antirez

46 ms·

751.
by antirez 3y ago
Appreciated.
752.
by antirez 3y ago
I agree. Incredible development environment for a huge number of MCUs. Super fast.
753.
by antirez 3y ago
Shooting at the moon should be about the kind of work you do. Quality vs size. 50 focuses persons can do incredible things.
754.
by antirez 3y ago
This is a joke, yet this is basically a very rudimental form of malloc() auditing tool. Because at the end of the program, it could show you all the stuff that is yet allocated, and even scan the program whole memory for the pointers and re
755.
by antirez 3y ago
Am I the only one that is shocked by the fact that a 1978 computer, even if a supercomputer (but still using the technology of the time) was 1/4 the speed of a Raspberry? The Pi, if you look at the big picture of computing, is a very f
756.
by antirez 3y ago
I believe that some hype / over-enthusiasm is acceptable, but I'm a bit more worried when a given community uses it as some kind of organized propaganda, arriving to excesses like saying that writing code in non-safe languages sho
757.
by antirez 3y ago
Rust programs are safer than C programs but almost universally more complicated to write compared to C with equivalent high level libraries for data structures and other operations. So I can't see how your argument applies here.
758.
by antirez 3y ago
Are we at this level now? The scheduler performance has nothing to do with the used language. Actually making use of the freedom C provides (while being unsafe) you have the ability to implement whatever complex algorithm you can envision i
759.
by antirez 3y ago
That's it indeed. Thanks.
760.
by antirez 3y ago
That makes sense, but if that's the case I would like to see something a bit more vague polish. For instance: 1. Ability to forward messages to app. 2. Ability to run the same questions to the different LLMs installed. 3. Some updated
761.
by antirez 3y ago
I see the efforts required to create the little app, but inference via llama.cpp or core Ml is trivial and the models are open weights, so it makes more sense to have a free app for this: most of the value is in the LLM which is free.
762.
by antirez 3y ago
Yes, but this is working. I literally didn't look at the code... but at the result. In other fields GPT can't deliver working code at all.
763.
by antirez 3y ago
Indeed, you are right. If you have an M[1,2,3] MacBook with enough RAM, you may want to run some model like DeepSeek-coder 34B in local. Or the smaller one, but it is going to be weaker.
764.
by antirez 3y ago
In general Copilot is much weaker than bigger/slower models, so if you have the feeling you are not using AI enough, the first thing to try IMHO is to chat with powerful models like GPT4 to see what is the current state of art in code
765.
by antirez 3y ago
@dang something is wrong with the ranking of this post.
766.
by antirez 3y ago
There is a misunderstanding going here. A linked list is a pure form of knowledge. What we see today is an explosion of arbitrary complexity that is the fruit, mostly, of bad design. If I learn the internals of React, I'm not really
767.
by antirez 3y ago
> is still a problem in philosophy Exactly! This is why I removed this fundamental questions from my post: in this moment they don't have any clear reply and will basically make an already complex landscape even more complex. I beli
768.
by antirez 3y ago
If the LLM is better than me at writing Torch code, it is a great idea for me to use an LLM to write my model definition, since the exact syntax or the reshaping of the tensors is not so important to me. If I want to create a convnet and tr
769.
by antirez 3y ago
Programming is special because 99% of times you can tell immediately if something works or not, so the risk of misinformation is very narrow.
770.
by antirez 3y ago
Maybe I can help you: https://gist.github.com/antirez/931cb4f814e2c852a8289b6fa876...
771.
by antirez 3y ago
This is my cut & paste failure (I didn't re-check the GPT-4 output that fixed the grammar). Fixing...
772.
by antirez 3y ago
Hey! With the maximum spreading, and the stock (very poor) omni antennas in the Lilygo devices, you get this: ~ 500 - 1km: urban landscape, tons of buildings in the middle. Like opposite sides of a very crowded block. ~ 5 km: open landscape
773.
by antirez 3y ago
Related: https://github.com/antirez/freakwan/ Disclaimer: I'm the author. The concept is similar to Meshtastic, but the goal was to make more documented and clear choices at protocol level, to have a much sim
774.
by antirez 3y ago
Awesome story! Yes, indeed PrusaSlicer is based on Slic3r :)
775.
by antirez 3y ago
I happen to be in a special situation here. One of my softwares, Redis, has a very permissive license that was used by many vendors to sell it. At the same time, I'm a lot into 3D printing. So I guess this gives me some perspective on
776.
by antirez 3y ago
Can't visualize the graphs well from mobile but the many layers are there because MoE models select a different expert at each layer, not just at each token.
777.
by antirez 3y ago
Odd that this is presented as a news. Milan tube has 3G and later 4G coverage for many many years now.
778.
by antirez 3y ago
I played a lot with this LLMs recently, and here they do not seem to be the vanilla versions. There in some prompt attached before or they were fine tuned to change the reply. In general they seem "broken" compared to the normal m
779.
by antirez 3y ago
I wonder if Cryptography Engineering, while updated, is a good book as Applied Cryptography, that really was a pretty inspired one.
780.
by antirez 3y ago
Are you sure PrusaLink has it? I've an MK4 with latest firmware and I can only see the usual temperature graph and file browser, and that's it. Not that the bed visualizer would be so useful with the MK4: this printer (like the XL

More ›