Given the context of the HN audience, it's probably something you can do.

Discrepancy is that we're mixing (lunar) months and weeks with solar timekeeping, in a solar calendar. These are fundamentally incompatible, so we've gone with cramming the approximate periodicity of the lunar calendar into the solar calendar, while ignoring the fact that we're no longer tracking the moon, and that the weeks don't line up with the year, and the fact that the months are randomly different lengths because they also don't line up and we don't want a weird half-month at the end.

Another potential fix would be having two calendars. A lunar calendar for weeks/months, and a solar calendar for seasons/years.

GC threads are generally often useful on multi-tenant systems or machines with many cores, as Java will default-size its thread pools according to the number of logical cores. If the server has 16 or more cores, that's very rarely something you want, especially if you run multiple JVMs on the same host.

Not JVM options, but these are often also good to tune:

    -Djdk.virtualThreadScheduler.parallelism
    -Djdk.virtualThreadScheduler.maxPoolSize
    -Djava.util.concurrent.ForkJoinPool.common.parallelism

In my experience this often both saves memory and improves performance.

You can get into difficulty with kubernetes here, as your jvm will detect all cores on the node but you may have set a resources limit on the pod/whatever, so it’ll assume it can spend more time doing stuff than it actually can, so often times it’s quite necessary to tune some things to prevent excessive switching etc.

Modern JVMs will detect orchestrator-set cgroup limits and size themselves accordingly. If you, for example, set a cpu limit for a pod to “1”, the JVM will size itself as if it was running on a single core machine.

It's in the picture, I presume. Just gotta brush up on that Koine Greek. Or if you read Egyptian hieroglyphs already, you can use the Rosetta Stone to reconstruct the Koine Greek from first principles.

It's not Koine and it's in verse.

Yeah, coins have distinct faces and can be flipped to generate random numbers.

Well it's a trade off right?

If the benefit outweighs the drawbacks, you say yes, and when the benefits evaporate leaving only the drawbacks, it's a no.

Well yeah but this guy portrays it like a moral decision

I didn't think it was a moral decision at the time. Now I do. But that's just, like, my opinion, man.

What decisions aren't?

The very open ended core question of the field of ethics is nothing more or less than "what is the best way to act".

Many decisions aren’t.

Nobody says “I don’t like potatoes” because they want to look like a noble good guy.

Many people say, “I don’t like ads” because they want to look like a noble good guy.

There's a moral aspect even to deciding to claim that you like potatoes.

For one, if you don't like potatoes, and you claim to do so, you're a liar, and lying is considered by many to be vicious.

On the other hand, if you genuinely do like potatoes, the moral thing would be to acknowledge this, rather than concealing this or claiming you're more of a rice stan, as again, honesty is considered by many to a virtue.

Is this the ads or the girlfriend?

Well, yeah. It's the explicit design philosphy of the language to wait and see what works as other languages do the experimenting.

I wonder if a language like Latin would be useful.

It's a significantly much succinct semantic encoding than English while being able to express all the same concepts, since it encodes a lot of glue words into the grammar of the language, and conventionally lets you drop many pronouns.

e.g.

"I would have walked home, but it seemed like it was going to rain" (14 words) -> "Domum ambulavissem, sed pluiturum esse videbatur" (6 words).

Words <> tokens

I think speculative decoding eliminates a lot of the savings people imagine they're getting from making LLMs use strange languages.

Yeah I'd prefer to be seen contributing to an ecosystem[1], rather than sucking all of the air out of web discovery.

[1] Of either collaboration or friendly competition. Keeps everyone involved honest.

I wonder if you can construct a function between the encoder and decoder such that for any given input, both the raw and manipulated embeddings decode to plausible meanings that are guaranteed to be different.