Welcome to mintCast

The Podcast by the Linux Mint Community for All Users of Linux

This is Episode 495

Recorded on Sunday, October 11, 2026.

… Joe; oh my aching back, I’m Bill; … Majid; …. Charles ; Seizing the means of production, I’m Jim

(Insert Intro)

• First up in the news
  Firefox gets a big update; more great news from Reddit; PC sales suffer; and Adobe hops on the “Vibe Coding” train.
• In Check this Out
  Give a hint of the product
• Wanderings
  Give a hint of the wanderings
• In Vibrations from the Ether
  Give a hint of what is coming up.

Show Segments

Please remember if you want to follow along with our discussions, the full show notes for this episode are linked in the show’s description.

• The News (15 min.)

  • Firefox 157 Nova: Significant change to Firefox’s user interface https://blog.mozilla.org/addons/2026/09/29/nova-is-here-what-changes-for-your-firefox-theme/. In my opinion it has ugly oval tabs and an annoying color gradient across the UI. A fast fix for me to get back to more rectangular tabs: about:config then go past the warning then browser.nova.enabled toggle to false. Themes are also available to change the UI as well if that aforementioned fix goes away.
  • Reddit Enshittification: Old Reddit, the far better and more compact interface on desktop, is already rate limited where I get error 429 Too many requests if I click around too quickly, and it is soon going away for many users unless they are using established accounts over the last six months that are logged in. Reddit RSS feeds are also being discontinued with the nonsense excuse that bots are taking advantage of them. So it will no longer be possible to follow subreddits or users in a RSS reader of your choice to aggregate that content for you : https://www.theverge.com/tech/1002788/old-reddit-ai-scraping
  • PC shipments crash: AI was supposed to sell PCs. Instead, PC shipments just crashed 20% as AI eats the world’s RAM, and Windows 11’s Copilot agents aren’t convincing buyers https://www.windowslatest.com/2026/10/11/ai-was-supposed-to-sell-pcs-instead-pc-shipments-just-crashed-20-as-ai-eats-the-worlds-ram-and-windows-11s-copilot-agents-arent-convincing-buyers/
  • Adobe Creative Suite Vibe Coded.: Someone vibe coded a Free Knockoff of Adobe Creative Suite. It’s janky for now, but it sort of works. https://gizmodo.com/someone-vibe-coded-a-free-knockoff-of-adobe-creative-suite-2000823322

• Security and Privacy (5 min.)

• Wanderings (25 min.)

  Joe

  • Last show that i was on i mentioned that i had picked up some hardware at the goodwill. Actually some pretty interesting finds. First up i found a thin client. Specifically a 10ZIG 5900q with a n3710 processor and 4gb of ram for 8 dollars. Not a powerful machine or anything but there is a lot of things that i can do with it. It took a little bit of work to get mint installed and i stuck to using firefox on it cause of the ram limitations but it can easily stream jellyfin or plex. I also looked into it and it can be upgraded to 8gb of ram which is ddr3. I also want to set it up to stream sports and hook it up to the tv in my room so that Jackie can watch hockey in there when she wants to. I am surprised at how well the device works. I know there is not a lot of ram and that the processor is not that good but it is a smooth operating device that has not had any issues with streaming or lagging during light tasks.
  • I also picked up a Roku stick for 4 dollars. Roku is not my favorite OS and i dont like the way that it dials back home but the price was right and i wanted it for wifi dex. But sadly it turns out that the generation of the device is no longer supported and there is no more updates for it. The generation id not new enough for it to have the screen mirroring ability either. So it turns out that it is ewaste now. It cannot do netflix or basically anything else because of the lack of updates. I will have to keep my eyes out for a newer device but still at 4 dollars it is not much of a loss.
  • I also installed readmeabook as a replacement for readarr which stopped getting updates a while ago and was causing problems with my library anyway. But while it is good for tracking new releases it is not good at managing my library of books. It also has a bias towards audible as the only resource for audiobooks. Granted most do come from there but there are other sources, especially if you like older books. It would not bulk import my library and i could go author by author and get it done but i am not seeing a point as it is not showing any of the ones that i have imported. Reading the information about the application it looks like i can attach a LLM to it and have it make reccommendations on new books. Maybe that is what the import is for?
  • Anyways it can be used to look up and get new books and to put them into the correct locations in the file structure with the names for easy parsing and importing to audiobookshelf. Will also sort books by series and can be used to organize in small portions. For some of the authors that i have that are very disorganized in my library i may try to clear out and reacquire so that i get rid of doubles and clean up the structure.
  • Tdarr is still running. It is moving along fairly well but all of my computers that are running nodes spend all of their time with maxed out CPU’s. Some of the tv shows move very quickly and some of them will take all day. I have not finished working on the first HDD yet so this may take more of a time investment than i want to put into it. I am thinking that i could turn it into a much longer term project and just dedicate one node to it and let it run for the next however many months against drives that i am not going to add anything to for now. It will take longer but it will also not max out all of my resources all of the time. Then when it finally finishes put it into a maintenance mode and just have it run against whatever new media i put on it. It does mean putting off the raid project for probably another year unless i happen into some more large capacity hard drives or maybe just nuking and rebuilding the entirety of it.  Then i was able to get through the documentation and the various setups and get the gpu processing working on one device.  Currently i have two that have dedicated gpus and i am working on getting the second one working.  But wow the speeds are literally 10 times as fast.  granted with only one currently going things are moving faster but not by a lot.  If i can get the second one working though it should make things interesting.  It also freed up my processor which is nice and i can do other things on the computer again but i cant use VLC because it is slowed down too much.  That kind of sucks but i can work around that and other things like youtube work just fine. Now i have to admit that i used the google AI from the web search a lot to get everything working the way that it should.  The first search doesnt provide much data but it does ask for details and when you give them it can help and point you in the right direction.  I needed to add a couple of parameters to node configuration on the server and then add the module that looks for the gpu before the one that does things by cpu.  Even if you tell it that it does not have to go in order this is still the case.
  • Well i did the setup last night and sadly the 970m is unable to do h265 transcoding but it does do cpu well enough. this will leave me at 5 devices for now which is the max without paying. This will keep the mini laptop free for day to day tasks which is how i use it.
  • I am also thinking of making some modifications to the second mini rack.  I want to redo a lot of the hard drive setup with all of the 2.5 inch drives and i looking to moving off one of the pc’s that is currently on it and putting on an itx board.  With some slight modification this will also allow me to put on a graphics card and possibly an llm.  I will also be adding an atx powersupply if i do it that way and i have found some interesting 3d prints that others have done that might allow me to mount the powersupply on the back and put some smaller items in from of it like the passthrough jacks and maybe the network switch.  The ITX board i am looking at that i have a processor and ram for would probably need 2u of space so i may end up having to add a couple of u to the rack but that is not difficult.  
    I did 3d print some simple things as well.  Previously i had made some risers with casters for my work bench in the garage and they did work well but i had printed them temporary from pla.  Well i forgot about that and used them for a couple of months and one of them finally broke.  So i saw where the print itself was weak and i bulked up those parts and reprinted all four even though only one was broken.  They are printed in petg this time so i expect them to last much longer.  Also with the temperature being more bearable in the garage i have started cleaning it up so that i can get some other projects finished.  I am still looking for my next run of 3d prints that i will do.  I have a lot if spools waiting.  Some of it will be more mini rack stuff but i am looking for other things too.  
  • I still need to listen to the last show that i missed and i am interested in the topic but things have been very busy.

  Bill

  • Fixed kernel regression with updated AMDGPU driver from AMD website.
  • MuteMe button now works with Pipewire
  • More server upgrades
  • More computers to build

  Majid

  Charles

  Jim

  • Vibe Coding a Firefox extension: Since the popular Xcancel Nitter instance was cancelled again due to legal threats from X, this time likely permanently, I decided to see if I could vibe code a Firefox extension of my own based on an old, archived extension called Nitter Redirect which I used to use that no longer works. I wanted to see if I could do this with a free account without paying a nickel. I decided to start with ChatGPT which I used once before instead of installing or using Claude in the browser. I created an account with an email and used Chat GPT 5.6-Luna in Firefox.

    • Me: I would like you to create a Firefox extension that is modeled on the one at the Github at SimonBrazell/nitter-redirect but that also adds the ability to listen for and redirect requests made to x.com to the Nitter instance of my choice.

    • ChatGPT: I created a Firefox extension implementing that behavior.

      What it does

      • Redirects x.com and www.x.com to your chosen Nitter instance.
      • Also handles the legacy twitter.com domains.
      • Preserves paths and query strings.
      • Handles pbs.twimg.com images through Nitter’s /pic/ endpoint.
      • Handles video.twimg.com media through Nitter’s /gif/ endpoint.
      • Includes an enable/disable toggle.
      • Lets you enter any HTTP(S) Nitter instance.
      • Uses Firefox’s blocking webRequest API, so navigation can be redirected before the X page loads. Firefox documents this API specifically for request redirection.
      • Uses Manifest V3 and appropriate host permissions.

      This follows the architecture of Simon Brazell’s original extension, which itself redirects Twitter requests at the network level and provides a configurable Nitter instance.
      Download the extension source:
      Download nitter-redirect-x-firefox.zip

      Installing it in Firefox

      1. Unzip the downloaded file.
      2. Go to about:debugging.
      3. Select This Firefox.
      4. Click Load Temporary Add-on…
      5. Select the manifest.json inside the unzipped directory.
      6. Click the extension’s toolbar icon and enter your Nitter instance, for example:
         https://your-nitter.example.com

      The original GitHub project has since been archived, so I deliberately used its behavior as the model rather than simply modifying its old Manifest V2 implementation.
      One important caveat: Nitter instances aren’t interchangeable with X in every respect, so some X URLs/API requests may not have a corresponding Nitter endpoint. The extension redirects the browser request mechanically; your selected instance determines what ultimately works.

    • Me: Thank you. But how do I download nitter-redirect-x-firefox.zip? There does not appear to be a valid link there.

    • ChatGPT: You’re right — that link was invalid because I pointed to /tmp/…, which isn’t exposed as a downloadable sandbox file.
      The ZIP does exist, but I need to re-create/copy it into the downloadable /mnt/data area before I can give you a proper download link. I’ll fix that.

    • Me: Nothing has happened, so I still need a link or another way to download that zip file you created for me through my browser.

    • ChatGPT: You’re absolutely right. The file is now in the downloadable area. Download nitter-redirect-x-firefox.zip
      Sorry about the earlier confusion. The previous link pointed to an internal temporary directory that your browser couldn’t access. This one points to the actual downloadable file.

    • Result: This actually worked! I downloaded the zip file, unzipped it, followed the above instructions to add it as a temporary extension in Firefox and it successfully redirects x and twitter links to the Nitter instance of my choice selected in the options.

    • Problem: It didn’t work correctly.

    • Me: Thank you. I installed and tried out the extension in Firefox. It works. However, one issue I discovered- whatever Nitter instance I select with the Save button is not persistent. What I mean by that is that it keeps re-setting so that the extension redirects to nitter.net instead of remembering my choice of nitter instance that I saved with the Save button. I keep having to click save again every time I want it to redirect to my instance of choice. Can you re-make the extension so that the instance I select with the save button is persistent and always used until I change it and it does not keep using nitter.net by default?

    • ChatGPT: Absolutely. I found the problem and rebuilt the extension as version 2.1.0.

      Download the corrected version

      Download nitter-redirect-x-firefox-persistent.zip

      What I changed

      The important change is that the selected Nitter instance is now treated as persistent configuration, rather than relying on the background process’s in-memory copy.
      Firefox’s Manifest V3 background pages can be unloaded and recreated. Firefox specifically recommends storage.local for state that needs to survive those background-page restarts.
      The new version:

      • Saves your selected instance in browser.storage.local.
      • Reloads that setting whenever the extension’s background process starts.
      • Uses the saved value when handling every X/Twitter redirect.
      • Avoids the startup race in the previous version where a request could be handled before the saved setting had finished loading.
      • Never overwrites an existing saved instance with nitter.net.
      • Keeps your existing saved instance and enabled settings when upgrading from the previous version.
      • Continues to support x.com, twitter.com, images, and video.
      • Is version 2.1.0.

      Firefox’s webRequest.onBeforeRequest also permits an asynchronous blocking listener, so the redirect can wait for the persisted setting when necessary.

      Installing the corrected version

      Since you installed the earlier version as a temporary extension:

      1. Download the ZIP above.
      2. Unzip it.
      3. Go to about:debugging.
      4. Select This Firefox.
      5. Remove the old Nitter Redirect for X temporary extension.
      6. Click Load Temporary Add-on…
      7. Select the new manifest.json.
      8. Enter your preferred Nitter instance and click Save.

      After that, you should be able to restart Firefox, close/reopen the extension, or otherwise let Firefox unload/reload its background process without it reverting to nitter.net.
      One caveat: because this is being loaded as a temporary extension through about:debugging, Firefox treats removal/reinstallation differently from a normally installed signed extension. The setting is persistent while that extension installation remains intact. storage.local itself is specifically designed to persist across browser and extension restarts.
      If you’d like, I can also make a next version with a proper Firefox Options page rather than having the configuration appear as the toolbar popup, which would make the extension a little nicer to use.

• Check This Out (5 min.)

• Vibrations from the Ether (10 min.)
  (Topics from the Matrix chat & feedback received)

  • RXVY

    • Hi Joe,

      I’m the developer of Auxo Linux, a new Arch-based rolling distro, and thought it might be worth a mention on mintCast.

      Auxo has a Calamares installer where you pick your desktop (Plasma, GNOME, Xfce, Cinnamon, or ready-made Hyprland, Sway and i3 setups), kernel, shell and an accent colour. After install, one tool, auxo-tweak, switches desktops, recolours the whole system from that one accent setting, or swaps kernels. It sets up NVIDIA and AMD graphics automatically, and takes a btrfs snapshot before every update, bootable from GRUB for easy rollback.

      It’s a small, brand-new project, so an honest take, good or bad, would mean a lot. Happy to answer any questions.

      Website: https://www.google.com/url?q=https://auxolinux.com&source=gmail&ust=1790576578321000&sa=E
      ISO: https://www.google.com/url?q=https://archive.org/download/auxo-linux-2026.09.26-x86_64/auxo-linux-2026.09.26-x86_64.iso&source=gmail&ust=1790576578321000&sa=E
      Discord: https://www.google.com/url?q=https://discord.gg/XbQ66dH5a7&source=gmail&ust=1790576578321000&sa=E

      Thanks for the show.

      Best,
      rxvy
      Auxo Linux

  • Hank Barta

    • Good morning gentlemen,

      I wanted to respond to a recent episode where one of the presenters discussed their experience using AI/LLMs for coding. My experience is different.

      I started exploring AI within the last year and have been using it entirely for coding related tasks. (I’m not interested in a sycophantic AI "friend.) My current situation is a retired S/W developer who has worked for decades in both embedded systems and the financial industry. Since I’m retired and thrifty (cheap!) I’m working with the free models available on the various web pages. At first it was frustrating because the models would propose a solution that almost worked and when presented with the error messages, promptly provide a fix. Why not produce the working solution on the first try? This just wasted the limited messages the free access was limited to. Since then either I’ve gotten a lot better at asking or the models have gotten a lot better at producing working responses.

      My greatest success at this point is using Claude to rewrite an ESP32 baseline project using the new and changed ESP-IDF v6.0.2 API. In the past I’d start a project by copying example files. That was problematic because they’re designed to display as many features and capabilities as possible and reducing that to just what I needed was challenging. This time I started with their canonical starting point and had Claude add the capabilities I needed including WiFi, MQTT, NTP, OTA update and a simple sensor (DS18B20 temperature.) The results were astonishing.

      * Claude understood the file structure that the ESP-IDF framework works best with.

      * Claude understood options to the build/flash/monitor driver ‘idf.py` including options that were scattered throughout Espressif’s  copious documentation and which I was not aware of.

      * Claude produced clean source files that implemented the features I requested. They were not always complete on the first try, but I could request a revision when there was an obvious deficit. For example, The WiFi module implemented error handling and retries for the initial connection and then had no code to handle a dropped connection.

      * When something doesn’t work as expected, Claude can produce a list of tests and suggestions for how to proceed.

      * When asked for a script to push the OTA update to a server that the ESPs could reach (on the IoT VLAN) Claude produced a working script on the first try. It was clean enough that I could easily go in and make changes to better suit my choices. It also included more error handling than I normally have including a check to make sure that the update could be fetched from the server. This is the closest I got to vibe coding with Claude.

      I did vibe code a Telegraf + InfluxDB + Grafana setup for monitoring a half dozen ESP32s using a Docker compose file. I understand Docker but remain fully ignorant of the inner workings of these other packages. This all worked out of the box, providing useful information on what the little beasties were doing.

      I’ve also used Claude and ChatGPT for producing Ansible playbooks and Systemd unit files. In these cases I was pretty specific about what I needed and often provided the model guidance for improving the result. They did not always work on the first try, but they do seem to be getting better. And when I run out of “free messages” I can work directly with the stuff and still make progress. Or do something else until the bucket is refilled.

  • Henrik

    • Hi,

      I listened to the episode a few days ago.

      494 – Good Vibes

      It was informative for me to listen to the live demo of creating
      software with vibe coding and the discussion around it. I also listened
      to the Linux OTC #81, which also reflected on it.

      I do not myself use any LLM for search, thinking, reading etc etc nor
      for any vibe coding. For time being I have no need nor desire.

      But I try to educate myself about AI in several ways. And this episode
      was very helpful to gain some insight in how AI can be used for vibe
      coding.

      Thanks!

  • Kennedy

    • Hi there,

      Episode 494’s vibe-coding segment, where Charlie built the Mint Radio AppImage by prompting, is the most honest demo of where this stuff is for everyday users. Your about page says you welcome guests from the community, so here’s a founder who does that workflow on open models all day.

      I’d like to pitch a guest: Josh Hikida, founder of Vresk, a workspace built entirely on open-weight AI models. His take is one your listeners will have half-formed already:

      AI is not going to kill you. Open models are free to download and already good enough that the biggest labs can’t hold their lead on merit alone. The doom story is what justifies regulation, and regulation is how a handful of companies get to decide what you’re allowed to run on your own machine. Open source has fought this fight before; this time the stakes are the model weights.

      Josh builds on open models from China, the US and Europe every day (Kimi, DeepSeek, Qwen, GLM) and talks from his own build, not benchmarks. For mintCast: vibe coding without the cloud. What Mint users can run today on open-weight models, what still needs a bigger model, and why regulation framed as safety would take the local option away.

      He’s a 2026 UBC Sauder grad who built Vresk in Vancouver and is based in Miami, so Pacific or Eastern time both work. To be upfront, he does have a product. The conversation is the thesis, though, and Vresk comes up if you ask.

      If that sounds like an episode, reply and I’ll send a few dates and a one-pager.

      Kennedy Booth
      Marketing Director, Vresk
      [email protected] · (786) 512-7596
      https://vresk.ai

  • Hello Jim,

    Just finished listening to #492 of the MintCast.

    Thank you for highlighting NomadBSD. I explored it briefly a few years ago. After hearing you talk about it today, I’m going to give it another try.

    I love that you have been experimenting with BSD distros like GhostBSD & Nomad. Linux is great. But the BSD variants have always held a certain attraction for me.

    Looking forward to hearing your updates on how GhostBSD is working for you in upcoming episodes.

    Thank you!

    Alejandro
    New York, NY

  • Hello Alejandro,

    Sorry for the slow reply, but thanks for the email. I wish the NomadBSD team was still actively developing, but even their forum was down again last time I checked. It’s amazing what a single outside person was able to do on their own with that independent fork I highlighted, working even on bare metal. The GhostBSD team in contrast is very busy with the developer Eric himself active on their online forum. Feel free to join us sometime on the mintCast Roundtable on alternate Saturdays at 3 pm ET if you have a webcam. It’s open to all.

    Jim

General

Thank you for listening to this episode of mintCast!

If you see something that you think we should be talking about, tell us!
Chat with us on Matrix at mintcast.org/matrix or send us email at [email protected]

Next Saturday October 17, will be a roundtable episode at 2pm US Central Time (convert to your time zone). Everyone is welcome to join us at mintcast.org/roundtable. Just click the link. No software required. Everyone’s welcome to take part, with a camera or not, or come join us in the Live Chat on YouTube.

The week after, October 25, will be a News/Wanderings episode at 2 pm US Central Time (convert to your time zone). Come join us in the Live Chat on YouTube.

Listen to other Podcasts of the Linux Syndicate Family

Join Bill, Majid, Eric and Charlie on Linux OTC, Linux Off The Cuff, for an unscripted and unpredictable conversation about all things Linux, open source and more.

Join Dale and Majid on Distrohoppers’ Digest to hear about their experiences and challenges with trying out different Linux distros.

Our Hosts

Joe – jeboylan on Matrix, [email protected], & The Linux Link Podcast, The Linux Lug Cast, Buy him coffee on Ko-fi as Joeeb.

Bill – Bill_H on Matrix, [email protected], wchouser3 on fosstodon.org.

Majid – AtypicalDoctor on Matrix, [email protected], & The Atypical Doctor Podcast.

Jim – GNU2Linux on Matrix, [email protected], & The Linux Shortcut on YouTube

Eric – esa1975 on Matrix, [email protected].

Charles – Mr PDX on Matrix, [email protected]

Acknowledgements

Before we leave, we want to make sure to acknowledge some of the people who make mintCast possible:

Bill for our audio editing, hosting & maintaining our Nextcloud & web servers.

Londoner for his contributions, from News to Time Syncs.

The Linux Mint development team for the fine distro we love to talk about. Thanks Clem & Team.

(Insert Outro)