I mean if you'd only care about the affiliate revenue, there probably are better niches to serve than citizens looking to protect themselves from tear gas.

You just click "buy" with another provider? They just host the model for money, nothing more.

To be clear: The models are open weights everyone can simply download, because many labs publish them as such. The providers in question are generic hosters. It's the same as if you would get some managed wordpress hosting somewhere.

> You just click "buy" with another provider? They just host the model for money, nothing more.

Start with where the GPU and rest of it comes from.

OK but what if, staying in that analogy, you just have to wait 6-12 months to become as fast as Usain is right now?

In this analogy Usain Bolt will get twice as fast too, right? Using a solution that is always twice weaker than SotA, would put a company (or a national olympic team) at a significant disadvantage to competitors which do use the SotA.

To add a few more details: llama.ccp now both has a web ui out of the box that even supports model switching, and easy model file downloads from huggingface using the cli: '-hf name_of_model:the_quant_you_want'.

.gguf is the native format of llama.cpp and is widely used for quantized models (models with reduced float accuracy to reduce memory requirements).

llama.cpp is the actual engine running the llms, ollama is a wrapper around it.

> llama.cpp is the actual engine running the llms, ollama is a wrapper around it.

How far did they get with their own inference engine? I seem to recall for the launch of Gemma (or some other model), they also launched their own Golang backend (I think), but never heard anything more about it. I'm guessing they'll always use llama.cpp for anything before that, but did they continue iterating on their own backend and how is it today?

"run" as in run locally? There's not much you can do with that little RAM.

If remote models are ok you could have a look at MiniMax M2.1 (minimax.io) or GLM from z.ai or Qwen3 Coder. You should be able to use all of these with your local openai app.

Looking at the data it seems that uk industrial electricity is the most expensive before taxes, is that not correct?

Wang explicitly said that support for Russia in the war is about diverting the US focus from China. If the US shows it doesn't care about Ukraine's future (and Wang told the truth), then I doubt Chinas support will stay as is for long. At most it will be a bargain in the coming EU-China negotiations.

Russia continues to be a significant player in North Korea and Central Asia as well, and this requires Russia-China cooperation. Additionally, if a hot conflit between the US+PH+JP+TW and China happens, China will have to utilize Russian supply and logistics chains to a certain extent (eg. the Chongqing–Xinjiang–Europe railway).

Pronouncing this out loud I wonder whether the name has been chosen on purpose for the marketing effect: "Where's my bitchat"

The comments here seem to show a typical US-EU divide in perspectives, people talk past each other. Yes, mass-market low-margin consumer products (the US view) probably won't be produced much in the EU in the next few years.

But much of the electronics industry in the EU is B2B and centers around producing high-margin products where 10.000 units of a product would be huge.

The company I work for, for example, usually produces a few hundred units of a product before the next revision replaces it. Whether or not the PCB costs 20€ more or not really isn't that important if you only plan to sell 100 devices of it per year for 10k€ each. Aspects like quality and regulatory conformity are way more important here.

That is an excellent point. The development costs of such low-volume projects are often way higher than the production costs. Having production in house or close often allows tighter, thus faster, prototyping and feedback during development, which in turn saves money.

The whole "But how can this be scaled and monetized" crowd here also does not seem to understand the point of such projects and Germany's Hacker community. It is about learning and just doing it, much less about building a high–revenue business.

I think so too, because the bar to economic sustainability is not incredibly high. I know a lot of one-man businesses operating with one Siplace pick and place machine and making a good living out of it.

Many people don't have the desire to expand forever. In my case I hope the company grows o 20 or 30 employees, and then I would work stabilizing it so it can last 50+ years. e.g. setting up a trust to oversee the well being of employees, the quality of the products, etc.

This is completely alien to most american founders and businessmen, in the words of Larry Elison: "it's not enough for me to win, it's about everybody else losing"

Hey, talk speaker here, that's a great point!

A discussion that got cut from the talk at the early draft process was defining what "small-series" and "large-series" mean.

To me, at a human scale and without dystopian monopolies, a small series is anything under 1k, a medium series around 50k and a large series 100-500k.

I wanted to define a special class of series, because to an american a small series is probably more like 100k, and a large one 1 million or more, last year something like 230 million iphones were sold globally and that's an absurd number imo.

Because my vision of a healthy electronics industry is 200 companies each selling competing runs of 1 million units, rather than apple selling all 230 million.

In my ideal world then, the only way for apps to be distributed is a marketplace that is regulated and prevents apple from imposing their 30% tax on every dollar spent on the app store.

How would this marketplace be regulated? Is the idea that there would be an open app format (so that devices from different vendors could run these apps) or a regulated OS?