Scene_Cast2

I had the same theory, but IIRC the H2 isn't much better with radio on.

They remind me more of the Korean TV series Cashero. There, the main hero has strength superpowers, but the power comes by crunching through cash in his pocket.

But I use the per-token APIs for my usage, not a subscription, so I'm guessing I'm in the minority.

I also use them per-token (and strongly prefer that due to a lack of lock-in).

However, from a game theory perspective, when there's a subscription, the model makers are incentivized to maximize problem solving in the minimum amount of tokens. With per-token pricing, the incentive is to maximize problem solving while increasing token usage.

Same here. Using it on two boxes, makes Linux sysadmin work easier.

Some more love for the updates page. E.g. select a subset of updates to install, be more clear that the last update time could be different if you installed updates via CLI, that kind of thing.

There's also Moth-class sailing or windfoiling for mere mortals.

Are there any good static (i.e. not runtime) type checkers for arrays and tensors? E.g. "16x64x256 fp16" in numpy, pytorch, jax, cupy, or whatever framework. Would be pretty useful for ML work.

Of typical homelabs that are posted and discussed.

The online activity of the homelab community leans towards those who treat it as an enjoyable hobby as opposed to a pragmatic solution.

I'm on the other side of the spectrum. Devops is (at best) a neutral activity; I personally do it because I strongly dislike companies being able to do a rug-pull. I don't think you'll see setups like mine too often, as there isn't anything to brag about or to show off.

It still seems to have the same pitfalls as all the other image generation models. I ran it through my test prompt (wary of posting it here, lest it gets trained on) - it still cannot generate something along the lines of "object A, but with feature X from Y", where that combo has never been seen in the training data. I wonder how the "astronaut riding unicorn on the moon" was solved...

EDIT: after significant prompting, it actually solved it. I think it's the first one to do so in my testing.

It's a shame that it's not on OpenRouter. I hate platform lock-in, but the top-tier "deep think" models have been increasingly requiring the use of their own platform.

Sweet! I love clever information theory things like that.

It goes the other way too. Given that LLMs are just lossless compression machines, I do sometimes wonder how much better they are at compressing plain text compared to zstd or similar. Should be easy to calculate...

EDIT: lossless when they're used as the probability estimator and paired with something like an arithmetic coder.

I just ran this with Gemini 3 Pro, Opus 4.6, and Grok 4 (the models I personally find the smartest for my work). All three answered correctly.

WizTree uses a similar idea - load the file system indices and works almost instantly.

Not the OP, but the new LLMs together with harnesses (OpenCode in my case) can handle larger scopes of work - so the workflow moves away from pair programming (single-file changes, small scope diffs) to full-feature PR reviewing.

Just to nitpick - compilers (and, to some extent, processors) weren't deterministic a few decades ago. Getting them to be deterministic has been a monumental effort - see build reproducibility.

Try looking into videography COBs. I would recommend something like the Zhiyun Molus G300, SmallRig RC 220B or similar.

The absolute cheapest lumens per dollar COB would be the GVM SD300D, although I highly question the reliability and light quality.

Oh hey, I have one of these! I really like it. It's quite a unique design. People (especially where it snows and gets gloomy) should have more lumens, and I'd recommend this lamp for others.

One downside is that the active fan cooling design is questionable - the air goes over the top of the LEDs, and there aren't any dedicated exit holes so the air is just squeezing through the very small gap between the glass and the heatsink. There are also blotches of paint that worsen the TIM contact between the PCB and the heatsink. I used a rotary tool to remove those blotches.

I bought a vintage (pre-WW1) circular slide rule. But it's too much of an "artifact" to use - I'm afraid I'd damage it.

Then why not geometric / Clifford algebra?

As someone who's worked with population data, I found that there is an enormous rift between reported opinion (and HN and reddit opinion) vs revealed (through experimentation) population preferences.

You're typically issued a corporate phone, it's the only phone that can open work email. You have a choice between an Android (something like a Pixel and a Samsung) and an iPhone, with some companies incentivising Androids with things like a faster upgrade cycle or more premium trims. The culture is split between having just the one free corporate phone and having two phones - one personal, one corporate.

I discovered it a few weeks ago. It solved my problem of having a hundred arxiv tabs open. Highly recommended. I'm also looking forward to trying out the new annotations feature.

It also percolates into reviews, too. When a nontrivial fraction of the community is buying dreams and is about collecting as opposed to using whatever it is, some reviewers style their content towards that crowd and overlook issues or benefits that pop up when actually using the gear.

I don't have a problem with collecting, but I'd love for the distinction to be more upfront.

I used to use Doxygen to create caller and callee graphs to understand code flow. Unfortunately, the tool hasn't really changed in more than a decade.

My baseline was non-HC "vanilla" residuals; I didn't do a meaningful HC run to compare.

My application has some particularities (important and easy to identify per-token signals) that result in values growing (about 3x to 10x) through layers even in the baseline.

I implemented this for a toy 8M ViT-style model. Got neutral results. This is just an anecdote and is not representative - I think mHC will help with larger parameter sizes and larger token counts.

Some types of software aren't too far from that paradigm.

Tangentially related - I was doing some highly sensitive analog electronics work. It's surprisingly hard to find capacitors that don't exhibit microphonic behavior - i.e. they don't produce (miniature) voltage swings when you tap the PCB.

There was previous discussion a few days ago on "Solving The Problems of HBM-on-Logic" that I think is relevant.

https://news.ycombinator.com/item?id=46302002

https://morethanmoore.substack.com/p/solving-the-problems-of...

I've heard from a friend who works in the embedded space that Tensorflow Lite is still the only realistic (supported by vendors) game in town for running ML models on microcontrollers such as ESP32, nRF, etc. The hardware support listed for this project seems like it's targeting much "fatter" MCUs (Android, etc).