Many companies now provide some kind of agentic tooling and expect its use. Some mandate which one, some are open. Most however, use US products and are sending all their data to Anslopic, Microslop, Goolag, ClosedAI, and so on. Not only that, they are expensive af. I've seen bills above 600€ per employee where management didn't even flinch.

There are "inference providers", but every single one on HuggingFace is a US company. There are also the model providers themselves but again, US or Chinese with the exception of Mistral in Europe.

To be able to convince management to use something else, it would have to be cheaper yet "powerful" (DeepSeek fits but is in China), have enough choice (it can't be a single model provider like Mistral, Anslopic, etc.), and be in Europe.

Does such a thing exist?

 

cross-posted from: https://lemmy.world/post/51211518

Context is this issue at neovim [dot] github [dot] io: https://github.com/neovim/neovim.github.io/issues/501

The neovim website has an impressions section, under which there's an impression by DHH:

"Full-screen Neovim looks cool as hell!” — DHH

Multiple people want that quote removed.

For those who don't know, DHH (creator of the Omarchy Linux distro) is seen as a problematic individual by many in the Linux & OSS community:

 

Code examples that use this convention are in my opinion terrible for comprehension - especially for domain specific examples.

Say you have a library called libprefix and you want to demonstrate how to use it

foobar example

import prefixer

foo = "foo"
bar = prefixer.prefix(foo, "bar")
print(bar) # foobar

vs representative

import prefixer 

to_prefix = "to prefix"
prefixed = prefixer.prefix(to_prefix, "not ")
print(prefixed) # not to prefix

The variables tell you what they represent

  • to_prefix a string that has to be prefixed
  • prefixed contains a string that has been prefixed

This is just a small example, but some example have 10, 20, 50 lines of code and when I start reading and it uses this foobarbaz crap, it makes me want to close the tab and move on. Naming things is hard, but come on...
Sometimes there are even domain specific examples where the effort is made to convert it to foobar convention just for the sake of it.

import traffic_light

foo = traffic_light.intersection("cross", 2) # 2 roads intersect
bar = { "north": 4, "east": 1, "south": 10, "west": 0} # vehicles at each light
baz = foo.simulate(bar) # creates a simulation object

print(baz) # print out the state of the intersection (bar in this cas)
quux = baz.step() # state of intersection after one step
print(quux)

We have all seen worse than this. Seriously, why?

 

It's that time again and I'd like to make my life easier. I have multiple CVs for different positions and in different languages. Recruiters and companies want the CVs in PDF.

My current workflow is to sequentially edit the .odts (wording then layout because things move around depending on length), maybe the style, then save and export to PDF. That can take a while, so here I am, wondering if it can be sped up by simply editing a text file and generating CVs with certain themes.

Or what your workflow is. Maybe this isn't worth automating...

Solution

Finally went with RenderCV (thank you @bert@sh.itjust.works), which also runs locally. YAML in, PDF, PNG, HTML, etc. out. It comes with 10 themes or so. Haven't looked at what it take to create my own (I'm lazy).

Contender was JSON Resume, but the site is down and I tried it years ago but never got it to work properly.

 

Equivalents because VSCode and its forks are not an equivalent.

TL;DR Jetbrains is getting worse, more like VSCode. I want out to something as powerful but not VSCode like (which is worse).

Jetbrains has been chasing VSCode's popularity to the detriment of its own product.

The modal commit dialog is now opt-in, behind a nearly hidden, advanced parameter and will be removed in the not so distant future. This was part of their effort to make the UI "leaner" and more like VSCode so now Jetbrains IDEs look more like it.

Some keyboard shortcuts have also had to be changed to be more like VSCode.

Jetbrains used to have a good docker integration, but that's either unmaintained or undergoing work to support dev-containers better - something that jetbrains practically had nearly a decade ago, which allowed you to code on another server (remote development).

AI Assistant and Junie (their coding agents which I have to use for work) have consistently been behind the curve and just keep getting worse. They have 2 separate products for agentic development and can't decide which to keep, so they duplicate efforts and fail at both. Now I'm asking my employer to just give me Cursor like the rest of the team, so I don't have to put up with this Jetbrains crap.

Privately, AI it's non-starter anyway, but my guess is that Jetbrains is also focussing more development efforts on that than actually fixing existing bugs.

 

I've found this very good website. Many torrents are unofficial and created by the website staff. Some torrents are official from the project maintainers.

I have been seeding torrents on me VPS for some time. Since I don't use most of the bandwidth. Now I have purchased an HDD IPv6-only VPS in the same datacenter to use as NAS and seed more torrents.

If you've got some spare bandwidth and disk space, you can seed the softwares you like.

 

I've been testing VPN services and somehow every one with a client has written their own. Proton, Windscribe, IVPN, Mullvad, AirVPN, NordVPN, and the list goes on.

Don't these services all do the same thing? Connect to a server to get a list of servers, then use WireGuard or OpenVPN to make a connection, and have some fancy stuff like split tunneling or a killswitch. It should just take configuration by adding profiles per service and it should be possible to use the same client.

What am I missing?

 

Goal: Take a recording of someone speaking, pass it through a CLI or GUI, output a recording with the speaker's voice replaced by an AI voice.

I know there are closed-source or hosted services that do this, but they are a privacy risk and most surely also continue training their models on what users submit. The opensource solutions I have seen (last I checked 6 months ago or so) were all extremely outdated (around the beginning of the ChatGPT release) and required training using thousands of voice samples.

Surely somebody has done this before and didn't need to be an expert in the field to do so...

Non-solutions:

  • hosted services (elevenlabs, chatGPT, Gemini, ...)
  • "train your own AI"
 

Following a complaint filed by association HOP – Stop Planned Obsolescence, Steering Group member of the Right to Repair Europe Coalition in September 2017, the printer manufacturer Epson has been summoned before the criminal court on charges of planned obsolescence and misleading commercial practices.

#PlannedObsolescence #RightToRepair #France

 

Archived version

A month or so ago, the Netherlands reportedly blocked a US company from buying the cloud provider that runs Dutch digital identity (there is a post about it in this community here).

The more recent news is that the Chief Privacy Officer of the Dutch government, who was behind this initiative, is about to lose his job.

...

For more than four months, Pieter van Oordt warned internally about the risks of the takeover. When these warnings were ignored, he brought the issue to the media, the Dutch Parliament, and the Cabinet. He showed that Parliament had received incomplete and misleading information and revealed that vulnerabilities in digital infrastructure had been shared with a U.S. company. His actions forced the Ministry of Economic Affairs to block the takeover.

Instead of protection, Van Oordt faced retaliation. He was excluded from meetings, his salary increase was blocked, and his request for protection under the European Whistleblower Directive was rejected, despite an expert report confirming that he should have been protected. He has been suspended, and his dismissal is mentioned in a written notice from the Attorney General on 22 May 2026.

The dismissal has not yet been implemented.

...

Someone started a petition (not me), I post this here as you may want to sign it (and maybe spread the word).

The petition calls on the European Parliament and the European Commission to act:

  • Protect Pieter van Oordt from an unlawful dismissal.
  • Ensure he can continue his work safely within the public sector.
  • Place digital sovereignty of vital infrastructure high on the European agenda.
  • Hold the Dutch government and the responsible ministries accountable.

[To read the English version of the linked text, you need to scroll down. It's below the Dutch version.]

view more: next ›