Basically you give it a directory with a few thousand randomly named meme images, and it asks a model to generate descriptions for each of them, that you can later search? Sounds useful.
I wonder how that policy change affects library usage, because while I've still never felt a need to code anything using an LLM, I rely on e.g. Valibot for input validation.
Docs for installing LanguageTool manually are here: https://dev.languagetool.org/http-server - version 20260528 works on Debian 13 under WSL, needs openjdk-25-jre-headless and the --public option. The extension settings have a localhost option, I didn't have to change anything else.
Note that premium as of today costs $2.50 per month if you pay for two years in advance. Running LanguageTool on a laptop isn't very fast for languages other than English, so paying them for the cloud might provide a better experience than self-hosting.
For plain text files, such as changelogs and translation updates, I have a pull request that makes FreshRSS track diffs, without overhead of a full browser: https://github.com/FreshRSS/FreshRSS/pull/8107
Thought "behind the scenes" in the title, three authors with good credentials, and featured article category, would mean the article presents some interesting technical details about the bugs, from which a software engineer could learn to plan better programs. But the article is just a verbose, non-technical ad for Claude, same as "The zero-days are numbered" by Firefox CTO in the Mozilla blog.
I'm still sure that direct monetary investments in work on software commons, from public budgets, without AI sellers in the middle, would result in similar or better fixes while producing public knowledge and new generations of specialists. Anthropic isn't doing a public good by donating early access to an LLM API to Firefox developers, it's doing extortion: we'll sell this model to anyone who pays, including your enemies, so give us good PR if you don't want them to get it first.
Because their app is essentially a website. News, videos, photo galleries. WP REST API is useful for writing a front-end using a different language than PHP while keeping the very convenient admin interface that most content managers are familiar with.
That's why I like it) Simple, accessible, big buttons, high-quality icons. Personally I use KDE when on GNU/Linux, but I definitely see what in GNOME would appeal to people sharing a computer with grandparents or children, or people who value design aesthetics more than extra functionality.
I'd also mention Nextcloud Forms. A common web development task is to automate something whenever a new form result appears. For example, to publish a review on the website while also sending a notification to the marketing department, recording a conversation in the CRM and sending the client a ticket number. This is often done by polling a CSV or JSON export of the results. Nextcloud offers this functionality, and it's reasonably private when self-hosted. Also due to its lack of end-to-end encryption its forms load much faster and on older devices, too.
My netbook has a keyboard without Home, End, PgUp, PgDn, Ins and Del keys. Vim and Emacs modes make it significantly more usable while remaining small. Vim is also convenient in Termux, for quick remote fixes from a phone.
Haven't had memory issues on my Windows 11 laptop so I don't check how much is free. On the other hand, free disk space is jumping between 0 and 10 GB without notification as to what appears or disappears. This is annoyingly unreliable because the space often gets closer to 0 right at the time when I need to Syncthing files with others.
Remembered that I made such a tool for myself ten years ago. Dusted off a backup, updated dependencies (and replaced some), refactored somewhat, changed license to AGPL and uploaded here: https://codeberg.org/nykula/imgie
Should be very easy to install because the backend is just ImageMagick and SQLite.
Beware of a 250M node_modules, though. My code is less than 1K lines in the initial commit, but the linters, bundlers etc are the same as I use for big projects.
Note: Lufi encrypts uploaded files and lets one share a link containing a decryption key. It doesn't let one expose images for other websites to embed. Thus a good tool but not for OP's purpose?
Upd 22:04: tried setting it up with Podman instead of Docker, and the instructions didn't work, first because of missing directories and then a permission issue. However, this can be because I tried on WSL rather than a dedicated GNU/Linux box.
If there's a need to introduce students to virtual DOM, I'd choose a library that doesn't require language extensions or editor plugins, but allows to easily code with just the syntax of JS itself: properties as Plain Old JavaScript Objects, map/reduce with arrow functions, ternary operator, variables as usual, components as functions. Adding Mithril to existing WordPress websites for making dynamic parts such as calculators or quizzes was quite straightforward. On the other hand, its API is somewhat less convenient than what React offers.
Probably like this: a threat actor [for authoritarians pushing ID laws and locked-in hardware], just not in a way that any of us [computing freedom enthusiasts] would consider as being a real theat [for the people].
Basically you give it a directory with a few thousand randomly named meme images, and it asks a model to generate descriptions for each of them, that you can later search? Sounds useful.