Ariel Millennium Thornton @arielmt@computerfairi.es

Like it or not, Microsoft has a productivity suite monopoly, and your program can't break that monopoly by being scarier and more confusing for ordinary users when saving files.

Opera didn't break Microsoft's browser monopoly, but Firefox sure did (before Google captured it again for itself), and it was by being genuinely easier and better for ordinary computer users, not just for tech-minded nerds like me.

But the whole reason my customer put up with this bullshit fueled batshit insanity is because the LibreOffice folks at the Document Foundation still haven't figured out that they need to treat Microsoft file compatibility as a **first**-class feature if they want ordinary people using it.

The human robot fumbled for another hour before getting stopped by the same problem.

But here's where things took a turn: The human fixed whatever broke.

After two hours of fighting against a convincingly confident bullshit generator passing for the as-yet-unrealized promise of AI.

The daffy boxes of scrap metal named Copilot and Store Assistant were cheerfully eager, utterly useless, and completely unhelpful. But after an hour of boxing both it and Microsoft's old-store not Store checkout (because of course Microsoft has two radically different things with the same name), it connected us to a human robot powered by natural intelligence.

Adding the digital download to the cart was fine, but the first step of checkout is logging in again because you're making a purchase.

And then needing to log in again and again, ad infinitum, never getting past that crucial first step. The login options are by password or by verification code, but the result is the same. It shows an error message when bad data is entered, but another login prompt when good data is entered.

Did you know that Microsoft still sells Office for home computer users as a one-time purchase with perpetual license, just like they did before pushing Office 365 and Microsoft 365 into everything?

One of my customers did, and they hired me to buy and install it for them. (On their MS account, with their card, and with their close supervision every step of the way invited, of course.)

Did you know that sometimes Microsoft doesn't want your money at all? Neither of us did.

However, the linked GitHub repo is gone now, and the user has no public repos at all. This may or may not be a fork of it: https://github.com/dpajak-synaptics/on-device-ai-assistant

Such promise, already ruined by other forces within three months.

LB ( https://petewarden.com/2025/10/16/why-does-a-local-ai-voice-agent-running-on-a-super-cheap-soc-matter/ ):
The article doesn't address which LLM that local AI is using, but a comment below it does: https://petewarden.com/2025/10/16/why-does-a-local-ai-voice-agent-running-on-a-super-cheap-soc-matter/#comment-181304

The answer I'm reading is, it doesn't use one at all. Just voice recognition, pattern matching to choose a pre-approved response, and TTS for the chosen response. So no chance of a dishwasher telling a customer to check for power by sticking a fork in the wall outlet.

Mozilla Developer Network, "Realizing common layouts using grids": https://developer.mozilla.org/en-US/docs/Web/CSS/Guides/Grid_layout/Common_grid_layouts

Reading guides like that while lamenting the shortcomings of vintage Web browsing, even via RetroZilla in a Windows 98SE VM, does strange things to a girl, y'know?

Related: CSS grids are 3 years older than Chromium-based Microsoft Edge and only 2 years younger than EdgeHTML (Trident)-based Microsoft Edge Legacy.

I'm reading that the CSS "display: contents;" on TR and (implicit) TBODY elements wrecks accessibility of tables as tabular data, rendering their TH and TD cells as ordinary content with no tabular relationship instead, but since these are tables "misapplied" for layout only and never meant to be tabular data themselves (one of the only ways to control page layout in pre-CSS and CSS1 days), that seems like a perk in this specific case.