Devs are lazy. Any productivity gains are negated by them simply choosing to work less and have more free time. Which is not a bad thing.

I think you can allow others to relicense your code, that’s what permissive licenses effectively do.

EUPL does not seem to do that however. It only lists other copyleft licenses that are compatible with it (IDK why it’s done this way).

The biggest hurdle are messengers, especially including cross-platform support and mobile devices.

She is the queen of the Oldest House.

Jesse Faden would manage it just fine.

AFAIK compilot just provides integration with visual studio and GitHub, it doesn’t have its own models. You pay through Microsoft to use use other companies’ models (like GPT or Claude). Though agent harness itself has a big impact on the performance of models.

Which ones don’t support it? It’s a very old format at this point. Every Android device in existence can work with it for example.

You can self host xray or sing-box server (which implement many of the protocols) on a cheap vps. A lot of documentation is not great and is either in Chinese or Russian but it’s doable. There are also paid services offering vless with various degrees of reliability and sketchiness.

Of course the real challenge when living in an authoritarian country is how to pay for it. I.e.you need to buy a vps abroad. I guess it will be easier in USA then in countries under sanctions or with internal restrictions where people generally can’t use Visa or PayPal.

Protocols that mimic regular https traffic in a way that survives DPI. There are many variants like vless+reality+vision, naiveproxy, etc. It’s still cat-and-mouse game since government’s DPI systems are constantly improving.

Not if they start using DPI and block VPN protocols. VPNs that use OpenVPN, Wireguard and other non-censorship-resistant protocols (including Mullvad) don’t work in China, Iran or Russia.

Mullvad and other mainstream VPN providers only work in free countries where VPN is allowed.

Some patched it to change the default in config.

There is zero chance it will be released (even if it’s true)

It will still run necessary calculations much slower than a gpu, so you will need to run it much longer for the same result, which is not only inconvenient but will likely use more power in the end anyway.

This will just waste even more energy since cpus are less efficient at this kind of task.

LLMs require insane amounts of compute, period. The hardware to do that needs energy and emits heat, there is no way around.

That’s still just moving the impact from data centers to end users’ machines.

Gamers (especially those that own 5090) are a tiny minority compared to llm users.

It’s really easy to not use ai. It’s not a necessity and not even a part of any “lifestyle” that some people would miss.

They are not magic. You are still using electricity, and for the same amount of work that cloud providers do you will consume the same amount. If every one of the billions of people using chatgpt or whatever buys rtx 5090 and makes it churn 24/7 for their “local” ai then nothing will change for the environment.

So you don’t care about environment, huh?

Yep, it’s all binaries.

In Java world package managers and package registries work with prebuilt libraries because for JVM bytecode it’s sufficient. And if they have native dependencies then they include compiled .dll or .so files that are built on developer’s machine. This makes things simpler but has its own drawbacks (these dependencies work only on architectures and OSes developer thought of, there is no guarantee that the library will be compatible with user’s environment, etc).

In Rust world everything is built from source, including native dependencies. Thsi makes build process much more complicated and necessitates running code on user’s (developer consumes the library as dependency, not end user) machine to build everything that’s not Rust.