exist270

HOLY FUCK. 🫪

I think I just discovered the biggest upgrade in #AI. 😲

It's not a frontier model, harness, workflow setup, API or any of that kind of stuff; it's a single, succinct system prompt that fixes almost all of the annoying stylistic nuances that we all hate about AI outputs. 😏

The system prompt:

"Only output US keyboard layout characters."

That's it. 🤷‍♂️

Em dashes, curly apostrophes, curly quotation marks & all the other dead giveaways that AI employs simply disappear & you get beautiful outputs that are essentially undiscernable from actual human writing because it only "types" characters that appear on your keyboard (like a normal fucking person would). 💁‍♂️

I sauced mine up to be:

"Use natural language according to proper grammar rules and only output US keyboard layout characters."

…but only because that's the style that I grew up with & prefer to read. 🙃

I have sought a solution to this issue for YEARS & in retrospect, the problem that I had was overindexing. I was explicitly adding more & more rules to the system prompt to get rid of the quirks that I didn't like, but the issue is that adding more rules decontextualizes the ruleset & reintroduces the quirks again later in the thread when the model needs to throw out some of the context in order to achieve an operation. 🫠

In theory, this system prompt reintroduces itself on every input (and is small enough to not be forgotten) & even if the model/API doesn't inject it before every input, the output itself keeps reintroducing it via the model's thinking & stylistic output, so it shouldn't ever dilute. 🤔

It turns out that less really IS more. 😂

Don't let the bots find this shit. 🤙

🩵 BLUE 🩵

This is a #VideoGame I'm currently working on. 💁‍♂️

It's a survival game set on a deserted tropical island. 🌴🌊🌅

The core concept originated as a #ShortFilm that I conceived ~15 years ago, but I could never get the talent, locations & timing to coincide, so it simply got shelved. I considered making the short film with #AI, but something felt empty to me about going that route, so I didn't do it. 🫠

I have an extensive history in #3D production & the tech finally caught up to be able to realize the concept as an open-world interactive game (as a solo dev), so I "broke ground" a few weeks ago. This is the first draft of the island environment. 🤷‍♂️

Perhaps releasing the #game will allow me to finally shoot the #film that I've always wanted to one day. 🙏

Ask me anything, #Nostr fam. 🤙

#AskNostr

AIN'T NO MOTHERFUCKING WAY THE #AI CLAPPED #BITKEY TOO, BRO. 🫪

Are these #Bitcoin / #BTC hardware wallet companies just not using leading models to penetration test their own shit & push updates/warnings? 😭

Nigga, what in the entire fuck are we even doing here in this industry? 🫠

SCAN YOUR FUCKING SHIT. 🤙

https://x.com/i/status/2083507377643606068

🚨 PRODUCTIVITY HACK!!! 🚨

You can use Brave browser's Leo AI tool to get your local #AI models to create a shortcut that reads YouTube transcripts & summarizes them for you into a nicely-organized cheat sheet. 💁‍♂️

Using this hack, you can consume video content faster than it's even possible to watch a video (even on 2-3x speed). I've "watched" 200+ videos in the past 2 days (some which were 2-3 hour videos); 2 weeks worth of #Bitcoin / #BTC content downloaded straight into my brain like Neo learning Kung Fu. 🥷

My current prompt:

Analyze the YouTube transcript by strictly filtering out only commercial breaks or advertisements. Every other element (including all facts, figures, names, arguments, opinions, side comments, conversational anecdotes, and references to individuals, regardless of how tangential they may seem) must be considered mandatory core material for inclusion. Using this fully retained content, generate one single, continuous Markdown table with exactly three columns: "###", "Topic", and "Details". This output must achieve absolute comprehensiveness, disregarding all concerns regarding brevity, censorship, or output token length. The table must exhaustively cover every distinct thematic segment of the retained transcript in strict chronological order. The "###" column requires sequential integers for each unique segment; the "Topic" column needs a concise, bolded summary header; and the "Details" column must list every single detail, argument, fact, or quote from that specific segment as an unrestricted bulleted list (•). Every single point listed in the "Details" column must be on its own separate line using "
" tags. Crucially, do not use any headers other than the three specified, add introductory/concluding text, or split the output; every row must correspond to a unique chronological segment of substantive content, synthesizing all dialogue from multiple speakers where necessary, ensuring that both the segment ordering and all detailed points follow strict chronological progression without exception.

Stay dangerous, #Nostr fam. 🤙

This is EXACTLY why I use local #OpenSource #AI & store my own checkpoints/models. 💁‍♂️

It blows my mind how people who understand #Bitcoin / #BTC aren't aggressively decentralizing & open-sourcing EVERYTHING in their lives; it's not just the money that's centralized & gatekept, it's EVERYTHING. 🤷‍♂️

If you're looking to go local, grab these first & expand from there:

LM Studio ( LLMs / text / code ) ComfyUI ( VLMs / images / video ) Docker ( Agentic AI "teams" via containers )

Not your machine, not your model… 🙉 Not your computer, not your code… 🙊 Not your iterations. not your images… 🙈

…you get the point. 🤙

🚨 BREAKING (AND ENTERING) 🚨

The #anonymous homies over at Anna's Archive just did a digital heist of #Spotify & rugged the entire subscription model for #audio media over the #Christmas / #NewYears holiday & nobody is even discussing the impact of that; this isn't gonna be the last time it happens to #digital #media & you just KNOW they're gonna scrape #Netflix next. 🫠

Meanwhile, the #Indie #artist / #podcast crowd are #streaming & dropping their #music & content on #Nostr services, getting paid (w/zero middle-man) & making more than they ever would on a centralized platform. And they're doing all of this while leaving it entirely to the free market to decide in a meritocratic way. 💁‍♂️

That's SO wild to juxtapose & I honestly down see how ANY the walled gardens of Web 2.0 survive the onslaught of the #anon #OpenSource army in an #AI age. 🤷‍♂️

Shoutout to all the supporters, services & #developers who are making all of this possible for the little guys. 🤙

CC:

nostr:nprofile1qyt8wumn8ghj7etyv4hzumn0wd68ytnvv9hxgtcppemhxue69uhkummn9ekx7mp0qqs8wk0myn8v2m79w4g82n9g7mfvvqvrmgjn0j8nsyy0mu5rkg9qukqahexty nostr:nprofile1qy2hwumn8ghj7etyv4hzumn0wd68ytnvv9hxgqtxwaehxw309anxjmr5v4ezumn0wd68ytnhd9hx2tmwwp6kyvtkx46kv7tgx3kxketnd3nhs7rrvdkxwwrxxp585ct6dpshwdmjwdexsanxw96hs7ndxfnxkd35vvmny6rswv6r2m3swc6n7cnjdaskgcmpwd6r6arjw4jsqgr98zf9a0akv86p3kx86p6tachg4lthsuqamzg8psk6jdk4w8j4cv6qcgkr nostr:nprofile1qyf8wumn8ghj7ur4wfcxcetsv9njuetnqythwumn8ghj7un9d3shjtnxda6kuarpd9hzuendqqstsekww6l9hqnxjkvqyjpj9wxlfsupvz8l5kjmglz082lqmrmk0fgupn0m2 nostr:nprofile1qy8hwumn8ghj7efwdehhxtnvdakqz9thwden5te0v4jx2m3wdehhxarj9ekxzmnyqqsq7gkqd6kpqqngfm7vdr6ks4qwsdpdzcya2z9u6scjcquwvx203dshersuv nostr:nprofile1qy88wumn8ghj7mn0wvhxcmmv9uq3zamnwvaz7tmwdaehgu3wwa5kuef0qqsv73dxhgfk8tt76gf6q788zrfyz9dwwgwfk3aar6l5gk82a76v9fgfngn5a nostr:nprofile1qythwumn8ghj7mn0wd68ytnsv9ex2ar09e6x7amwqyv8wumn8ghj7urjv4kkjatd9ec8y6tdv9kzumn9wsqzq5edsvxllcyuz0n4azc5tjp9wx8uz2cqq0mp6c0fqamjr3llly7tksuz3y nostr:nprofile1qy2hwumn8ghj7etyv4hzumn0wd68ytnvv9hxgqgdwaehxw309ahx7uewd3hkcqpq8m76awca3y37hkvuneavuw6pjj4525fw90necxmadrvjg0sdy6qsmthtls nostr:nprofile1qy2hwumn8ghj7un9d3shjtnyv9kh2uewd9hj7qghwaehxw309aex2mrp0yhxummnw3ezucnpdejz7qpq0r8xl2njyepcw2zwv3a6dyufj4e4ajx86hz6v4ehu4gnpupxxp7s85uvay

https://decrypt.co/353282/shadow-archive-says-it-copied-virtually-all-of-spotifys-music

Well, it finally happened; the local SSD that one of my #Bitcoin / #BTC #Knots nodes is stored on filled up to red zone levels. 😭

I've been doing tons of local #AI LLM/image/video/audio generation experiments lately (running from the same SSD drive) & had NO idea that checkpoint models, diffusion models, LoRAs, text encoders, outputs & installations would take up drive space so aggressively (a single base model alone can EASILY be 5-30GB & you need many different models to accomplish different tasks). 🫠

All of these models need to be run off of an SSD so that performance isn't degraded to the point of non-usability, so it was finally time for a data re-org. 🤷‍♂️

==========

STEPS TAKEN:

1. Close Bitcoin Knots.
2. Copy/paste "Bitcoin" folder from existing SSD → Archival HDD.
3. When the copy is complete, change the name of the old "Bitcoin" folder to "Bitcoin-old" & re-open Knots (so Knots thinks the folder doesn't exist anymore).
4. Choose the new "Bitcoin" directory in the "New installation" prompt & uncheck all boxes to make sure you are still unpruned & verifying all blocks (it will falsely report the disk space available because it assumes the existing data in the "Bitcoin" folder is unrelated data).
5. Knots will then recognize & verify the existing blocks, then sync any new blocks that were found during the downtime.
6. Once you've verified that the migration worked & you're fully synced, delete the "Bitcoin-old" folder on the old drive.

That's it! You're migrated & fully up-to-speed without having to download the entire blockchain all over again. 🥳

IMPORTANT NOTE:

Make sure you are only copying your own verified blockchain so that you are sure that the block data is legitimate. DO NOT grab the blockchain from any other unverified source, as this can lead to incorrect data issues, misallocated funds & double-spend attempts. 🚨

==========

I'm happy to report that after 8 hours of downtime (I forgot how slow HDDs actually were at transferring data since I run most daily tasks off of SSDs), all drives are happy & I will be able to host this complete, unpruned blockchain node for another 5-8 years before having to re-org any data again (by which time it will be completely feasible to cheaply migrate to a completely dedicated SSD drive that will last for another decade or more before filling up). 💁‍♂️

Time to accelerate the AI experimentation into hyperdrive. 🚀

Thank you for coming to my TED talk. 🤙