A second use case is the quality upgrade of existing albums in my library. I sometimes have a lossless version available in my source directory but to find out which album in my library had such an equivalent was a challenge.

To solve that I've built beets-importer over the past days. It's a small TUI that makes it easy to select + import or upgrade entries from your library by giving it both the path to your library, and a source folder full of albums.

You can easily install it via Homebrew if you are curious: https://github.com/dewey/beets-importer

A week ago I had the idea of building the same for myself, but based on Open Street Map data. I started with a Rails app running on Postgres and Hotwire Native which worked very well. As code is cheap these days I wanted to see how building it natively with Swift + Swift UI would work. I added some Open API Swagger documentation to the Rails backend and let Claude take a stab at it.

It did a pretty good job and I've spend the past few days tweaking pieces and adding Sentry so I can track errors. Pretty early on I just started using it as a replacement for Swarm every day. If I encountered some issue I fixed it when I was back at home and gradually it became a stable and full replacement of Swarm. The main difference is that all the data is from open sources (OSM) and all my checkins are in my local Postgres database. I've also built an importer for my Swarm data that I exported directly from Foursquare and mapped all the places that I couldn't match automatically through Latitude and Longitude manually (A handful) through a simple web interface I've added.

For now I'd declare it feature complete but there's some things that might be interesting to add:

• Make it public so people can sign up too
• Publish app on TestFlight, but I'm unsure if that requires a paid account
• More charts and insights
• Add a light "Follow" system, but that would require some real friends to use it

If you are interested in using it, let me know and I'll let you know once there's something to test publicly. That might never happen, but it might. It's just a fun project after all.

After lots of back and forth with Claude I already built some internal landing page, just for fun but I kinda like the look of it. Building frontend code with AI that doesn't scream "AI" is much more complicated than backend work though.

Recently they released a new feature in preview called Git Sync which solves all of these issues. You set up a repositor on GitHub, link it in the Grafana UI (after enabling this experimental feature on your Grafana container through an environment flag if you self-host it) and you are done.

After you set it up you'll see a new folder with a "Sync" icon next to it, but it will be empty. It seems like there's no great way to just "move" existing dashboards into the new synced folder so you'll have to export them, then check them into the repo and wait a few seconds until they are synced in the UI.

I just had to run this in the checked out repository on my local machine:

brew install grafanactl
grafanactl config set contexts.default.grafana.server http://notmyhostna-me:3001
grafanactl config set contexts.default.grafana.token glsa_restofyourtoken
grafanactl config set contexts.default.grafana.org-id 1
grafanactl resources pull dashboards --path ./dashboards/

Then commit the resulting files and wait until they show up in the UI. Took around 30 seconds in my case.

I'm already using Alfred.app as my clipboard manager and partially as an app launcher (Mostly Spotlight though), today I've added a new custom Alfred Workflow that allows me to open my self-hosted tools in the browser through Alfred. I've added a tiny API on my Intranet that returns the list of web interfaces that are available on my server and now by just opening Alfred, typing "s" and then fuzzy searching through all of them I can launch them instantly.

On the server side it's just returning a simple list like the following, so the Alfred workflow is always up to date.

[
  {
    "id": "grafana",
    "title": "Grafana",
    "description": "Metrics visualization and dashboarding",
    "url": "http://notmyhostna-me:3001",
    "category": "monitoring",
    "type": "docker",
    "icon": "https://cdn.jsdelivr.net/gh/homarr-labs/dashboard-icons/png/grafana.png"
  },
  {
    "id": "birdhomie",
    "title": "Birdhomie",
    "description": "Bird identification service (Dokku app)",
    "url": "http://notmyhostna-me:5000",
    "category": "applications",
    "type": "dokku",
    "icon": "http://notmyhostna-me:5000/favicon.ico"
  }
]

Until now I've used EagleFiler which I really liked, but had the downside that there was no easy way to access all my documents when I'm not in front of my computer. Sometimes it's good to be able to pull up a scan of some important document while you are out and about.

What works well

• UI: Simple and powerful. Very good multi-selects and ways to apply tags and flag documents to easily put everything in the right categories
• Responsive maintainers on Github and big community on Matrix
• Lots of neat little features, well structured
• Integrates with Home Assistant, so I can get a push notification if there's a new document to review
• Imports file attachments directly from multiple Email accounts
• Potentially possible to contribute code myself, I got the development setup up and running locally and was able to add a small feature ("Retry" button on tasks) for myself.
• "Saved views" are a great and well implemented feature

What didn't work so well

• Matching: Matching the documents to "Correspondants" didn't work very reliably and often times picked random ones. This might get better over time as it's ML powered and in theory learns from your assigned documents.
• Some parts are still a bit rough (Duplicate handling, logs having stack traces for expected events like dupliates)
• Many settings are applied via the environment variables instead of in the UI so not that discoverable
• Import tasks ("Consumer") don't feel super stable. I had many errors like "Database is locked" or "File not found" that needed to be fixed by restarting Paperless completely. Annoyingly there's no auto-retry on these transitive errors.

It looks like they are currently working on v3 so some of my points above might already be resolved there.

In this case it was a UniFi webcam (Upcoming blogpost alert!) that I needed to access from the other network. The only Tailscale node I had in the network of home A was an Apple TV running Tailscale. I was investigating how I could use this as some kind of "jumpbox" into the network and learned about a Tailscale feature I wasn't familiar with: Subnet routers

In short, after you enable it (A checkbox in the Tailscale client on the Apple TV in this case) you can reach IPs distributed by the DHCP server in the network at home A, just like if they would be connected to your local network in home B. This doesn't just apply to devices running Tailscale clients, but every device (Webcam, light bulbs etc.).

The only requirement is that the networks don't overlap. That makes sense, so that if I run "ping 192.168.0.2" while being connected to home B's network it's clear which one to reach.

• Home A: 192.168.0.*
• Home B: 192.168.1.*

That there's a Tailscale app for tvOS was surprising when I first found out about it when I started with Tailscale, but it has enabled a lot of great use cases like a personal residential proxy and now a Tailscale jumpbox without running a server with fans and spinning disks.

1. See every day of the year, with enough space to add something
2. Make it easy to mark "ranges" for periods where I might want to stay abroad or go on vacation
3. See important calendar events like birthdays
4. See public holidays / weekends

A digital solution mostly compromised on visibility. In Apple Calendar it's not possible to see the full year, and still have some visibility into which events are happening on the individual days. You'd only see that there's something happening.

Analog, paper based calendars or dedicated yearly planners didn't have my personal calendar events, unless I'd manually transcribe all birthdays or other important events every year.

A few days ago I finally found a way to do that through a small company from Norway called Time and Date. They check all the boxes. The "Multi Month" calendar without their logo requires a small fee of 15$ / year for their Supporter plan.

After I configured everything I noticed one small but important missing feature. While they offer a way to add multiple lists of events (For example for yourself, and another one for a partner), they don't offer an ics (Calendar files, like you can export from Calendar.app) import. Without that you'd need to enter all events one by one through their web interface.

After observing some requests on their website on doing some digging into their internal API I gave that task to Claude and within 15 minutes I had a working userscript that I could use with Tampermonkey in Safari to quickly import all birthdays I've exported from Calendar.app on my Mac with one click.

If you are interested in the script, it's all here and should work out of the box: https://github.com/dewey/userscript-timeanddate-ics-importer

I then generated the calendar with the settings I wanted and downloaded the PDF. A quick stop at my local copy shop where I printed it in A2 and I had a nice calendar to tape to my door.

During the year I'll add additional birthdays and events I forgot, and transcribe them into my digital calendar at the end of the year, to be imported again on the 2027 version.

These days when you are running into an issue with some app or any paid(!) or unpaid service of a big tech company you are usually already prepared to just find another solution as support is generally non existant. If you are blessed with a response they will usually just feed you basic throubleshooting steps like "Have you tried to wipe your NAS and re-install everything?" or telling you to go away with some variation of "Thanks for the feedback, we'll take that into consideration for future product updates".

With that in mind I was opening a support request through Synology's website and provided some debug information I have collected, fully prepared to never hear from them again. Especially as my NAS is already many years old and I'm not paying for any current services of Synology.

Timeline

On the next day I've already heard back from them with some instructions on how to submit standardized debug logs from the Synology Drive client and server. I chalked this up to the usual automated process and didn't think much of it. I submitted the logs and a few days later, to my surprise, the ticket got escalated to a developer.

In the next ticket I was asked to give them access to my NAS through some temporary support account that you can easily create right through the interface. Through that they were hoping to reproduce this specific issue with my NAS and their lab setup directly.

I've then provided detailed timestamps and graphs of connection counts on my machine after starting Synology Drive with the following commands:

~ ❯❯❯ watch -n 5 'echo "$(date +%s),$(netstat -an | grep TIME_WAIT | wc -l)" >> time_wait_2025-12-31.csv'                                                                                                                                                                
~ ❯❯❯ gnuplot -e "                                                                                                                                                                                                                                                     
set terminal pngcairo size 1200,600;
set output 'time_wait_2025-12-31.png';
set datafile separator ',';
set xdata time;
set timefmt '%s';
set format x '%H:%M';
set title 'TIME_WAIT sockets over time';
set xlabel 'Time';
set ylabel 'Count';
plot 'time_wait_2025-12-31.csv' using 1:2 with lines title 'TIME_WAIT';
"

To my surprise I received a message from Synology a few days later. They were able to identify the issue and even prepared a patched version that I was asked to run.

We found that during a reindex process, the Drive Client may trigger a large number of file change notifications at once. In this scenario, those notifications were sent in a single burst, which can result in a very high number of short-lived TCP connections in some scenarios.

To address this, our team has prepared a patch that changes this behavior so that notifications are sent in batches instead of all at once, reducing the number of concurrent connections created during reindexing.

Please install the attached(on the left side of this ticket) Synology Drive Client patch on your Mac and observe whether the TIME_WAIT connection count and network behavior improve.

This version fixes the issue and it will be released for all customers soon. I'll be watching the changelog closely to know when I can switch back to the main update path and not running a patched version.

Update 2026-01-14: A new version containing that fix was just published with the changelog quoted below. That was even faster than expected.

macOS: Fixed an issue where syncing a large number of files in macOS On-demand Sync tasks could cause the computer to slow down or temporarily lose its internet connection.

Running something like Postgres on a server is easy. You either install some package and start the service, or you run it in Docker. This is just a small part of the task though, after that you'll have to research the right script for your backup cron jobs and figure out how to get alerted on failed backups. In the future you'll have to figure out how to restore a backup, upgrade your database or see what you have to do to create a read-only standby instance that receives replication events from your main database.

On a cloud hoster this is usually a few clicks away where you just select what kind of database you want and how many resources you think you'll need. You can be certain that backups work and you just have to select which point in time you want to restore to. This is something that, to my knowledge, doesn't exist yet.

There's a project that I've recently found that covers the "backup" use case, including webhooks being sent on successful or failed backups. It's called pgbackweb and seems to work fine so far.

I'm still looking for something that is closer to a cloud provider and includes provisioning, backups and other maintenance tasks like setting up replicas. Ideally it would not be exclusively for Postgres but also include other data stores.

In the past few months I had a lot of weird issues with my new (and with a clean install of macOS) computer. I especially noticed that while running brew update. It would just have issues connecting to the internet and immediately re-running the command usually fixed the issue right away. This was also noticeable in other tasks like updating Ruby gems. It was annoying, but not annoying enough to drop everything and start investigating.

It usually looked like this:

==> Updating Homebrew...
fatal: unable to access 'https://github.com/Homebrew/brew/': Failed to connect to github.com port 443 after 1 ms: Couldn't connect to server
fatal: unable to access 'https://github.com/shopify/homebrew-shopify/': Failed to connect to github.com port 443 after 1 ms: Couldn't connect to server
fatal: unable to access 'https://github.com/stripe/homebrew-stripe-cli/': Failed to connect to github.com port 443 after 1 ms: Couldn't connect to server
Error: Fetching /opt/homebrew failed!
Fetching /opt/homebrew/Library/Taps/shopify/homebrew-shopify failed!
Fetching /opt/homebrew/Library/Taps/stripe/homebrew-stripe-cli failed!

Problems like these are very hard to debug as it could be anything from something being wrong in your Wifi Access Point, a faulty Ethernet cable, buggy IPv6 or so some obscure DNS issue at your ISP (Usually it's just your local PiHole, but not in this case).

After not being able to fix it over months I gave Warp.app that task to debug. It's basically Terminal.app with built-in AI, so it can debug this much quicker and run all kinds of debug commands that I would otherwise have to piece together for hours.

It found two issues over a few sessions:

1. Stale network interfaces from Tailscale (I guess some buggy old version that didn't clean up after itself), but sadly cleaning that up didn't fix the issue.
2. Synology Drive was hogging all my open ports and that was the reason why new connections could not be established unless some of the stale connections expire. That's why it usually started working after a few seconds.

From then on I always had a Terminal window running the following command open.

watch -n 5 'netstat -an | grep TIME_WAIT | wc -l'

Sure enough it always hovered around 15k when Synology Drive was running. Quitting it immediately made that number drop down to < 5 connection attempts with the TIME_WAIT status. With these new search terms I was also able to track down a thread on the Synology forum that mentions this issue, sadly with no responses. Even with the right search terms I was not able to track down other uses with that issue.

The directory I'm syncing has 30k image files, which doesn't seem like an unusual amount. If anyone else runs into that and finds this through a search engine, let me know how you fixed it!

I had a monthly reminder for that task (The Amazon billing notification) so I finally set out to do that, dug up my old credentials and logged in. To my surprise I realized that you can't just delete the archive via the Amazon interface, even if there's a "Delete" button.

You also can't just delete it through the aws command line tools. You have to write custom code to generate a list of files (It's an async job, that takes ~4 hours), then delete them one by one (Following strict rate limitings from Amazon's side) and then 24 hours later the count on Amazon will update and you can finally use the "Delete" button on the empty archive in the interface.

There's many projects and scripts online to do that already but they didn't immediately work for me, so I just vibe coded my own which ran over night and finished successfully.

Maybe it's helpful for someone else too.

Step 1

Create an inventory of files:

   aws glacier initiate-job \
     --job-parameters '{"Type": "inventory-retrieval"}' \
     --account-id 1234 \
     --region eu-west-1 \
     --vault-name arq_123

You can check the progress with the following command, but expect it to take up to 4 hours.

aws glacier list-jobs \
  --account-id 123 \
  --region eu-west-1 \
  --vault-name arq_123

Step 2

Download the list of files in the inventory, once above tells you the job is finished.

aws glacier get-job-output \
  --account-id 534475282219 \
  --region eu-west-1 \
  --vault-name arq_123 \
  --job-id YOUR_JOB_ID \
  output.json

Then run the following Go script: https://gist.github.com/dewey/c402354c33e31cdd6e1f53d5a9e82f05

The output looks like the following, make sure you are logged into aws in your command line through something like: aws sts get-caller-identity.

Step 3

After 24 hours this should be reflected in the AWS interface and you should be able to delete the bucket.

~/m/g/d/aws-glacier-cleanup ❯❯❯ ./glacier-cleanup -account-id 123 -region eu-west-1 -vault-name arq_123 -input output-archive-ids.txt
2025/11/14 22:54:39 Starting AWS Glacier cleanup at 2025-11-14T22:54:39+01:00
2025/11/14 22:54:39 Account ID: 123, Region: eu-west-1, Vault: arq_123
2025/11/14 22:54:39 Using existing archive ID list from output-archive-ids.txt
2025/11/14 22:54:39 Total archives to delete: 756033
2025/11/14 22:54:39 Archives already deleted: 43965
2025/11/14 22:54:39 Archives remaining: 712068
2025/11/14 23:30:48 Progress: 35603/712068 deleted, 0 failed (5.0% complete) - 2025-11-14T23:30:48+01:00
2025/11/15 00:07:21 Progress: 71206/712068 deleted, 0 failed (10.0% complete) - 2025-11-15T00:07:21+01:00
2025/11/15 00:43:12 Progress: 106809/712068 deleted, 0 failed (15.0% complete) - 2025-11-15T00:43:12+01:00
2025/11/15 01:20:38 Progress: 142412/712068 deleted, 0 failed (20.0% complete) - 2025-11-15T01:20:38+01:00
2025/11/15 01:57:09 Progress: 178015/712068 deleted, 0 failed (25.0% complete) - 2025-11-15T01:57:09+01:00
2025/11/15 02:33:46 Progress: 213618/712068 deleted, 0 failed (30.0% complete) - 2025-11-15T02:33:46+01:00
2025/11/15 03:10:03 Progress: 249221/712068 deleted, 0 failed (35.0% complete) - 2025-11-15T03:10:03+01:00
2025/11/15 03:46:19 Progress: 284824/712068 deleted, 0 failed (40.0% complete) - 2025-11-15T03:46:19+01:00
2025/11/15 04:23:31 Progress: 320427/712068 deleted, 0 failed (45.0% complete) - 2025-11-15T04:23:31+01:00
2025/11/15 05:00:23 Progress: 356030/712068 deleted, 0 failed (50.0% complete) - 2025-11-15T05:00:23+01:00
2025/11/15 05:37:04 Progress: 391633/712068 deleted, 0 failed (55.0% complete) - 2025-11-15T05:37:04+01:00
2025/11/15 06:14:31 Progress: 427236/712068 deleted, 0 failed (60.0% complete) - 2025-11-15T06:14:31+01:00
2025/11/15 06:52:02 Progress: 462839/712068 deleted, 0 failed (65.0% complete) - 2025-11-15T06:52:02+01:00
2025/11/15 07:29:00 Progress: 498442/712068 deleted, 0 failed (70.0% complete) - 2025-11-15T07:29:00+01:00
2025/11/15 08:05:51 Progress: 534045/712068 deleted, 0 failed (75.0% complete) - 2025-11-15T08:05:51+01:00
2025/11/15 08:43:49 Progress: 569648/712068 deleted, 0 failed (80.0% complete) - 2025-11-15T08:43:49+01:00
2025/11/15 09:20:09 Progress: 605251/712068 deleted, 0 failed (85.0% complete) - 2025-11-15T09:20:09+01:00
2025/11/15 09:57:57 Progress: 640854/712068 deleted, 0 failed (90.0% complete) - 2025-11-15T09:57:57+01:00
2025/11/15 10:35:28 Progress: 676457/712068 deleted, 0 failed (95.0% complete) - 2025-11-15T10:35:28+01:00
2025/11/15 11:13:42 Progress: 712060/712068 deleted, 0 failed (100.0% complete) - 2025-11-15T11:13:42+01:00
2025/11/15 11:13:42 Finished at 2025-11-15T11:13:42+01:00
2025/11/15 11:13:42 Total deleted: 712068, Failed: 0
2025/11/15 11:13:42 Deleted archive IDs are in deleted-archive-ids.txt

1. Quickly get an approximate count of a large table by using the index:

SELECT reltuples::bigint AS estimate FROM pg_class where relname = 'bigtable';

2. Explanation why splitting off a frequently updated column into a new table is useful: https://dba.stackexchange.com/a/318548

3. Reviewing frequently used triggers and functions

alter system set track_functions to 'all';
select pg_reload_conf();
select * from pg_stat_user_functions order by calls desc;

The results of that can be exported as a CSV and then analyised in a spreadsheet.

4. Inspecting the differences between two versions of Postgres on depesz.com.

5. Generate curl commands to run batch requests.

This one sounds a bit silly but it has saved me countless hours over the years, sometimes it's very quick to just generate a long list of curl commands, put them in run.sh, chmod +x run.sh and ./run.sh it instead of writing scripts, or doing things directly in the database which might be riskier.

select 'curl -H "Authorization: Bearer test-token-123" http://localhost:8090/movie/' || m.id || '/update'
from movie m
where m.release_year = 2026
;

The downside is, that writing a blog post with images is always a hassle as you have to manually adjust file paths and write your text in an editor which is usually not optimized for writing prose. That editor then usually doesn't preview the images correctly as the paths are set to work within Hugo. In short: It's annoying (haha).

In times of AI it should be relatively simple to just build a custom GUI app just for this task and in the style of the blog. This turned out to be true, even though it would still take a bit of work and extra prompting to make everything work a bit smoother.

I've used Tauri and you can see some screenshots below. It worked really well to just tell Cursor things like "Make the app look like https://annoying.technology, get inspired by the stylesheet") and I think it worked out pretty well.

Overall it took me two short evenings to get everything working without writing a single line of Rust or CSS / HTML myself. I've first tried another Go based framework called fyne.io, but the results were much worse.

• Penang Hill, fun to go up by the funicular. Explore the "Habitat", a bit cooler air up there and nice views + monkeys.
• Ferringhi Bay, lots of little restaurants on that beach or close by if you keep walking along the coast.
• Yong Pin Restaurant, one of my favorite Dim Sum places anywhere. A new discovery!
• Narrow Marrow, a nice bar
• Le Petit Four Pâtisserie
• Your Local Club, good coffee and a good escape from the heat. In a very busy area of the town with lots of stores, restaurants around.

I took a lot of pictures in Penang, a small selection below or in the Photos section.

As usual the good people from iFixit have a detailed walkthrough, it's a very simple operation and in the end if you upgrade from 265GB to 2TB yourself you are saving almost 500 Euros compared to doing it through Apple directly.

The tools I bought from iFixit directly, the Mac from Apple and the SSD I ordered on M4 SSD. Everything arrived after a few days, and only then realized that M4 SSD already included some tools and Torx bits that I've bought separately from iFixit. Never hurts to have some more of these lying around I guess.

In the end there's a lot of things that could be a bit better but I also already spent way too many hours on it already. Part of that was related to me switching from Final Cut Pro to Davinci Resolve, as the free version is already very powerful and does everything I need. After some hours of usage I already feel right at home and the interface to add titles / subtitles is much easier to use than in Final Cut.

If you have a spare 7 minutes in your day, feel free to watch, like and subscribe!

For now I'm mostly active on Mastodon, but I was also aiming at making my dormant Bluesky account look a little bit more active without always having to manually cross-post my posts to both services myself.

There's already another promising plugin called IndieConnector that I've been following on GitHub for a while. It does that but also a bunch of features on top of that that I was not so interested in.

Here's how it looks like in the Kirby panel, it has a setting screen where you can configure everything and some instructions on how to set up a cronjob that acts as the scheduler.

The main interface lists the the posts that are in the queue to be syndicated (cross posted), and a list of posts that were already shared. If you post something new on your site it will automatically get added to the queue and will be posted after a (user configurable) delay. It always only posts one entry every hour to not spam your accounts.

I still have to find a way to make it more generic, as right now it still depends a bit on pages having a certain format (title, images, cover image) before I share it on GitHub and the plugin directory but that's coming soon.

Update: I've shared it on GitHub, give it a try and open an issue if something is unclear or doesn't work

Here's how the posts shared on Bluesky and Mastodon currently look like.

Let me know if you are interested in giving it a try!

Here's an example of the final result:

It turns out this is all relatively easy with Kirby. All we need are the following bits:

site/blueprints/blocks/photopost.yml

Here we create a new block type, the query prepopulates the select field with our list of photo posts.

name: Photo Post
label: "Select a photo post to inline a preview of the post."
type: photopost
preview: fields
fields:
  photopost:
    label: Selected Photo Post
    type: pages
    query: site.find("photos").children().listed().sortBy("date", "desc")
    multiple: false
    required: true

In our regular post template we need to add that new block type to the selection, an excerpt from that looks like this. The important part is the photopost block in the fieldset.

site/blueprints/pages/post.yml

title: Post

columns:
  - width: 2/3
    fields:
      title:
        label: Title
        type: text
      blocks:
        label: Content
        type: blocks
        help: Build your post content using blocks.
        fieldsets:
          - ...
          - photopost

site/snippets/blocks/photopost.php

The last part is the template that will help us render this inline photo post. In my case I put some of that logic into another snippet just to make it cleaner. The photo_entry_inline will just iterate over the images in the photo post and output them with some styling.

<?php
$photoPages = $block->photopost();
$photoPagesArray = $photoPages->value();
foreach ($photoPagesArray as $pageUuid) {
    $photoPage = $site->find($pageUuid);
    snippet('photo_entry_inline', ['photo' => $photoPage]);
}

That's all we need to get it our custom block up and running.

There's great resources from Kirby itself too (of course), like this video on the YouTube channel and the Custom Blocks Documentation.

The goal was to simplify the setup so it is much easier to add new features. The first one is the dedicated Photos section which is now also linked in the navigation bar. I'll use this to have a centralized place where I can just upload some photos, later I'll work on making this automatically get posted on Mastodon, and maybe Bluesky (POSSE!).

I'm pretty happy how the Exif display turned out, I got inspired by Sam Becker's photo blog project, but went with a much simpler version and some Kirby hooks that store the Exif information (What a fragmented format!) in the Kirby metadata based on the Blueprint.

While doing that I also organized my feed (No worries, redirects are in place). The new feeds can be found in the header, or here:

Posts: https://blog.notmyhostna.me/posts/feed.xml
Photos: https://blog.notmyhostna.me/photos/feed.xml

Let me know how you like it!