The idea is to create cryogenic electronics that operate near absolute zero with minimal power. https://www.techbriefs.com/component/content/article/54866-an-ultra-low-power-technology-expands-quantum-and-space-potential

There are also ideas for putting quantum computers on dirigibles in the stratosphere https://discovery.kaust.edu.sa/en/article/26198/green-quantum-computing-takes-to-the-skies/

And China has already been doing a lot of experiments with quantum communication using satellites https://en.wikipedia.org/wiki/Quantum_Experiments_at_Space_Scale

It turns out that there might be an elegant way around bloated context windows and slow inference speeds. Instead of stuffing thousands of text tokens into a prompt like standard retrieval setups do, Memorilla approach encodes an entire document collection into just 16 query dependent latent tokens. The system drops these soft memory tokens directly into the input embedding space of a frozen LLM. Doing so averages 394 times fewer document tokens than a full context baseline and 113 times fewer than a standard fifty document retrieval method. Aside from saving massive amounts of compute overhead, the system also beats full context inference on complex multi document reasoning tasks. It hit 84.33 percent accuracy on the FactKG dataset compared to 80.43 percent for full context while using 108 times fewer tokens to represent the source documents.

This approach also has implications for continuous learning and agentic workflows. Having persistent memory prevents the agent from getting stuck in stupid navigation loops between previously explored items. Because the initial document encoding is a one time shared cost, the decoder only processes a small number of memory tokens. You get continuous adaptation to new environments without overwriting prior knowledge or blowing up your active context limit. It’s a really clever way to decouple persistent knowledge storage from the active context window.

One obvious advantage is that the ambient environment is already cold. Also, there are reasons to want ultra-cold materials in space. For example, the substrate used to do quantum computing needs to stay ultra-cold, so putting quantum computers in space makes sense. https://www.uni.lu/snt-en/news/how-quantum-computing-is-transforming-space-research/

Oh no, I don’t have the VRAM to run them locally. I can run Qwen 27b, but it’s not quite there yet for doing bigger projects. I find hexbear is very reactionary in terms of AI.

I mostly use DeepSeek and Kimi. I find DeepSeek is good enough to do most coding tasks, and Kimi does a good job with planning. And I suspect most people shitting on the ArtCraft apps haven’t actually used them. I’ve been using them for around two weeks now, and if anything, I’m impressed how well they work for being such new apps.

PhotoCraft is better for me than GIMP and even PhotoGIMP because it’s much snappier, the UI/UX is far better because it just apes Photoshop, and the tasks I typically do all work fine. If you’re a professional doing complex things with layers and effects, it’s obviously not a replacement, but for casual use it’s great. And I can only assume it will continue getting better and features will be more fleshed out going forward.

sadly

we’re gonna see a lot more medical tourism going forward I expect

This is from China where they already have a healthcare system that works for regular people.

turns out that neural networks may in fact be a lot cleaner than people assumed https://already-clean.goodfire.com/

Fire Emblem series is one of my fav turn based tactics games. Other ones I’ve really enjoyed are Advance Wars/Wargroove, XCOM, Shadowrun, Wasteland series.

Oh I built my own from scratch https://github.com/yogthos/newsroom

I haven’t looked too hard to be honest I find I prefer a laptop for any sort of editing.

I use PhotoCraft and LightCraft for my photo processing now which replaced two commercial apps for me. I tried using GIMP and DarkTable before, and found the UX was just too clunky for me. Neither app fully replicates the functionality, but they do enough of what I use day to day, and they feel snappy.

Personally, what I’ve found to be more interesting though are the custom tools I can now easily build for myself. For example, I was able to make my own news aggregator that goes through a bunch of different news sources using rss, scraping, reading youtube transcripts. It doesn’t care what language the source is in, so I can get news from many different countries. Then it aggregates them to tell me what news are most widely reported, how stories evolve over time, find connections between them and give me a big picture overview. It took around a week to build, and it goes through hundreds of news sites each day to give me a briefing. Being able to make apps like this quickly that are customized to your exact workflow is something that just wasn’t easily doable before.

I disagree, biological neurons evolved under the exact same selection pressures the article explains. Their complexity doesn’t change their fundamental nature and the problems the brain solves.

Maybe if it was a publicly owned utility then sure. I wouldn’t even trust a non profit here.

not my fault your instance can’t federate properly

I love that you’re so mad now that you’re actually stalking me 🤣

Wow, still going with that are you?

Whatever helps you sleep at night I guess.