jmalicki

27 ms·

601.
by jmalicki 8mo ago
You care when you're trying to generate random vectors which may be of a different size, and if you are biasing your sample. Is it enough to truly matter ? Maybe not, but does it also matter if 80 bit SHA1 only has 61 bits?
602.
by jmalicki 8mo ago
I don't think it probably helps with a lot of the super easy stuff like creating a pool with a line of configuration - fair! I (personally) would rather spend the fixed several hours of doing a few things like that manually, vs. poundi
603.
by jmalicki 8mo ago
I love Diesel for ORMs - it's very much in the same spirit as a type safe sqlalchemy, so it depends if you like that. The type safety is a great feature, it always saves me from writing incorrect queries (a huge one is nullability is
604.
by jmalicki 8mo ago
What about a Porsche vs. a Toyota Camry for half the price?
605.
by jmalicki 8mo ago
For a web backend? Rust is pretty mature there, it doesn't even feel like an innovation token - it's by my favorite thing to use Rust for. You have very mature webservers, asyncio, ORMs, auth, etc., it's very easy to write,
606.
by jmalicki 8mo ago
I am aware of tests vs. ChaCha20 here https://www.pcg-random.org/index.html , I am not aware of tests vs. AES-256-CTR. However at some point, 100x faster performance w/o an exploitable attack vector is also relevant! (t
607.
by jmalicki 8mo ago
There has never been a perfect CSPRNG.
608.
by jmalicki 8mo ago
"before the computer in question broke down." A good MCMC simulation might test that! E.g. say, training a large diffusion model. It takes way more computing power than the average time for a single computer to fail. Also, the s
609.
by jmalicki 8mo ago
It all comes down to actual specific statistical tests, and how hard they are to break in specific applications. No CSPRNG is absolutely perfect, no CSPRNG has ever absolutely passed every statistical test thrown at it. In MCMC, it stresses
610.
by jmalicki 8mo ago
MCMC can be difficult for this reason. There are concepts like "k-dimensional equidistribution" etc. etc... where in some ways the requirements of a PRNG are far, far, higher than a cryptographically sound PRNG, but also in man
611.
by jmalicki 8mo ago
The number of possible assignments has to be effectively close to an integer multiple of the number of shuffles. It doesn't matter how many universes it would take to generate all of them, there are some assignments that are less likel
612.
by jmalicki 8mo ago
I've experienced that AI relies even more on the abstractions humans need than humans do. People talk about stripping away the ORM, or just writing object code, but the limited context windows in my experience make AI require even mor
613.
by jmalicki 8mo ago
A ton of GPU workloads require leaving large amounts of RAM resident on the GPU and running computation with some new data from the CPU.
614.
by jmalicki 8mo ago
Oh you made vibe coding work? Well then it's not vibe coding. But any time someone mentions using AI without proof of success? Vibe coding sucks.
615.
by jmalicki 9mo ago
"Grad Student did this". Co-authored by <Famous advisor 1>, <Famous advisor 2>, <Famous advisor 3>. Is this so different?
616.
by jmalicki 9mo ago
The nice thing about modern LLMs is that it's a relatively large static use case. The compute is large and expensive enough you can afford to just write custom kernels, to a degree. It's not like CUDA where running on 1, 2, 8 GP
617.
by jmalicki 9mo ago
I have it do svgs. The layout isn't always great on first shot, but you iterate on that. They can also natively generate e.g. github markdown mermaid diagrams (github markdown has a lot of extensions like that)
618.
by jmalicki 9mo ago
HPC is also not a normal data center but also usually doesn't have the scale of hyperscaler AI data centers either.
619.
by jmalicki 9mo ago
It's a whole wafer. Basically all chips are made on wafers that big, but normally it's a lot of different chips, you cut the wafer into small chips and throw the bad ones away. Cerebras has other ways of marking the defects so th
620.
by jmalicki 9mo ago
The title is wrong. The title of the paper is "Silicon Formalism: Rules, Standards, and Judge AI" When they say legally correct they are clear that they mean in a surface formal reading of the law. They are using it to characteri
621.
by jmalicki 9mo ago
A game engine is sort of just a UI toolkit for interacting with 3d objects. If you want a 3d model you can interact with, you call that a game engine. Should they invent something new because they dislike the word "game"?
622.
by jmalicki 9mo ago
At the time YouTube was acquired their infrastructure costs were quite high. Not as crazy as today's AI companies, but in the same way a lot of people were questioning if they could ever make money because of it.
623.
by jmalicki 9mo ago
> Whilst true, this is a kind of "you're holding it wrong" argument. If LLMs had model of what differentiates good code from bad code, whatever they pull into their context should make no difference. Good code is in the ey
624.
by jmalicki 9mo ago
They ran out of passively collected data. RLHF allows them to gather deeper more targeted data.
625.
by jmalicki 9mo ago
There is a lot of RLHF effort around this.
626.
by jmalicki 9mo ago
> A programmer might write a function, notice it becoming too long or doing too much, and then decide break it down into smaller subroutines. I've never seen an LLM really do this, they seem biased towards being additive. The nice t
627.
by jmalicki 9mo ago
SkyNet. When it can do the hard stuff, why do you think we'll still be around for project management and prompting? At that point, we are livestock.
628.
by jmalicki 9mo ago
What do you think the entire issue was with supply chain attacks of skills moltbook was installing? Those skills were downloading rootkits to steal crypto.
629.
by jmalicki 9mo ago
I find that AI allows me to get into algorithm design more, and the intersection of math and programming more, by avoiding boilerplate. You can indulge even more by letting AI take care of the easy stuff so you can focus on the hard stuff.
630.
by jmalicki 9mo ago
In cursor you highlight and hit Ctrl-L, and use the voice prompting - I can do this today!

More ›