VonReposti

From a data standpoint it's also pretty bad. Your request to Netflix is a few hundred bytes. Their response is gigabytes.

And it's not the data transfer costs that's the monetary bottleneck. It isn't even the data storage costs (unless you run a media server, but let's ignore that for a moment). It's the computation costs that is the largest cost center for most websites. You can run simple static pages on a potato but if you have significant visitors you need CDNs or significant infrastructure to handle it. If you start to make dynamic pages, maybe even web apps, then your computational workloads will increase significantly too.

Paying for an HTTP transfer, for example, would not be able to distinguish between a simple HTML website or an AI prompt.

Capitalism may have some ugly sides but the fact that you pay to prompt Claude but can see my simple static HTML website for free is a way to distribute the money to where your activity costs money. If you want to burn down the rain forest to power your AI prompt, then you pay for that. If you want to see a blog about kittens, then you pay by seeing an ad (or maybe not at all if the website is run purely as a free hobby project). My personal website is running off of my old gaming PC so the only cost is the electricity. But I am running most of my services I need on there too, so the public website is negligible in comparison. A byte transfer wouldn't even be worth a penny while Claude probably could be measured in dollars.

And the depreciation of hardware is magnitudes faster than fiver optic cables. You can't slowly snatch up hardware over the course of a decade. In a decade it would be electronic scrap.

Fra en opfølgende artikel:

Ruslands ambassadør i Danmark efter DR-afsløring: Det er 'en løgn', at vi vil angribe europæiske Nato-lande

Så kan vi trygt læne os tilbage som da oprustningen ved Ukraine var en løgn.

Why do the sticks point the same way while in the mirror they point in opposite ways?

Jeg kan ikke se hvordan konkurrence mod DSB vil løfte kvaliteten af den infrastruktur som Banedanmark stiller til rådighed.

Vi ser virksomheder der brænder en måneds AI forbrug af på under en dag efter at GitHub gik over til usage based pricing. Og selv med det så bruger de store AI firmaer omtrent 10 kr. for hver krone du giver dem. Jeg forudser at der kommer et markedschok som ender i en markedskorrektion på niveau med dotcom boblen (og kan måske sammenlignes med Great Depression for USA hvis de fortsætter deres økonomiske vej).

Det betyder at det går op for folk had AI reelt koster og de finder ud af at bruger enormt mange ressourcer til at få AI'en til at "tænke" for dig er for dyrt. Jeg vil mene at vi går et skridt tilbage mod en glorificeret autocomplete, da systemer som Lovable har svært ved at lave "enterprise-grade" kode (maintainable, extendable, sikkert, etc.). Det kræver en dygtig udvikler til hele tiden at påpege de fejl som AI'en laver som brænder igennem tokens.

Så lige nu ser jeg en lille nedtur for udviklere nu hvor at virksomheder kunne få en gratis omgang, men på sigt er der intet at bekymre sig om. Cobol var også lavet så forretningsfolk selv kunne udvikle og dermed gøre udviklere arbejdsløse. Det samme med fjerde generation af sprog, CASE (Computer-Assisted Software Engineering) og til sidst BPMN. Udviklere ville blive erstattet. Men vi er her stadig. Det viser sig at der er brug for folk til at forstå hvordan man bruger en skruetrækker, uanset om den er med bits og strøm eller ej. Det samme med kodning. Udvikling forsvinder ikke. Vi er bare i en lidt mørk tid indtil det går op for virksomhederne hvad det reelt koster at udvikle "enterprise-grade" kode udelukkende med AI.

Det sker efterhånden ret ofte. Gad vide om det er aldrende køreledninger på Sjælland eller pantograferne der er misligholdt. Jeg kan ikke forestille mig at det skulle være andet.

Jeg vil nok give det en uge mindst og så komme med en opfølgning. En god leder forstår at man er spændt, venter afklaring og måske skal videre i livet. Her kan en lille reminder hjælpe, så du ikke bare er passiv, men er en aktiv søgende. En leder der bliver frustreret over at få en reminder en uge eller to senere er højst sandsynligt heller ikke en leder du ønsker at samle på.

I loved it when I first encountered it. It was simple, elegant, and easy to use. Over time though, things got bad. I relied on an extension to show the taskbar on all monitors as well as the "Other locations" tab in the file manager to see disk usage plus some other things but these two broke the camel's back. Every upgrade the extension stopped working for a few months, but at least I could delay upgrading for a bit. That was until the extension maintainer went AWOL, so no update was made for at least a year.

The "other locations" tab showed disk usage just a click away from my usual workflows. As someone who has a habit of making high utilisation of my disks, keeping an eye on disk usage was required. That was until Gnome decided we were too good for such an easy location so now the only place to see disk usage is in the disk usage analyzer that 1) is rarely used 2) takes a while to start up while it's scanning the entire disk. My habit of checking disk usage thus died. Until I had to upgrade to the new distro version. And it turned out I didn't have enough storage left to carry out the installation so my laptop bricked itself halfway. I was lucky that I could boot with a rescue image to clear some space and continue the upgrade but the first thing I did when it was finished was to install Plasma and kill Gnome.

That's what happens when you leave it in a dry dock for too long.

I haven't been running experimental, so I've yet to experience the new features

I have been looking forwards to the new vehicle patching, I hope it'll solve the issues with collisions my current paths have.