0x01 = 00001 0x0c = 01100 0x12 = 10010 0x1f = 11111 0x1c = 11100 0x19 = 11001 0x02 = 00010

00001 01100 10010 11111 11100 11001 00010

5bit byte to 8bit byte

00001011 = 0x0b 00100101 = 0x25 11111110 = 0xfe 01100101 = 0x65 00001101 = 0x0d 00000100 = 0x04

0b25fe650d04!

Hey @ek, I made a series of live streams in 2024 explaining how BOLT 11 is implemented in Core Lightning. I'm no longer working on LN, but I guess the implementation has not changed much, and neither has the protocol, at least for BOLT 11.

I think you may find part 4 interesting.

I hope this helps.

The literary agent thing has always seemed like a scam to me. Traditional "Publishing" (especially novels) is apparently a world where you have to pay someone who is friends with a publisher to tell them about your book. The whole thing feels icky. And also: don't we have the internet and half the world with pocket devices that can display text? If ever there was an industry that needs to burn to ashes, traditional publishing sure feels like it.

Being able to say you wrote a 2000 words a day while cleaning toilets at a hostel in Morocco sounds like time well spent.

I've found that being present and listening attentively to someone is bafflingly powerful; in fact, there's basically an entire psychotherapeutic school built on that idea.

You probably did her more good than a week's worth of whatever the halfway house accomplished, and that's not even knocking on the institutions of halfway houses.

Thanks for the comment. It's strange that everyone expects writers to both have commercial goals and be hitting them.

When people ask 'how many books have you sold?' it's just so they can do the napkin math on how much money you have. The economics of writing (even self-pub book sales) are comically bad.

I always wonder why people dont ask hobbyist rock climbers if they are going to go on the pro tour, or ask those who sing in a choir when their next platinum disc will come out.

When you do write for money, it is usually to sell stuff for other people, and you crestive freedom is somewhat restricted.

Still, then important thing is to write. Reach whatever goals you set, and don't let anyone drag you down.

Yes, A Moveable Feast is a riot. Sun Also Rises is probably my favorite though (I even used to live in Pamplona and write in the cafe he frequented).

Really fascinating stuff.

I'm also interested in how the engineers who implemented this feature at Instagram felt. I might have been tempted to resign. It is just so stupid and dystopian.

But, honestly, I am not surprised. Stacker News is not representative of the population. In my interactions people, I notice that many of them do not have confidence in their ability to write meaningful or thoughtful things. They believe that the AI writes better than them, so they think that by having AI generate a comment or an email, that it will be of better quality than what they could write.

I think they are wrong, of course. I'd rather read a clumsy, genuine comment than a flowing, fake one. But still, many people really do think this way. Coupled with inherent laziness, and you can see why people would actually make use of this feature.

The only ai writing application I like using, is the suggested email replies.

There are usually three options and one usually sounds right to me. Is that substantially less authentic than me taking a minute to land at the same place either way.

To be clear, I’m talking about the pro forma courteous replies, not matters of any substance.

One of the unseen things on Trump’s side is how the larger counterfactual deficit would have been handled.

Tariffs are bad, but they may be less bad than inflation and other available forms of taxation. There’s good reason to suspect that’s the case, at least for the domestic economy.

I’ve never lumped all those professions together in one garbage men category. It’s a great point.

The broken windows fallacy certain does apply to Trump's logic of raising tariffs to "bring manufacturing jobs back to America", because what that logic misses is that by having China do the manufacturing, that frees Americans up to do other, potentially more desirable jobs.

That being said, the broken windows argument isn't going to be convincing to people who are having trouble adapting to this new economy. It should also be said that a knowledge-oriented economy may be more cutthroat and "winner-takes-all" than a manufacturing-based one, because one person can more easily service millions of people with their knowledge production (this includes music and entertainment as well), than one factory worker can service millions of people with their factory work.

The consequences of the loss of manufacturing work and the transition to a more knowledge-oriented economy has resulted in degraded lives and loss of dignity for a lot of people, which is a big reason why Trump gained popularity by speaking into their problems.

I'll add to this that a lot of the "knowledge based economy" is potentially welfare-reducing make-work, whereas manufacturing is more likely to be a net positive overall because you're actually making stuff that people can use. But certain roles like government regulators and the associated compliance officers, lawyers, accountants, etc, are jobs that aren't necessarily value-enhancing, but only exist because of the complex maze of bureaucratic rules that have sprung up due to excessive regulation.

It's kinda like saying the economy is growing because we're employing a lot more garbagemen. More garbagemen isn't a sign of good economic health: it's just a sign of more waste. So more accountants, lawyers, and compliance officers isn't a sign of economic health, it's just a sign of more value-destroying burdensome regulations.

Another, much more salient example, is the growing spending on healthcare. Growing spending on healthcare isn't necessarily a sign of economic growth, if it means a population is getting sicker and/or spending more on wasteful treatments.

I'm actually open to this.

But I'm at the same time not open to this conclusion, simply because if I were to instruct an AI to leave a note exactly how I want it to be, then I expect that it would be slower, not faster, than writing it myself?

So what we get are low effort interactions in settings of low expectation. But then, is instagram a place you go for the high quality interactions? It's a match... slop4slop?

Me when an old crone tries to read my palms at a flea market

Everything will become about proof of work.

I see how well-written comments on LinkedIn gain more traction than posts. That's because 50% of the people are posting drivel, and the other 50% are commenting with AI.

The value of words trends to zero when producing them costs no time.

That's why words produced with thought and effort WILL continue to be valuable.

Final thought: If platforms like Insatagram become pure AI drivel-flows, people will cease to find it valuable and will eventually migrate elsewhere.

I use Claude regularly for extracting the important ideas out of articles I feel that if I read I would be disappointed that I spent that amount of time. It's the equivalent of watching a video on YouTube at 2x speed. Sometimes you just need a tool or technique to discern for you whether a wall of text is worth your time before you give your time to that wall of text. Often I only learn one or two useful details from a long article whose author is conditioned to hit a word count rather than cogently and quickly getting to the fucking meat. Having read many thousands of articles before LLMs came along I feel confident that Clause is getting the salient points out better than the author could.

On the writing side it's similar. I think it's time consuming to communicate an otherwise good idea cogently, and LLMs are good at that. If I think I know what I'm going to say and how I'll say it as soon as I start typing, like this comment, I'll just type it into the field and post as is. But if I feel I need help tightening it up I'll copy paste into Claude to clean it up, then paste it back in. That makes me reader's life easier, even if they end up using an LLM on the other end.

Yea, no thanks. Let the retards keep their bots and atrophy their neurons.

There can be a lot of information in an email reply saying "Ok, thank you!", or "Umm, can you explain? what's this?" - in a business setting. I used to send tons of emails like this.

Those are the ones I’m talking about. There are all these little social conventions that go into those and that are tedious to think through, so when I’m offered one that sounds right I just take it.

I asked chat gpt about the premise of your title. It gave me a BS answer at first and then a page-long extrapolation. I got the real answer pushing back demanding a better up front response. This is what I got:

"AI doesn’t write the way humans “tend” to write. It writes the way systems that are rewarded for not causing problems write.

That’s the core reason."

Don't be. This is exactly why teamwork is important.

Must've been a copy/paste screw-up. Did you figure out the conversion yet? I tried it again and fail at hex 0x02.

0x01 = 00001 0x0c = 01100 0x12 = 10010 0x1f = 11111 0x1c = 11100 0x19 = 11001 0x02 = 00010

00001011001001011111111001100100010 -> 1496314658 ✔️ 00001011 00100101 11111110 01100100 010 0x0b 0x25 0xfe 0x64 0x02

Those 3 bits (0x02) might already be part of the next byte. Makes sense 'cause 0x50 is 01010000.