Their FAQ actually has an audio clip, haha https://forgejo.org/static/forgejo.mp4

The only way to stay sane

Anubis was originally created to protect git web interfaces since they have a lot of heavy-to-compute URLs that aren’t feasible to cache (revision diffs, zip downloads etc).

After that I think it got adopted by a lot of people who didn’t actually need it, they just don’t like seeing AI scrapers in their logs.

I learned coding at age 7-8 by messing around with the scripts of the built-in demo stacks in HyperCard. It was close enough to English that you didn’t need to study syntax but could easily learn from example

Devices like laptops, tablets and phones, usually do not have Ethernet built in, or are too mobile to make it practical to use

What I did in the living room was plug a USB-C dock with a 2.5 Gbit Ethernet adapter into the wall outlet with a 2 meter USB-C 3.x cable.

So I sit down in the living room and plug in my laptop/phone in to charge when I’m using it and they automatically get a 2.5 Gbit network connection. Even iOS natively supports the common Realtek 2.5 Gbit chipset.

This is complete BS, I could find zero sources for that claim, and several debunking it.

The only tangentially related thing I could find was that in colder climates, they need heat to de-ice the wings, and at one point, the power supply to a Scottish wind farm was cut off, so they put in some temporary diesel generators on-site to power the de-icing system to get the turbines going again.

Have you checked the SMART value 199/0xC7 “UltraDMA CRC Error Count”? This should tell you how many checksum errors happened on the SATA interface between the CPU and the drive. If this is higher than zero your hypothesis is correct and there’s something bad with the connection, if it’s zero then the problem is more likely to be elsewhere.

BTW the Brother scare about them adding DRM that was in the news a while back turned out to be false, it was just a random guy on Reddit with a bad third-party cartridge, and Brother replied that they do not block third-party cartridges.

That said, I’m not a huge fan of their weird PPD installer on Linux that installs some random, undocumented crap

It could even be something “innocent” about how ad-blockers have started to interact with the site as YouTube ramps up their anti-adblock measures and the ad-blockers have to change how they work. Like maybe the ad-blockers have started blocking the JavaScript callback that logs the views.

Even before this drop in views, the rule was you have to watch a video for at least 30 seconds before it counts as a view, as a way to combat clickbait where people instantly bounce from a video. Maybe they have changed this further? Or they change some kind of bot detection?

On broadcast TV, a 30 minute timeslot had only 23 minutes of actual content and 7 minutes of ads.

That’s what we’re heading back to. 20% of the watch time is ads.

Everyone here is speculating about their content, but the simple answer is YouTube just changed how they count the view number. The change basically happened overnight, so it’s not some slow attrition of views. They said in the WAN Show that while the view count halved, the number of likes hasn’t changed (the view/like ratio doubled), and the revenue they earned hasn’t changed (CPM doubled). All of this points that the same number of humans are watching, but what counts as a view in the “views” number just changed.

They literally don’t know. “GPT-5” is several models, with a model gating in front to choose which model to use depending on how “hard” it thinks the question is. They’ve already been tweaking the front-end to change how it cuts over. They’ve definitely going to keep changing it.

All outlets should just be replaced with IEC C13. Robust and compact.

Instagram is extremely popular, and it’s heavily promoted inside of there, with Threads content embedded to almost look like Instagram content but when you tap on it it hops you over to Threads. I’m not surprised that they’ve been able to build a user base while X declines

In Japan they have an IR beacon system to track traffic congestion which works anonymously and lets offline car navigation systems have good-quality traffic info.

I picked up a used 2018 Fujitsu office PC with an i5-7500 for $60 (from a physical recycle shop, with a 14 day warranty) and it draws 15W idle. Way better value than a Pi (once you’ve added case, cooling, PSU etc) for running home server stuff.

A Pi still kills for “Arduino plus plus” use cases where you need the size, GPIO or can optimize the heck out of power usage on a battery.

iOS does have an API for apps to record the screen throughout the OS these days through Broadcast Extensions, but it has to be user-initiated through the control center screen recording toggle (where they then get to pick what app to record the screen to instead of just saving as a video), it wouldn’t do that people think the T-Mobile app is doing

1. I do have 10 Gbps, I pay $35/mo here in Japan (not even a big city like Tokyo, this is a depopulating, rural capitol)
2. More importantly, even my 5 year old, 4-bay spinning rust Synology NAS can saturate 2.5 Gbps copying files. With soldered storage in modern machines, faster networking is cheaper than replacing my whole machine

Yeah what I’ve settled on is one of those $40 generic Chinese 4x2.5G PoE+2x10G SFP+ switches. Gives me:

• 10G for internet/router
• 10G for my main computer
• 2.5G for secondary machine
• 2.5G for NAS
• 2.5G PoE for WiFi
• One port chained to a 16-port Gbit switch for all the slow junk that doesn’t need performance

Would be great to get another 10G for the NAS as well!