LLMs are to copyright as samplers were to music in the 80s they let you remix anything. they aren't intelligent. they are search and replace engines with a lossy, very expensive compression algorithm. they can deal with text, binary data, and structured data (eg, images, 3d models, stuff like social graphs, maps etc), and make mostly accurate recompositions of the input data.

don't let the corporate AI hype deceive you LLMs are souped up search and replace systems that's why they sometimes do the most crazy shit

organisations like the ethereum foundation are the recipients of humongous amounts of support from fiat banking. on top of this, they are spewing tokens on the market like there is no tomorrow. to describe the "gifts" of this endlessly debasing, fiat-propped shitcoin as being "charity" and saying that is the same thing as someone giving their own hard earned bitcoin to a cause. uh uh. bitcoiner's made a sacrifice. the shitcoiners pressed ctrl-p that's aimed at you rage bate lola l33ts

i was just sitting here scrolling the last 5 minutes and completely failed to register that i wasn't wearing my glasses and was having no problems reading the screen. it's almost back. i guess my kidneys must be starting to show signs of recovery. being that i'm finally not drinking, i was hoping that i'd recover a bit faster, but the alcohol messes up your sense of thirst. idk how long this effect lasts but so far it's a month. i probably need to up my salt intake too. so good being able to see without anything between my eyes and the screen. woot.

just relating that to that article someone published recently about programmer identity crisis the problem is that almost nobody understands the real identity of this technology. it has been conflated with what our brains do - although it is related, and similar in some aspects, as far as we understand how nervous systems encode memory and search for data. it's just the memory part. not the survival imperatives of instinct baked into the hardware.

#orly🦉#devstr #progressreport i'm currrently working on a little tool that aggregates all of the content it can find on the network from a single npub. it uses these relays for seeding: "wss://nostr.wine/", "wss://nostr.land/", "wss://orly-relay.imwald.eu", "wss://relay.orly.dev/", "wss://relay.damus.io/", "wss://nos.lol/", "wss://theforest.nostr1.com/", and when it finds the follow list events in the results it adds the relays on them, schedules them to pull the whole time window. https://git.nostrdev.com/mleku/next.orly.dev/src/branch/main/cmd/aggregator and as an extra bonus, most likely it will fetch nearly all events of of the npub that exists on the network, and then it prints out the bloom filter which can be used to recognise the events you pulled in that run, which could be used later, perhaps, even, it could be added as a parameter to specify a file with a past filter in it. yeah, ok, it will do that too, so you can copypasta the bloom filter and run it again later and it won't waste any time trying to download events that were already found previously. the bloom filter output shows the window of the aggregation run at the top of it, and can be copypasta to a file, and use that to eliminate existing already found events, and it appends the events in this mode instead of overwriting the whole file. https://next.orly.dev/ tag v0.17.16 has the full version with this extra feature added. there may be bugs in it somewhere, beware. i ran it for my npub and found about 160mb of events, which includes all mentions of my npub... probably the larger part of that list is my presence on other people's follow lists with their hundreds of follows. it's uploaded to wss://relay.orly.dev/ now... should be the complete collection of my things. the only thing is now, after importing those events there seems to be a bug in the relay web UI, it's showing events repeatedly and popping up some popup message about some error in the javascript. anyway, there it is. if you ever wanted to snarf all of the events that probably exist relating to your npub, there is now a tool.

it's not even news. quantum computers are known to be able to outperform standard super computers for some time now, at least on a power consumption basis. the supercomputers are cheaper to build though, to be fair.

i like what you did with the column buttons. you could actually make it just automatically switch to single column starting from like 1280 wide and remove the button. most of the time you want single column in a narrow window.

https://github.com/yourbasic/graph the Virtual Graphs looks especially interesting, since it doesn't build the graph until you ask it to show you parts of it.

want. also with the ability to do all that neat stuff like electron engine and use different languages and run them inside containers instead of using javascript for anything more than reactive interfaces and instant config and things like this

Describe the problem you are having: Use the drag and drop or paste to add images of the problem. What device and operating system are you using? Device: Linux x86_64 Operating System: X11; Linux x86_64 Browser: Chrome/141.0.0.0 Safari/537.36 Language: en-US (en-US) Screen Resolution: 3840x2160 Viewport Size: 1288x2066 Color Depth: 24 bits Device Pixel Ratio: 1 Timezone: Atlantic/Madeira Touch Support: No WebGL Support: Yes WebGL2 Support: Yes Service Worker: Yes Notifications: Yes Geolocation: Yes Local Storage: Yes Session Storage: Yes Online: Yes Cookies: Enabled What steps did you take to reproduce the problem? What did you expect to happen? What actually happened? Please provide any other relevant information

i've started collecting material to train an LLM... so far i've got mises, tesla, szabo, a bunch of DST theory papers, some space weather, the bible, enoch, jasher, jubilees, critias... gonna just pause every so often and trawl for more resources until i've got way too much for me to make sense of. i'm going to make a little model that runs on my 16gb radeon 7800xt, that is specifically focused on economics, cryptography, mathematics, chemistry, physics, computer science, and i've still gotta dig up a ton of Go stuff, repos, papers, documentation of various kinds. i might start training it before i've collected all the data tho. from what i've read it's a weeks long process. but if i specialise it in the kinds of topics that i have use for being able to dig through to back a hypothesis, it can help me check what i'm doing and make sure it matches up with the various models i've fed it to build from. i'm not convinced that general purpose AIs like Claude and Gemini et al are even slightly optimized for science. i'm betting the woke content percentage of those models is too damn high. too much marxist theory, too much mainstream narrative bias. especially not computer programming. oh yeah, they say claude knows basically all the languages but my eyes must be lying when i watch it constantly forget salient syntax matters over and over again like a green n00b learning to program.

btw, has anyone else noticed that it was the rise of bitcoin mining that led to the tech that made LLMs practical (ASICs that do hash functions really fast). i'm pretty sure that the evolutionary pathways of the tech directly ground the answer that solved the problem of markov chains and made them better than filler generators was about the hash function. crypto hashes are hashes that are extremely, practically impossible to reverse. LLM hashes are the opposite, they let you find things that are similar. in every other way they are the same thing: a mathematical formula that changes a number deterministically into another number. bitcoin made it valuable from there it graduated into becoming useful. hash grinding is now a major industry in several areas. it's always comical to me how people think that it solves spam problems. it's just a friggin ratelimiter that you can't cheat.

brazil is a movie made by that python guy, Terry Gilliam, who also made 12 monkeys.

#bitcoin price is free to stay down for the next 24 hours so i get paid at this price :)

so, the wasps in the garden that i've had a couple of run-ins with are called "yellow jacket wasps" and they are actually kinda useful parasites of some pests that cause problems in gardens. but i'm not getting stung again. so, the best way to do them in, is to make a big pot of boiling hot soapy water, pour it through the openings in their nest, then some heavy black plastic covering a 2x2m or so area, seal it up so they have trouble getting out just gotta find a giant pot to do this with and get a sheet of plastic. they aren't really a problem but they are so close to the house you see them flying near the house a lot. they do sting but not usually except when you hammer their nest with a line trimmer. fire is useless against them, though i have thought about the idea of phosphorus or thermite. lol. actually, now i'm thinking about it, they usually build on old tree stumps that have mostly been submerged, so some potassium nitrate or guano might be good for helping reduce the chances of them ever returning to that spot. after they are gone, dig down and put the powder on it and then cover it back up and let the bacteria go nuts making amino acids out of it all. there could be another nest in the garden too, i haven't fully cleared it yet, in fact 😅 another one down on the side near the sorta cliff edge. fingers crossed i can just rid this one nest and not get stung trimming the weeds out. also need grass seeds and clover to drown out most of the weeds. the stings of yellow jackets is not as bad as a bee sting but it persists for a few days and it does swell and redden a little. definitely not recommended garden pet.

incidentally, ordinals in latin: primus, secundus, tertius, quartus, quintus, sextus, septimus, optimus, octavus, decimus

an important thing that just happened is the project that puts together a map of the magnetic field strengths and alignments across the planet every 5 years, and make forecasts about the direction they expect it to change. well, 5 years ago they had to revise it a year early because it was affecting so many industries, especially aviation and navigation. these people are now saying that they have to do this again. second 5 year cycle that has wildly diverged from their forecast. https://www.ncei.noaa.gov/products/world-magnetic-model - The WMM predicts an intensification of the South Atlantic Anomaly (SAA) between 2025 and 2030. - The WMM uncertainty analysis presented in the report has been updated with new data, resulting in a slightly revised error budget. - The WMM report provides altitude ranges for varying levels of geomagnetic activity. https://www.ncei.noaa.gov/news/world-magnetic-model-receives-upgrade "error budget" their forecast was already totally known to be wrong months ago. error budget