kristianp

> That sentence had the exact pattern the skill removes.

Even that sentence sounds AI to me. Humans don't write like that, with "the skill removes" at the end of a sentence.

Which was 816B parameters.

Yeah, I use Openrouter with the zdr parameter.

You're absolutely right!

I share your disapointment at the lack of options for bitmap fonts. For linux users the lack of Wayland support means terminals need to do the font support themselves for bitmapped fonts. After using them in Konsole, I wanted to have the same sharp fonts on the web, but that also seems a lost cause.

And it's so easy to see that it's bullshit. Just go to the web tab on Google search and there's the results without an AI overview. The search results still use conventional algorithms, the AI overview is an LLM interpretation of the search results, unless the query is so simple the small LLM can answer itself.

Grammarly does use LLMs for a while.

Just a clarification, Google's search revenue actually went up in their latest quarterly reporting. It just didn't go up as far as analysts were predicting.

> Three separate height restrictions pressed against the building from above, and the car claimed its space from below.

Very AI style in this whole article. Some good photos though.

> efficiency of the generators with jet engines could be increased a lot if the heat from the exhaust gases would be recovered and used to generate extra electric power, with a steam turbine

I'm guessing most gas power plants use the combined cycle, otherwise you'd be wasting the energy of the exhaust gases. That uses a steam turbine heated by exhaust gases from the gas turbine. https://en.wikipedia.org/wiki/Combined-cycle_power_plant

Yes, Should probably be a sub-section or a single sentence in https://en.wikipedia.org/wiki/Novelty_bias

Why is there no FaceId? Is the camera too low-res?

Looks like Qubes OS is still alive: https://github.com/QubesOS/

Interesting difference between Go and Python, `c.counts[name] += n` can allocate if name isn't already present, whereas in Python you get a key error.

That some people think in black and white?

Alphabet shares down 13% today: https://www.cnbc.com/2026/10/29/alphabet-google-microsoft-ms...

They've reduced size in every dimension, here's the 14 inch size from the specs for example:

Oled:

    Height: 1.50 cm (0.59 inch)
    Width: 31.08 cm (12.24 inches)
    Depth: 21.63 cm (8.52 inches)
    Weight (M5 Pro): 1.38 kg (3.0 pounds)
    Weight (M5 Max): 1.40 kg (3.1 pounds)

Ye Olde:

    Height: 1.55 cm (0.61 inch)
    Width: 31.26 cm (12.31 inches)
    Depth: 22.12 cm (8.71 inches)
    Weight (M6): 1.55 kg (3.4 pounds)
    Weight (M5 Pro): 1.60 kg (3.5 pounds)
    Weight (M5 Max): 1.62 kg (3.6 pounds)

The lack of a 16" model for the M6 is consistent with Apple's current minimum of M5 Pro chip.

Why a Pi 4 Model A and not a Pi 5? Perhaps because LPDDR4 is cheaper/more available than LPDDR4x?

Deepinfra has a "contact sales" link. I assume personal accounts aren't available.

I found the transparency of windows in Vista and 7 to be a useless distraction. Found Windows 8.1 to be refreshing in that respect, although the menuless apps were poor.

Fascinating images. The article linked by the parent has some info https://arstechnica.com/science/2026/10/meet-the-2026-winner...

His website is https://www.macrofying.com/ which is mostly links to the usual image/video sharing sites.

I don't understand why HN chops off the leading "Why", it completely changes the intent of the title.

Edit: removing reference to the leading "How", that seems to be done editorially by the poster of https://news.ycombinator.com/item?id=50227356

I reckon the 64GB variant might be good value for running Qwen 27b. I'm not happy with the flat gapless keyboard though. It looks uncomfortable to work on. Has anyone had any experience with this type of keyboard from Dell?

If running LLMs, I'd want to know about the thermals. Looking at the 3D view, there are no holes underneath, just a large one on the side and at the back.

They've shown the calculation for number of concurrent requests limited by kv cache size. I guess for each new request loaded into the queue, the kv cache for that context will need to be loaded from a supporting storage server, that will need to have a lot of ram to store the active caches of multiple requests?

I don't think it will all be paid off. I said once it is gone. I think, like the Dotcom bubble, many companies will go bust and their debt will be worthless.

I think a lot of people on hn intuit that Rails isn't actually a jit compiler from their knowledge of what Rails actually is and haven't bothered to open the article or comment here on it either. It's clearly some kind of clickbaity metaphor, but I wouldn't know what, as I didn't click on the article either.

Or the code has been added to the training data.

Once most of this debt is gone, whether through companies going bust or the debt being paid off or inflated away, AI will become cheap enough for frontier models that any productivity improvements will be visible in the user's bottom lines.