This is the approach recommended for starting a hobby as well. Buy the cheapest set of tools and if they break or are not meeting your needs.

I've been watching seaweedfs for a few years now. It is very cool tech and I'm itching to use it but unfortunately the docs are very bare bones.

Do you have plans to improve the docs? Some of the info I needed was inside GitHub issues / discussions.

Overall very cool project. One of these days I'll reach out about enterprise support for my usecase.

I am not sure whether you checked seaweedfs github wiki. Many people did not notice the many wiki pages there.

If you're a fan of single player checkout the mod scene. It's very good. UED First Light is excellent.

In a way, it might reflect accurate space and our universe. We really might be alone.

What is "sticky" in this context?

Experts vary per token in MoE, there is maximum flexibility. Good for driving down loss, bad for locality/gpu memory/bandwidth.

If expert selection were more constrained, inference systems could take advantage of it. Keeping experts cached would mean not needing to load them from disk/ram every token.

I don't understand the purpose. What am I missing?

The author "prints" to the eink display. Why not load say the pdf directly?

I've wanted something like this for a long time to print recipes that I'm about to cook. I don't want my phone in the kitchen getting gunk on it, so I'll usually print on paper, get the gunk on the paper, and then throw it out.

I'm not going to create a recipe PDF and load it on an e-reader. That's too much work. But I will happily press control-P on a recipe web page and select the eink display as the target.

I am no cook, but you my friend are cooking something here

Right! I like this usecase very much. For uses existing software infra to perform the required task.

I wonder now what else we've reinvented when it comes to sending data...

From the 4th line in the article:

> But getting stuff onto it was tedious. I had to join its hotspot and open a little upload website in my browser. Yuck.

Anyway, very creative, I like the idea a lot.

> I was thinking of this, but rendering PDF takes a lot out of a 400KB RAM'd device. It also needs a ton of things to actually work which the ROM can't hold.

> Kindles have a ton of RAM, this one didn't.

It's an e-reader. According to the specs it has native support for PDF.

It's an esp32 MCU, even if it can somehow run the PDF render pipeline, it won't be more ideal than just accept the raw byte stream and display.

I'd like my e-reader to show up as a printer, and when I print something on it it actually gets saved as a pdf on the reader. Like in the article, but multi-page pdf.

I often do 'print to A5 pdf => copy pdf to e-reader'. Would be nice to do that in one step from any program that can print.

> Bill's frustration is not unreasonable.

I find this single sentence not only awkward but frustratingly annoying. Bills frustration is beyond reasonable.

The lack of books and therefore lack of reading leads to reduce vocabulary. I read somewhere the average of known words by an Arabic speaker was substantially lower than English speakers.

When I switch to speaking Arabic (poorly) I struggle to find words because my vocab level is so bad. Imagine for an every day person, not being able to effectively express themselves. It would lead to a ton of frustration, miscommunication and misunderstanding.

I really need to improve my Arabic!

That really interesting.

To be fair, modern "English" is a nebulous smorgasbord of words from a lot of different languages. I'm sure we could afford to drop a lot of the words from clusters of synonyms.

> we could afford to drop a lot of the words from clusters of synonyms

Not really, because every synonym imparts a slightly different meaning, usually perceptable only to native speakers. That's why they persist, because they are useful even if not strictly necessary.

I recall a team working on machine translation starting by translating English into a simpler form before translating that into the target languages. This was a couple of decades ago. Perhaps not a new idea even then.

You can take those 'useless', redundant synonyms from my cold, dead brain cells!