stevefan1999

I used to hear that when you finished eating ramen and you should say this as a replacement speech of "thank you". It basically means "It must be tired of you to go around to design and prepare for the meal" which implicitly implied an appreciation

Another fun fact: itadakimasu is categorized as a keigo, or more specifically, kenjougo, an honorific expression that depreciates speaker itself. And since itadaku is an advanced version of morau to show that "I am of lower status", so the whole expression should be (from the gods/kamisama/the people above me who brings me here for the feast) I humbly receives.

This is also why you say itadakimasu even when alone, despite there are apparently no one superior than you exists, but no, it is now the gods you appreciate

Hmm, I've got Rybelsus, it did something. I can reduce my daily meal from 3 to 1 lately, and even on 2 meals I can eat half of it compared to what I typically eats ordinarily, to get full appetite.

Still, with all the things I've tried I still gained weight. And I have been noticably getting harder to sleep at night because the hungry feelings comes back. Maybe it is the oral Rybelsus waning off idk

But the problem is not that your model is fast.

Sure, you can go ASIC and go even faster, but the thing around GPU is that they scale well for both training and inference, and the technical floor is low.

The level to get into FPGA design is insanely high, you've got to read timing diagrams, you need to know combinatorial and sequential logics and good sense of boolean algebra, you need to have an asynchronous signal based mindset which is vastly different from CPU/GPU, you need to know netlist and you need to endure the time it takes for the EDA to finish generating it. Yosys is still years behind Xilinx

There is a reason GPUs are called accelerators; it sacrifices and does not try to really specialize on one particular thing, except high parallel dataflow and branch-free calculation. Otherwise we will all be using DSPs

> Part of this is probably unconscious, but another part is probably explicit assumptions about the person's self-discipline, e.g. the fit guy must get up at 6am every morning and go for a run, and the obese guy probably sleeps in, eats trash, etc.

More like stereotypes.

For some reason I took both GLP-1 and Ritalin everyday, having moderate exercises despite throwups, only see myself gaining weight rather than losing. I have no idea why both drugs that claims to lose weight only to do the exact opposite, and perhaps I'm the kind of people who can only get fat and not losing it despite some joint efforts.

Perhaps it is not enough and I have to go full aerobatic and ketone mode. Money money money.

So uh, this is just trying to promote fat shaming and lookism using science as a judgement tool. How cool is that? /s

Containers do solve dependency hell by isolation. It's like venv/uv in Python that you switch between different environment with different sets of dependency. Or speaking essentially, you lifted yourself away from having dependencies shenanigans since you have virtualizations of it

but the dependency hell for human is inherently political

You technically can sideload. For 7 days or on TestFlight

> On Android though, you can soon see such app stolen and on some of the apk warez sites. It just breaks the model.

Think about Windows and software privacy

It only happens iff that market is full of intelligent crowds that can truly reflect their "voting" and exchange of their idea.

No, it's more about maximizing the utility function and finding the "shield" where you'd start getting diminishing return beyond that

I wonder why LLM love this word so much. Same as mint, seam, tier.

> This way one person can implement a whole new hobby project instead of only writing a single specialized part in an existing project.

I still do start from bottom up to refactor everything into workspace members and this way, parallelism is way easier

> LLMs reward seeing forest instead of trees.

Not just forest. LLMs reward seeing structures and designs, and it still requires a lot of creativity to design it well. I have to do a lot of architect works for that, like package layout, testing strategies (BDD vs TDD vs E2E), setting up access to platforms and digging out what strategies to apply...etc.

> People focused on the trees will however lag behind as implementation details matter less and less.

Trees like demonstrations and example code are even more important. This is because those are literally post-training materials for the LLMs

> Specialists will be fed to the model weights

I don't think so. New labor laws are certainly going to be introduced by the politicans to tackle this.

> generalists will inherit the world.

Now the problem is how far do you want to stretch the systemic thinking, and you also get burned out quicker.

Well to me using an LLM is akin to using microwave for food, and programming is getting your hands dirty in the art of cooking. I used to do the latter but the pace is too much, so I'm forced to go for the former. I still miss the days where I can cook myself

I wonder if LLMs are useful in reconstructing a real compilable game that reverse engineers all functions.

No, not in the sense that I want to shove "AI" into everything vibe coding, but that I seems to me that those PDBs are insanely valuable artifacts for data sourcing and doing posttrainings for.

We have all those variable names, and perhaps a little bit of optimization change such as the control flow, and the hard inverse problem of figuring out if likely/unlikely's effect on branch predictor, that means guessing the likely/unlikely without knowing real code on the code generated by compiler and the general feature of the CPU of that era, and sometimes even guessing the compiler version right too, since new versions of compiler may have better optimization down the line.

This I would say is something that even the best compiler wizard can't really do, but LLM is ruthless and reluctant on trying, until they find one.

Of course, things like vtable and structure layouts/paddings is almost impossible to truly reverse. Not even human, even LLM would struggle.

> much simpler type checkers like Rust's

Linear/Affine types aren't easy, although deep down it is about enforcing XOR.

Yeah, nommu Linux basically cannot run normal ELF since there is no virtual memory which is needed for relative addressing and relocations. You are mostly left with classical formats like AT&T a.out only

Why does it looks like...Shamir's secret sharing scheme to me?

> Anthropic has never advocated for a ban on open-weights models.

Shut the fuck up

5. A gazillion amount of telemetry data and social media data to sell on the data broker market. That alone is enough to attract another billion dollar of potential investment consistently, especially for training data and intelligence...you know, selling to the Mossad or CIA.

In a world of money, every bit you can earn, you have to earn, while I know it is immoral to sell the data for spying purposes, morality alone don't mean shit now, that I never see any capitalists and corpo dogs have any kind of morality anyway, except when massive backslash, congress hearings and boycotts, which is, ahem, Meta/Facebook and Cambridge Analytica. But given the unique position Musk used to have in Trump's government and his right leaning position he is safe to do so. At least for now, he is immune to holding congress hearing for a scandal I suppose.

And not to mention that Grok is literally trained over X/Twitter, and keep in mind Bluesky is federated and strangely, no one seems to be scraping it over ActivityPub. I believe Thread is also built on the same foundation but it's sure that Meta made good measures for antiscraping, I expected as a data oriented company.

Alas, keep in mind SpaceX merged with xAI and nonetheless don't forget their data side

Yeah, they can't guarantee a SLA and cost reimbursement policy, such a shame

Not a surprise for them to be a dick despite they seems to have considered security through obscurity is better than security through openness

Where's the reset...

> early mover monopoly

Well, while I do agree it is an early mover (I believe Steam is not the first to do online game platform, but it should indeed be the first one to use a CDN for the solution), I'm sorry that you have Ubisoft Connect, EA Origin, Epic Games Store, itch.io, GOG, Humble Bundle.

So you do have a choice. That renders the statement of it being a monopoly impossible. It's simply because every other platforms sucks.

In other words, it's like Nvidia. You choose Nvidia because AMD and Intel sucks, not because you don't have a choice.

> engineering quality is mediocre

You should see how CS2's smoke system works: https://www.youtube.com/watch?v=ryB8hT5TMSg

tl;dw: it is a voxel ray marching plus flood fill with Mie scattering. I'd like to point this out that is not only a really good game design, but it also requires ingenious engineering quality. I tried to improve upon it by adding octrees (or in other word, making it sparse voxel octree) but it doesn't seems to add much.

Oh and plus their new bomb shock wave system. I tried to reverse engineer it with vector field, but I couldn't come up with the ODE equations to recreate it. I believe it has to be some kind of field, because in a high energy explosion, it scatter through air into fluid-like structure and hence we can use some aerodynamics for simluation

> PYTORCH_ROCM_ARCH="gfx900;gfx906;gfx908;gfx90a;gfx942;gfx950;\ gfx1030;gfx1100;gfx1101;gfx1102;gfx1103;\ gfx1150;gfx1151;gfx1200;gfx1201"

That kicked off my PTSD to setup ROCm on my 6600XT and 9070XT, not even with different containers. I have to build an image for EACH architecture. That's duping 15GB for each arch on my already terribly small SSD.

That only works if you have all the unit tests and making sure the behaviors between different GPUs are expected, and can tolerate difference in implementation details such as numerical stability and complexity for some functions. AMD and Nvidia are doing vastly different computation for different implementation to the same algorithm for their APIs (can't tell due to NDA, but one algorithm Nvidia did it using broadcast, while AMD is using tile), and that's one of the reason I gave up. And perhaps only for light kernels doing basic BLAS operations and GEMMs. With a standard to follow, that's one of places where it indeed is giving it a good chance, for example on Tensorflow with ROCm or PyTorch. You basically have an insurance of the behavior and a translation layer for ROCm and CUDA by Python.

But anything other than that, nah.

Edit: Maybe I remember what algorithm that is. It is a trick to do coding theory stuff by using NTT (Number Theoretic Transform, the friend of DFT, the Discrete Fourier Transform, but on a ring and hence applicable to finite field) on GPU to do computation over a GF(2^8) field, for my former employment but I went back to using GFNI. Still I have a NDA to adhere to and can't really explain what happens.

I rather replace bun with deno, given that now both bun and deno are written in Rust, and Deno have even more node compatibility than bun. Deno and Node both runs on V8, while Bun runs on JSC, and Deno has a killer feature that you don't have to even run npm install to use package, you can just top-level import or await import an URL (given that you allowed it with a command prompt or bypass it competely with -A but discouraged).

I've used Deno, CucumberJS and Playwright to write E2E test suites. Zero npm install and not even deno.json or package.json

but you have to rewrite all your software to ROCm. And ROCm, to this day, still sucks.

ZLUDA basically crash on high memory demand, and only accounted for ~70% of CUDA API coverage (and it is still buggy).

But hey, at least it does run on Rust-CUDA. I'm one of the few who ported it and fixed a few bugs on ZLUDA. I used it to run a simple SHA256 kernel and it ran sure, but I gave it up because of those fundamental problems on AMD GPUs. You can't believe how messy ROCm is. I wonder if Vulkan compute kernel using SPIR-V would be a better choice.

Perhaps we should learn from the ADHDs on how to resist the dopamine hacks.

Oh wait, I am one and I can't resist it since day 1 on this world.