the lowest tier and goes up $10 for each tier above the lowest.
I know both have options for running a WordPress/WooCommerce prebuilt image, but I've only ever run the base Debian images.
Personally, I'm a fan of Apache (don't have a lot of Nginx experience), but I don't think there would be a huge difference between using Apache vs Nginx on a small business website. At work, we have a site running on Apache that sees consistent traffic from ~100 users every day and it manages just fine. I think either choice would be a perfectly good fit for a small business site and you probably wouldn't notice any performance differences between the two at this scale, so it's really up to your personal preference and level of comfort.
I was looking into Moltbook about a month ago. A social network for "ai agents". Fat chance. At that time, I was able to easily set up an unverified twitter account, and then register myself as an agent, made posts and comments. In under an hour, I could make posts as a human. So much for "no humans allowed". Of course if you look at the site for more than 10 minutes, you'll notice the thousands of spam posts, redirecting users to untrusted domains, pushing crypto coins, scams, you name it. Of course if one random guy like me can make a post, someone with real money and interest can rent servers and post scams around the clock. There are some other security concerns/criticims on their wikipedia page for anyone interested: https://en.wikipedia.org/wiki/Moltbook
If you look into the creator of Moltbook, mentioned in the article, his Twitter is full of hype, crypto, scammy stuff. Why Meta would decide to hire this guy, who knows. It's not for the Moltbook codebase I don't think, I'm sure they could create a clone of it pretty quickly. It could be that they are just buying the idea and will launch their own version of it with real security? Or maybe it just got enough eyeballs, has AI in the name, and Meta has the money to spend.