rfgplk

7 ms·

1.
by rfgplk 3d ago
> What do you all think about Astra now that it's been out for a few weeks? Best model put out so far by any of the frontier labs. Way better than Anthropics models, especially in actual text generation. Claudes fodder heavy text is
2.
by rfgplk 3d ago
It's possible to do this even with existing frontier models. You just have to angle your prompts such that it doesn't invoke "pentesting" anywhere in it's chain of thought. This also allows you to get the models to
3.
by rfgplk 3d ago
My general piece of advice to anyone looking to do this is not to even bother. Just strip the hardware package and replace it with an arduino/bb or an equivalent sbc figure out what the motors are and wire it up yourself. Way faster an
4.
by rfgplk 3d ago
You can use an LLM but you have the provide the spec of the motor and the surrounding hardware apparatus. Not every motor is the same.
5.
by rfgplk 3d ago
Pretty much anything hardware based is like this. Everything is 100% closed source and bolted down such that inspecting it is outright impossible without extensive reverse engineering.
6.
by rfgplk 6d ago
Any company still doing them is wasting their time. I can show up hungover and half asleep with an Astra/Fable subscription and ace any leetcode problem with zero difficulty.
7.
by rfgplk 6d ago
Why would the subscription matter if you're not even getting interviews? The only relevant metric is your personal network, everything else is done.
8.
by rfgplk 6d ago
This would cease to be a problem if OpenAI remained true to their founding motto and... actually open sourced their training/inference pipeline.
9.
by rfgplk 6d ago
Under EU rules it doesn't constitute consent.
10.
by rfgplk 6d ago
Under current understanding of the law, anything produced purely by LLMs (with no substantive human input, which is what OpenAI claimed in their post) is firmly in the public domain. So OpenAI can "claim" anything they want, it do
11.
Micron: A pure C++ (re)implementation of Libc(github.com)
1 points by rfgplk 8d ago|1 comments
12.
by rfgplk 8d ago
Greetings, from the last time I posted this micron has received large extensions and is now (more or less) feature complete. In addition, I've also ported it to baremetal, so you can use it (more or less drop in) in embedded environmen
13.
by rfgplk 8d ago
We are very likely at the begging of the next industrial revolution.
14.
by rfgplk 8d ago
> Why can't they rule it out? Is even OpenAI unable to track the provenance of all of their training data? Probably? I have a few hundred TB of training data for various small scale models and I can attest that I have _no idea_ what
15.
by rfgplk 8d ago
> And what's a better way of empowering people than robbing them. Better than the walled gardens of most journals where you can't even read half the papers without shelling over thousands of $$$
16.
by rfgplk 8d ago
Skill issue. Also lean is meant to be executed, not read.
17.
by rfgplk 8d ago
> This is utterly shocking. Even the AI optimists did not expect this to happen in 2026. Wow. Wrong.
18.
by rfgplk 8d ago
This isn't really true.
19.
by rfgplk 8d ago
Something I've been going on and on about for months now and no one seems to listen. LLMs today are allowing _anyone_ to access cross-discipline knowledge that was previously entirely inaccessible without a) extremely deep pockets or b
20.
by rfgplk 8d ago
I don't see how companies like this can survive going forwards. DaVinci Resolve is effectively ~200 Astra/Fable prompts, maybe even fewer if you're willing to forgo polish.
21.
by rfgplk 8d ago
This is something I've been pondering about. The article is well written but doesn't mention that even mere specs of dust can have catastrophic consequences for spacecraft (literally a spec of a few milligrams). A logical solution
22.
by rfgplk 8d ago
What languages do you work with? The two models perform differently on different tech stacks. Generally as of today, GPT-6 is downright the best overall model at raw code generation / cross discipline understanding / bugfixing. Th
23.
by rfgplk 9d ago
> People should realize by now that they aren't giving out thousands of dollars worth of compute in a $20 subscription out of the goodness of their hearts. The base level plans aren't meant for any kind of serious work beyond t
24.
by rfgplk 9d ago
Unique language! To me the definition of low-level vs high-level strictly comes from the indirection the language runtime provides for you. If the language compiles down to asm, it's low level. It literally does _not_ matter what it lo
25.
by rfgplk 9d ago
> C doesn't expose modern CPU features like speculative execution, therefore C is not low-level This is a 100% skill issue of the author, as is always the case. C does expose it fully, except it's implicitly implied by your cod
26.
Heterodox C++(meridional.xyz)
1 points by rfgplk 9d ago|0 comments
27.
by rfgplk 9d ago
> It also means that the "hey Claude vibeslop me a CUDA Anubis solver" route is on its way to being fundamentally dead. If they "vibeslopped" WebAssembly in Anubis they would have been done a year ago.
28.
by rfgplk 10d ago
You're thinking of pre-LLM times. Nowadays GPT-6 can probably advise them better than all the worlds law firms combined. And probably win too, considering the incompetence (and tech illiteracy) of human lawyers.
29.
by rfgplk 11d ago
~99% of the cost is due to red-tape regulations. It's so obscene that they require certified backup genererators for the backup generators and 3-4 layers of fail safes. Then you need to wait YEARS to get all the permits and licenses. O
30.
by rfgplk 12d ago
> Could I hook up a SOTA model the 2D computer puzzle game Gruntz (1999) so that it can read it from screenshots and act on it through keyboard and mouse inputs in a way where it would learn how to play and progress through the game? I d

More ›