Simple means different things to different people.
I self-host Ghost and find it pleasant to use and low maintenance. It is a single Docker container plus MySQL. I recommend a reverse proxy in front of it like Nginx. There are importers from many other blog formats.
Years ago there was a voice to text transcription service sold as automated that worked by people listening to your voicemails and typing them out.
For bookmarking: https://raindrop.io/
But it’s not self-hosted and I’m not sure it supports offline reading.
Framework has an opportunity here: they may not be able to build in an ergonomic keyboard with Kailh keys, but they _could_ split the space bar and make it into 2 or 6 keys.
Framework has announced a “One Key Module” which will be available later this year and will allow building custom keyboard designs for the Framework 16, which already uses QMK.
Some details are still unknown, but it seems like something like a Corne layout will soon be possible.
What keys to you use to activate custom layers with Katana?
A nice thing about custom keyboards is that they designed with easy to reach keys that can be used for layer switching.
First keyboard with copy and paste keys: 1987 First release of QMK: 2016.
About 30 years later.
It isn’t hard when every works perfectly but there is a tremendous amount of complexity in some of these apps and a huge range of quality, documentation and required env vars and mounts.
And so, so many ways for things to break.
You still have manage upgrades due security vulns in all the features you are ignoring.
Yes. DMZ on router 1 exposes router 2 IP to internet.
No, this is all happening in the browser, there are no other image manipulation tools being called.
I just tested the new release. Consider defaulting PNGs to convert to JPEGs unless they have a PNG-specific feature like transparency. Lots of screenshots are initially PNGs, but not because they need any PNG-specific features. Consider: In a test screenshot, it compressed 3.4% with the default 80% setting and PNG->PNG, but for PNG->JPG, it compressed 84.6%. 
MCP sounds like a standardized way for AI clients to connect to data sources, the Model Context Protocol.
https://www.anthropic.com/news/model-context-protocol
It sounds like it may compete some with Google’s A2A protocol, which is for AI agent to agent communication.
Both share the same goal of making services easier for AI to consume.
Do you have a source to cite for the literal 99%?
The top-rated answer to this question on the Security StackExhange is “not really”. https://security.stackexchange.com/questions/189726/does-it-improve-security-to-use-obscure-port-numbers
On Serverfault, the top answer is that random SSH ports provide “no serious defense” https://serverfault.com/questions/316516/does-changing-default-port-number-actually-increase-security
Or the answer here, highlighting that scanners check a whole range ports and all the pitfalls of changing the port. Concluding: “Often times it is simply easier to just configure your firewall to only allow access to 22 from specific hosts, as opposed to the whole Internet.” https://security.stackexchange.com/questions/32308/should-i-change-the-default-ssh-port-on-linux-servers
Using a nonstandard port doesn’t get you much, especially popular nonstandard ports like 2222.
I used that port once and just as much junk traffic and ultimately regretted bothering.
I’m glad to have some competition for the Frost Oven Squoosh, which is being lightly maintained. I opened some issues in the Mazanoke issue tracker for some features to consider.
One feature I started on for that project but got stuck on was implementing a STDIN / STDOUT CLI workflow.
https://github.com/frostoven/Squoosh-with-CLI/issues/10
As I said there, the goal was a workflow where I take a screenshot, annotate it, optimize it, copy it and paste it into my blog… without creating any intermediate temp files.
At least on Linux, all the the steps of the pipeline are solved, except for a CLI image compressor that could accept an image STDIN and produce a compressed image on STDOUT.
Sounds like an oversight. Consider filing a bug with them.
Nice. I use Squoosh for this, which is also free and runs in the browser.
Https give you encryption in transit. The files you view will be accesible to the host.
Same idea with email.
Nope. Some algorithms are fastest when a whole data set is held into memory. You could design it to page data in from disk as needed, but it would be slower.
OpenTripPlanner as an example will hold the entire road network of the US in memory for example for fast driving directions, and it uses the amount of RAM in that ballpark.