I got hired by a local company exactly for my experience with running linux servers at home and in a semi-professional capacity. They had hired someone with perfect credentials from university with several masters degrees, but they had to let him go after half a year because he did not fit in to their linux environment that has been operating since the 90s. I had no formal education, but many years with tinkering, self-hosting and operating linux machines for a small number of customers, and they could not hire me fast enough. They told me that it is really hard to come by people with this mix of experiences, everything is in the clouds these days.
I do NOT consider the Fediverse and the myriads of implementations of it to be social media, but rather a social web. More like websites with the abilities to communicate and interact in different and interesting ways.
Social media is dead, and has been for a while. Many use it still, but it is not primarily social. The social part was mainly a ploy to get peoples attention and then badly abusing it in ever more creative and sinister ways.
I do NOT consider the Fediverse and the myriads of implementations of it to be social media, but rather a social web. More like websites with the abilities to communicate and interact in different and interesting ways.
Social media is dead, and has been for a while. Many use it still, but it is not primarily social. The social part was mainly a ploy to get peoples attention and then badly abusing it in ever more creative and sinister ways.
EDIT: This comment was meant to be posted to the parent comment!
But for how long? The subsidized phase is probably short, and then what? I run Qwen 3.5 27 Dense om my old AMD RX7900XTX at about 45 t/s and barely use my Claude Code subscription anymore.
We have set up two DGX Sparks at work and are self sufficient for our AI needs. It is not SOTA, but it works really well for our needs. No matter what happens around cloud-hosted AI in the future, we will have decent in-house AI without further investments or expenses. We are a company of 24 people.
What they need to figure out is what topics peaks their interest. Kids need exposure to a broad spectrum early, get interested, and then have mentors that know how to run with it and harness that motivation. Later on these kids can tolerate learning more mundane, boring stuff if that brings them closer to a goal they have set for themself. But motivation has to come first!
I have used it almost daily for the last few months, with only 3 simple extensions. On this project I spent a few minutes adding context to an AGENTS.md and having it write a todo list based on the spec, then I start a new chat for each task on the list. Most changes are just a few lines of code so it is very quick to review. But this a very small website project though.
I planed to make something like this as a skill for my self, but pasting this into any agent chat works like a charm. I just had my local model (Qwen3.6 27B / pi) make a list of all the required standards an older hugo site of mine was missing, make a todo list and then run through the whole thing one by one, giving me chance to review each change.
It even made the missing favicon by cropping out the symbol from the logo, and it looks good!
A window into novel technology, commented on by knowledgeable people from around the world. Also, proof that some online communities can survive and thrive for decades while staying largely the same (no big redesigns etc).
Nowadays I use our local setup 95% of the time, but it is not that long since that flipped for me personally.
Context: I have a $20 Claude Code subscription, and have used it for a handfull of small-ish projects the last year, in parallel with local models on my AMD 9700XTX (24GB) at home. Mostly Ministral 14B and more recently Qwen3.6 27B Dense 4q.
Historically, the tooling (interferens engines and harness) has been the biggest challenge when using local models, a lot of the benefits from Claude Code was a rather unified and well oiled agent system. Local setups often bring with them sutle incompatibilities between models, inference engines and agent systems that are not obvious from initial testing, but cause trouble on projects larger than a couple of files.
The Spark setup at work is now at a point where I do not miss Claude, like at all. A big part of this is the harness and the tools available to the agent, most critically a good tool for searching online. I use my Kagi subscription to allow the models to fetch up-to-date information, and the Kagi MCP I use also has a summarizer which is very helpful in avoiding rapidly filling up the context window.
I mostly use Zed and it's native agent, which only recently got muuuch better, and on the terminal I use Pi with a minimal selection of extensions (currently pi-kagi-search, pi-smart-fetch, pi-btw and pi-diffloop). I also have Pi in Zed via the ACP, but it does not work so well with some of the extensions, especially the lack of a built-in permission system is a problem, when YOLO-mode is the only mode :)
Honestly, as long as you have a model that is decent at tool calling, your good. Having a solid and stable frame around your model makes a huge difference. The only caveat in all of this is that I spend most of my time on smaller projects and debugging on linux base systems, not huge and complex code bases, so your mileage might vary.
The next phase at work is to set up a chatGPT-like webinterface, and so far LibreChat is at the top of my shortlist. We had OpenWebUI for a while, but it is so bad at using MCP tools that it is practically non-functional for us. LibreChat is a bit more work to set up, but the interface and it's MCP story is much more solid. The goal is to plug in our internal helpdesk, docs and task manager system to LibreChat via MCPs to give us a quick way to query and gather information that is currently very time consuming to do on your own.
I have experimented with both q8 and q4 for KV cache. I can't find any difference between q8 and fp16, but q4 suffers more when the context grows. q8 seems like a good compromise and gives us enough ctx for about 6-8 concurrent, full context sessions. But we have not fully tested those limits yet, as the context windows rarely reach the limit.
In our company of 24 employees, we get by with two DGX Sparks. We don't use AI heavily, but each Spark can serve about 6-8 concurrent requests with a full context lenght of 256k, which is decent. We get about ~35 t/s depending on the model we use (currently Qwen3.5 122B A10B and Qwen3 Coder Next), but we might set up a smaller model too for simpler tasks.
This works for us and will work for years to come. It is not SOTA, but it works darn well for our purposes, and we control the compute and data flowing through it, so totally worth it.
I remember using some kind of software around the time of windows xp i think, that could replace the chrome/shell so you could design your own GUI entirely – but I can't remember what it was called! I spent a lot of time iterating and experimenting back then, replacing iexplore.exe or whatever the main process was called.