lopotubez

Oh my god...I just realized SpaceX = Space Sex?

My biggest regret in life was that I was introduced to vim before emacs and got really comfortable with it.

Did not realize that everything I was building in my vim configs to make it better was already default in emacs i.e dired, compile mode, magit, org mode, etc.

If I wasn't able to multiplex with tmux I would have switched to emacs a long time ago.

It baffles my mind that trains never had the technology for sensors so they could atleast slow down before incoming collisions.

Stuff like this just further enhances my fear of collective travelling.

Nice but I prefer the much powerful version of 3 points being collinear precisely when ((p-q)^2 + (q-r)^2 - (p-r)^2)^2 = 4(p-q)^2(q-r)^2 which works for any field(!) with non-characteristics 2.

If they are non-collinear/degenerate then the difference becomes 16 * area^2 of the triangle PQR, in which case one can easily derive Heron's formula.

So the solutions to the n'th roots of unity of the form z^n = 1 are n-gons on the unit circle. This means for an arbitrary w in the complex field z^n = w gives solutions on a circle with radius n'th root of modulus w and are 2*pi/n apart so it is also an n-gon.

Pretty fun stuff to plot around using Geogebra to give a geometric intuition how the complex numbers behave.

Yes, this is all good knowledge in theory for testing local ai. There are some more technical stuffs to consider that are happening in the background that is well outside my expertise to explain. I have been testing every MoE model I can fit on my system every year to find out which is the best for me.

Last year, it would have been impossible to run a 35B A3B model on old Pascual gpu and make it somehow usable. But for some reason now it just works even with a better finetuned version. So clearly there is more to it that just having it theoretically fit the system. Look for example how Strata works on certain models and systems compared to its llamacpp counterpart.

Models and inference engines are getting better and more optimized everyday so it is definitely worth experimenting with your GPU. You should probably check on newer finetuned and or REAP models Huggingface that could fit on your system.

I believe in your case the model quality matters a lot. To my (trust me bro) experience any MOE model, finetuned or not, with params < 19B even with Q4-quant always give lackluster results. So any Qwen2.5 wouldn't suffice to even simple coding tasks. For your case maybe you can bump it up to a Gemma4 E2B/E4B or Qwen3.5-9B. Also make sure that your current build of llamacpp actually uses your GPU with CUDA.

I feel like these types of benchmarks mostly test what the mainstream tells you is possible and or works. You rarely get to hear the different implementations of inference engines, different types of setups (e.g. hybrid ones or older hardware) and their pros and cons. Also it seems these were tested with very minimal token usage.

Currently on my 9 year old laptop with a GTX 1070 Max-Q 8GB vram, 16 gb ddr4 ram and i7 8750h. I am running a finetuned Qwen 3.5-35B-A3B (Tiel-Coder-35B-A3B-GGUF-MTP) with 4-bit quant on a llamacpp fork (Turboquant) with 131k context at around 16-19 tk/s and 350-500 tk/s prefil using Pi harness. It is amazing for writing scripts and doing medium coding tasks and summarizing text for me. After around 100k context it goes down to 9 tk/s but that is to be expected.

If you ever ask any frontier LLM or these types of benchmarks, they would probably say that the best model I can run on my system is a crappy 9B model or something, which is just completely false. So I really wonder how much more you can pull out from more powerful systems and much better open weight models.

Idk maybe depends on what you study or what your college campus offers. But for my case studying EE, getting access to expensive lab equipments and also being able to network with companies looking to hire juniors makes it quite valuable to me.