bccdee

The only really useful thing I run is a secure pastebin.

I often need to copy a sensitive string from one computer to another, and I don't want to worry about it getting preserved in the edit history of a Google Doc or whatever. A small pastebin service with basic authentication where deleted notes are deleted permanently. I use it all the time.

My exact setup. Main repo on a second internal disk, one replica on an external disk, second replica in backblaze. Replicas are kept in sync with `restic copy`. Backup integrity gets checked automatically with `restic check`. It all works great.

Oh this is interesting. Thanks.

If something behaves as if it has a goal, that's a functional goal. When I say this chess engine is "trying to take my queen," I'm basically correct, insofar as this is a useful way to understand my computerized opponent. If I give it access to my queen, it'll capture it, because that's what it's trying to do.

If LLMs are "just" statistical models, humans are "just" a bunch of neurons squirting chemicals back and forth. There's no pixie dust in our brains that makes us special.

LLMs are not conscious: They have no analogues for feelings or senses and no construct of selfhood. But if a statistical model had those things—if it did all the mundane, physical bookkeeping our brains do to produce "real" emotions and motivations—then there's no reason it couldn't be as conscious as we are.

Ok I agree they have functional goals and can functionally go rogue, but how do they functionally feel? They express feelings, but so do actors.

LLMs can't begin to have functional emotions unless they're (for instance) explicitly seeking out or avoiding situations based on the emotions those situations would produce in them. Has a chatbot ever responded to one of your requests with, "no, doing that would make me sad"?

Unless you're already manually setting temperature to zero on the models you work with, the deviation from the "optimal" path imposed by the watermark should be no greater than what you already get from the randomization.

> “He leaped at the chance” and “He jumped at the opportunity” are very similar sentences expressing the same general sentiment, but they are not the same. The exact words we choose when writing matter. I want any LLM I use to choose the very best, most precise words at every single decision point.

Neither of these is "better" or "more precise"; in fact, LLMs will generally choose randomly between these candidates based on temperature, and SynthID should not distort the output of an LLM any more than the default temperature settings do already.

I agree that those are not the same sentences, but if the difference matters to you, you shouldn't be using an LLM. This difference exists at the level of what sounds better and is more evocative; to an LLM, nothing sounds like or evokes anything. They simply do not write good prose.

I've turned against them, partly because they're too flexible. Many of the things a dash does can be done better by more specific forms of punctuation like semicolons, colons, or brackets, which give your sentence a clearer visual structure.

It's also easy to get too used to dashes and start sticking them in places where they don't belong. If you're using a coordinate conjunction like "and" or "but," a dash is awkward and redundant. "A period is great to express a complete thought — and it shifts the register of the next letter," can more easily be "a complete thought, and it shifts" or "a complete thought — it shifts."

Dashes take up a lot of space. Like exclamation points, they draw the eye. Does the eye really need to be drawn to "a complete thought; it shifts"? If you let them, dashes will replace all your other punctuation marks, and your writing will become a soup of horizontal lines. Dashes are a sometimes food.

The vast majority of readers can tell the difference between - and —. It's very noticeable. This sentence:

> I use dashes in place of commas, semicolons, colons, and brackets all the time - I just use the standard keyboard dash.

is noticeably weird-looking. Even a spaced en-dash looks a bit small; a spaced hyphen is just odd. But if you type it like this:

> I use dashes in place of commas, semicolons, colons, and brackets all the time—proper dashes, though, not hyphens.

it becomes much more conventional and legible. If you're going to use hyphens, it's better to admit the typewriter convention and use two, like so:

> I use dashes in place of commas, semicolons, colons, and brackets all the time -- I just use the standard keyboard hyphen, but twice.

I'm introducing examples. It's logical to judge a thing based on its context and circumstances, because those can tell you about the thing. A menu can tell you about a restaurant in the same way that the presence or absence of a cart can tell you about a street-food vendor.

> hotdogs are shit in my opinion

Then by all means, substitute the street food of your preference. Unless you disdain street food in general, which would be irrational, unless eating at a specific place is meant for its societal signaling value rather than for its food :)

> Is it a garage that has the same hygiene standards and medical equipment as an office?

Maybe! A garage can be cleaned very thoroughly, especially if you renovate with that in mind. I wouldn't bet on it, though. A good doctor probably could operate out of a garage, but they wouldn't. And a good restaurateur would take time to think about their menu.

These signals select for people with domain knowledge and experience. Status has nothing to do with it. Even the cheapest street food will be evaluated against its competition based on external signals, and so better, more experienced vendors will tend to know how to present themselves well.

All else being equal, would you rather eat a hot dog from a cart with a nice sign, or from a guy sitting on the sidewalk who sharpied "Hotdog $3" on a piece of cardboard?

No, it's entirely rational to judge a service based on external signals.

If you need medical treatment, will you go to a doctor operating out of an office, or a doctor operating out of their own garage? Theoretically, either one could be good or bad at medicine, but in practice, professionalism is a useful signal.

An ugly AI-generated menu shows a naive disregard for the actual experience of your customers. This attitude will bleed into the food quality.

> Every dollar spent on design is a dollar that is not going to the quality of my food.

Then throw something together in MS Word. That's free. It's not bad for something to be low-effort; it just needs to be commensurately simple.

Exactly. Modifier sequences and escape presses show up as special characters and can be inserted with `C-v Esc` or whathaveyou. It works seamlessly.