alex7o

6 ms·

61.
by alex7o 4mo ago
True but also for a one of piracy exists just use a cracked copy of it and be done with it
62.
by alex7o 4mo ago
We already get TFLOP per watt so you can compute how much flops you are doing while cycling
63.
by alex7o 4mo ago
I have been thinking a lot about this and my simplified understanding is that each model can be seen as a bell curve over human knowledge and each model has a different distribution. Using multiple models would allow us to change the distri
64.
by alex7o 4mo ago
Always happy when I can use a smart model in a sane harness like pi or mastracode. I only wish I was able to run this locally
65.
by alex7o 4mo ago
I mean by some measure a trap can be worse, as many people have died from mines years after no conflict.
66.
by alex7o 4mo ago
I have been using all of ataraxy labs tools for the past few months and that have been indispensable for models to make less mistakes for me. From better git diffs to impact analysis and code reviews.
67.
by alex7o 4mo ago
Oh that is a nice approach whish more benchmarks did cost per successful
68.
by alex7o 4mo ago
I would like to have deeper comparison with alternatives like rtk, which are already fast and written in rust, also the previous comments mentioned something that has been a know problem with rtk that it sometimes strips the thing that the
69.
by alex7o 4mo ago
To be fully hones no I haven't, but for me to switch I need some proof that I would have good experience. Like sell me on having good experience and I will try your product for a week, have a good experience and I will use your product
70.
by alex7o 4mo ago
I live in the EU and I use kagi, but I think you fail to understand why kagi is good and useful. In the end of the day I need a search engine to find stuff and kagi is better than google for the things I and my AI agents are searching for.
71.
by alex7o 4mo ago
He is talking about jai the programing language from Jonathan Blow, which is quite cool but there is no way to access it.
72.
by alex7o 5mo ago
It would be super cool, if we can get this to compile the typescript compiler and see how fast it is against the go version. P.S: does it have PGO for collapsing numbers to ints when possible etc
73.
by alex7o 5mo ago
I think the whole NaCL thing was able to do the ssh directly from the browser without any tunneling, but maybe I am misremembering
74.
by alex7o 5mo ago
I do understand why NaCL and PNaCL are undesirable and why wasm is much better, but as a student the NaCL ssh app had saved my computer science homeworks more than once, and this is something that still doesn't have an alternative alth
75.
by alex7o 5mo ago
Obligatory https://strudel.cc/ mention, same thing bit different, have made music in any of them. But I follow artists that use strudel.
76.
by alex7o 5mo ago
From all the models that do toolcalls the only thing I am confused is why did you pick the worst? Or maybe they are only bad in agentic work it fine for one shot toolcalls?
77.
by alex7o 5mo ago
I mean the openai API is the industry standard for allowing apps to communicate with models, llama-server has it, oMLX has it, ollama has it, vLLM has it, lmstudio as well. I don't think this is such a hard thing to do, but it requires
78.
by alex7o 5mo ago
There is no other way than shipping your own model, because you will want an abstracted API over the inference, and you don't know what the user has installed. Also you can ship 9b fp4 model but it all just depends
79.
by alex7o 5mo ago
How is it mother day isn't that 8 march
80.
by alex7o 5mo ago
I don't know about you but kimi 2.6 from the kimi subscription has been absolutely bad and useless for the past 1 week so I canceled my sub and stopped using it.
81.
by alex7o 6mo ago
Turboquant on 4bit helps a lot as well for keeping context in vram, but int4 is definitely not lossless. But it all depends for some people this is sufficient
82.
by alex7o 6mo ago
Because when you pay for a subscription they don't silently quantize the model a few week after release, and you can no longer get the full model running. Otherwise no need for full fp16, int8 works 99% as well for half the mem, and th
83.
Mastra Code(code.mastra.ai)
2 points by alex7o 6mo ago|0 comments
84.
by alex7o 6mo ago
The thing is I have never used the function keys on my laptop so that was not a problem form me, but also some of the custom functions I hard can just be mapped to fn keys so it is bit like it it us a huge loss
85.
by alex7o 6mo ago
Fight me but I miss the touchbar, it was customizable to be super useful with better touch tool
86.
by alex7o 6mo ago
Maybe a bit misleading. I have used in in two places. One Is for local opencode coding and config of stuff the other is for agent-browser use and for both it did better (opus 4.6) for the thing I was testing atm. The problem with opus at th
87.
by alex7o 6mo ago
Ok I find it funny that people compare models and are like, opus 4.7 is SOTA and is much better etc, but I have used glm 5.1 (I assume this comes form them training on both opus and codex) for things opus couldn't do and have seen it m
88.
by alex7o 6mo ago
Choosing an eccentric stack makes the llms do better even. Like Effect.ts or Elixir
89.
by alex7o 6mo ago
Think of it like that 8b model needs 8gb of vram. Also search online there is a lot of info on that
90.
by alex7o 6mo ago
Also https://github.com/rtk-ai/rtk but some people see that changing how commands output stuff can confuse some models

More ›