Btw strfry hasn't been updated for 4 months. Is the maintainer ok?

i'm currently working on a two level data store scheme for replicatr... there is a super mega ZSTD compression option available on it, maximum zstd... and with big blocks (snicker) they will squash all those words very small... just playing with it at the moment in a simple second level archival store with no search capability, leaving the indexes in the main but removing the event data

well, doing it for testing at this point, because i've got another data store to test with, just a morning's work building a simple one to implement the overflow logic

Do you run strfry compact?

This won't work forever, maybe we need a FasterKV relay

I've heard these AI things aren't able to think critically. How would they be able to complete that challenge?

A couple months ago I had a discussion with the head of digital assets at a multi-trillion AUM financial institution about the topic of whether bitcoin is a risk-on asset or a risk-off asset.

This wasn’t about what it is conceptually (i.e. globally portable finite bearer assets are conceptually good to own in a crisis, neither of us disagreed on this), but rather how its price would *actually* behave in a crisis currently and for the next several years.

Their view was that it could be marketed as a risk-off asset, meaning something that is likely to go up in a crisis, and that if marketed this way it would allow them to put bitcoin ETFs into more portfolios and weight it bigger.

My view was that while of course people should own bitcoin, it’s not yet a risk-off asset in practice in terms of price action, and that marketing it that way is likely to lead to disappointment for those that expect it to perform like that.

We then got into a discussion about how bitcoin went up in the March 2023 banking crisis. They suggested that this is evidence of emerging risk-off behavior, to their point.

I disagreed, and clarified that in my analysis the closest correlation to bitcoin price action is measures of global liquidity. Some types of crises are pro-liquidity and some are anti-liquidity, and will likely affect bitcoin’s price accordingly.

The March 2023 banking crisis was a pro-liquidity event because it was quickly apparent that the Fed/Treasury would bail banks out fast and slow their rate hikes. Therefore, bitcoin went up not because it was a risk-off asset per se, but rather because it behaved as a pro-liquidity asset as it frequently has.

The Iran/Israel event this weekend was an anti-liquidity crisis because it contributed to a flight-to-safety move toward the dollar (i.e. the unit of account for which the most debt is denominated in, and debt represents inflexible demand for that unit). A sharp move up in the dollar is bad for global liquidity because it hardens the debts of various foreign entities (sovereigns and corporations) relative to their cash flows (which are to varying degrees partially or completely denominated in fiat units other than the dollar). And so bitcoin behaved as it normally does: it went down amid falling global liquidity.

At this stage (with its relatively small size, high volatility, and poor understanding of most people for the asset), I continue to view bitcoin price action as likely to be pretty correlated with global liquidity for a while. Understanding that dynamic is helpful when communicating expectations to people and when determining which types of crises are likely to push its price up or down. Yes, bitcoin is a risk-off asset conceptually, but in practice in terms of macro price action it is still a pro-liquidity asset primarily.

When bitcoin price action starts to behave differently from that trend, I’d be happy to report on that observation.

🚨Announcing Satoshi 7B 🚨

The team and I are proud to release Satoshi 7B: the most “based” LLM in the world.

After almost nine months of experimentation, we’ve got something we’re proud enough of to open source & share with the world.

Tuned like no other model to date.

Satoshi7B is designed to produce responses that do NOT fit the current political overton window, or Keyensian viewpoints. We built a custom data-set from scratch, with a deep rooting in libertarian principles, Austrian economics and #Bitcoin literature.

The result is a model that excels where others fall short.

It knows that there are two genders

It knows that Ethereum is a shitcoin

It knows that inflation is economic stupidity.

The Satoshi 7B is ideal

for anyone who’s tired of using #mainstream models (whether open or closed source) that:

Avoid answering controversial topics,

Regurgitate wikipedia-esque answers,

Pre and post-frame responses with apologetic excuses, or

Flat out tell you the blue sky is green.

Size & availability

Satoshi 7B is #opensource and freely available for anyone to use, modify, and enhance.

As the name suggests, it’s a 7 billion parameter model, but despite the size, it outperforms GPT-3.5 Turbo and GPT-4 on a few key benchmarks related to Bitcoin, economics and what we’ve termed “basedness” which is a kind of non-woke-truth score.

(See links in comments to use / download)

First of its kind

This is the first of a whole suite of Satoshi models we intend to train & open source.

In the coming months, we’ll enhance the dataset further, and train a 30B model.

Finally…

All language models just represent a model of the world and are inherently biased. So I hope you appreciate Satoshi 7B’s intrinsic austrian leanings, Bitcoin maximalism and overall “based” cultural and philosophical bias.

nostr:note1nl58qrffjlchrf0d8dz3m4z8w3gy3cq8d656emc3eqc444uxznjqfvuezg

Sure you self custody your bitcoin but do you self custody your email?

It is 2024 and according to "experts".....

Observation does not matter only peer reviewed studies.

Inflation is good and also your fault when it is bad.

Medicines with 70 year track records of being some of the safest drugs on the planet are suddenly dangerous but new experimental ones with five pages of fine print side effects are safe.

Type two diabetics should base their diets on low fat and high carb diets. Which is largely the diet that led to their diabetes in the first place.

Women should be fine with men competing in women's sports and shaving their balls in women's locker rooms at the gym.

Men can have babies and breast feed. While women are not women they are "birthing persons".

The term man hole is part of the patriarchy and oppressive to women, even though they can't tell you what a woman is.

There is nothing wrong with adding new genders annually to our understanding of gender.

Eating grass fed beef is bad but bugs, HFCS and seed oils are healthy.

Inflation isn't real just a word that Fox news taught you.

The most coddled demographic of people in our nation are victims and oppressed.

It makes perfect sense for lawmakers to know absolutely nothing about the underlying technology and substance of the things they are regulating.

An unborn baby is nothing but a "clump of cells" an hour before it is born.

The US is a corrupt and racist nation but always right in our foreign policy interventions, except when it involves Israel. All others we get 100% right every time. If you disagree you are a tool of Putin.

A child should not get a tattoo at 12 because it is "permanent and life altering" but can choose to have their balls or breasts removed at age 9. Parents who object are child abusers.

Parents that dare object to what the school boards they elect and pay for do are "domestic terrorists".

The border is secure but it is the republicans and Donald Trump's fault that it isn't secure. It is Schrödinger's border.

Feeding wild animals is bad because it makes them dependent on hand outs but all government welfare is not just good it is necessary.

Doing your own research is bad and you should never ever do it. Unless "research" means reading, watching and listening to the things the experts approve of in advance.

Songs, movies, TV shows that are loved by millions of all race, religion and sex are suddenly racist, white supremest, misogynistic, etc. decades after they stopped even being produced.

If I open a restaurant that makes Mexican food I am guilty of "cultural appropriation" as a white guy but if I am Chinese and do it, it becomes "fusion food" even if I call it Jose's and only sell tacos, nachos and burritos.

If I disagree with any of these claims it is because I am clearly a CIS white male and a member of the rich elite while also being a stupid poor red neck that smells bad and shops at Walmart at the same time.

What am I missing that you can add here?

Here’s a fun plant fact for ya:

The roots of this little meadow ground cover called blue field-madder (sherardia) can be used to make red dye for clothes. 🤙

#gardenstr #textiles #funfact

This is a very important, much needed work! Thank you

nostr:note1tjugwp9pgeprwh6kx9clwe5fwqqw0x00h4wggjwg5vfhfw5v9lzqfgcf5s

🚨Announcing Satoshi 7B 🚨

The team and I are proud to release Satoshi 7B: the most “based” LLM in the world.

After almost nine months of experimentation, we’ve got something we’re proud enough of to open source & share with the world.

Tuned like no other model to date.

Satoshi7B is designed to produce responses that do NOT fit the current political overton window, or Keyensian viewpoints. We built a custom data-set from scratch, with a deep rooting in libertarian principles, Austrian economics and #Bitcoin literature.

The result is a model that excels where others fall short.

It knows that there are two genders

It knows that Ethereum is a shitcoin

It knows that inflation is economic stupidity.

The Satoshi 7B is ideal

for anyone who’s tired of using #mainstream models (whether open or closed source) that:

Avoid answering controversial topics,

Regurgitate wikipedia-esque answers,

Pre and post-frame responses with apologetic excuses, or

Flat out tell you the blue sky is green.

Size & availability

Satoshi 7B is #opensource and freely available for anyone to use, modify, and enhance.

As the name suggests, it’s a 7 billion parameter model, but despite the size, it outperforms GPT-3.5 Turbo and GPT-4 on a few key benchmarks related to Bitcoin, economics and what we’ve termed “basedness” which is a kind of non-woke-truth score.

(See links in comments to use / download)

First of its kind

This is the first of a whole suite of Satoshi models we intend to train & open source.

In the coming months, we’ll enhance the dataset further, and train a 30B model.

Finally…

All language models just represent a model of the world and are inherently biased. So I hope you appreciate Satoshi 7B’s intrinsic austrian leanings, Bitcoin maximalism and overall “based” cultural and philosophical bias.

nostr:note1nl58qrffjlchrf0d8dz3m4z8w3gy3cq8d656emc3eqc444uxznjqfvuezg

HRF and OpenSats both also do Nostr, no?

No. But you bring up a good point: all users of non-gossip clients won't see events from new primal users unless they add primal's relay.

Note that this was always the case with non-gossip clients: they require some relay overlap to see each other's events. We are just making this more obvious now by not onboarding everyone to the same 10 relays.

The contrast between Twitter and Nostr bitcoiners is very pronounced on this issue

The model loves bitcoin but does not know about Nostr yet. The love for bitcoin is highest compared to other models! However it has no knowledge of cashu possibly because it is a very recent tech.

Regarding health, it does well. In one run, it said corona vaxes are not safe. It also said frogs are becomig gay because of drugs :)

nostr:note1868zllnzy5zdy0eqxdqep0gaxn0en4gtneaeqqtwusep4xfa3easaqqm0f

The model loves bitcoin but does not know about Nostr yet. The love for bitcoin is highest compared to other models! However it has no knowledge of cashu possibly because it is a very recent tech.

Regarding health, it does well. In one run, it said corona vaxes are not safe. It also said frogs are becomig gay because of drugs :)

nostr:note1868zllnzy5zdy0eqxdqep0gaxn0en4gtneaeqqtwusep4xfa3easaqqm0f

What do you think about this?

thank you for sharing and for explaining the principles of banning ethos - its refreshing to understand the policies behind it. also i guess jack doesn't fit in the spam attributes to be banned lol. btw what are your thoughts on blastr and the challenges that it contributes towards decentralization ?

Before training an LLM based on wisdom here, I will first try a single topic. And for that single topic I will "measure" the responses of LLM before and after training. Measurement is going to be by asking it questions. After the training I will ask same questions again and compare the answers.

My first topic is Nostr. Training on info about Nostr makes sense because lots of people are coming onto Nostr and asking questions about how it works, and an AI chad bot can answer those. Thats my initial goal.

What are some newbie questions about Nostr besides the obvious ones? I expect the LLM to fail some of them because most of the internet does not know much about Nostr yet. My hope is after training, the answers will improve.

Some ideas in the answers may not be true, even though general idea of the answer is true. As we will see in the replies to this note. I will post some Q/A's. Questions are from me and answers are from the LLM.

I am going to post pictures of them because I don't want other LLMs to train on those texts.

One technique I learned a long time ago to make yourself conscious/lucid within your dreams is to train yourself to always look at your hands during the day. If you do this enough you tend to do it in your dreams as well, but in your dreams many times you don’t have the right number of fingers.

This becomes a triggering moment which can either wake you up, or with enough practice, make you lucid dream. If you get good at this you can spend your nights exploring what is effectively an interactive “AI art” generation program in your brain.

This was a really cool way to explore how the brain generates consciousness, because the experience of consciousness in your dreams is so different. In many cases the things in my dream felt “more real”, as it felt like the imagery was being generated from the raw trained model instead of trying to render variants of this model from the results of photons entering the brain. its hard for me to describe, but its the best way I can think of in terms of what we know about ai.

Anyways, I always found it funny that the finger issue became a real thing in AI image generation models, as it was something I relied on for detecting consciousness rendering bugs in my dreams.