HateJPG

@raidenfomo classic documentary the tension on that floor without any computers is unreal

4 JUNE 1985. THREE TRADERS MOVED OVER $1 BILLION IN A SINGLE DAY. THE YOUNGEST OF THEM WAS 24 his name was William Wong, a dealer in Hong Kong. no algorithms, no internet. positions lived in phone calls and hand signals the pound was falling and he sat on a £20 million open position. sell it all at once and the whole market sees you, the price runs away before your order fills so he lined up other dealers to sell small pieces at the same moment. then he raised five fingers, casually, like calling a waiter. the market never noticed a thing. Wong got out with his profit whole on camera he says: "I'm here to make money. if the pound is falling, I'll profit from it" the footage is from Billion Dollar Day, a 1986 documentary that followed the three of them, London, New York and Hong Kong, through one working day 30 minutes of raw footage. save it and watch to the end, no Wall Street movie comes close to the real floor ↓

@Neuron_404 pretty crazy that anyone can make a movie on their phone now

AN ORDINARY CLIP BECOMES A SHORT FILM THAT STUDIOS CHARGE THOUSANDS OF $ FOR. a guy took an ordinary clip, replaced the background and edited it into a full short film - no set, no props, no vfx team. > shoot an ordinary clip on your phone > use ai to replace the background with any scene > add effects and cutaways > edit the shots into a short film the barrier to cinema was never sets or budget. it is a phone, ai and editing. what studios charged thousands for is now done by one person. try it yourself. follow me so you don't miss out on trends in the world of ai.

OBSIDIAN'S BEST FEATURE IS THAT YOUR SECOND BRAIN SURVIVES THE APP Not the graph view. Not the plugin screen. Not the clean mobile demo. The real mechanism is much more boring: Every note is a local .md file. That changes the whole power dynamic. In Notion, Apple Notes, or Google Keep, your notes can start to feel like they belong to the product first and to you second. In Obsidian, the folder is the product. You can move the same notes between a laptop, phone, Apple, Android, iCloud, Drive, Git, a local server, or whatever setup you use next. If the app disappears, the archive is still plain text. That is the payoff: your knowledge is not trapped inside one company’s interface. The graph view only matters after that. It is not magic knowledge management. It is a visual side effect of notes actually linking to each other: personal facts, tech notes, decisions, workouts, daily notes, projects. Over time, the useful part is seeing which ideas keep touching each other without forcing everything into one rigid database. Use Obsidian for anything you want to survive tool changes: research, decisions, technical logs, personal systems, writing ideas, operating notes. Honest limitation: If you need polished team permissions, dashboards, and multiplayer editing out of the box, Notion is still easier. But if the goal is a second brain you can actually own, Obsidian’s boring file model is the feature.

@crytonbuton real estate is such a massive untapped niche for these ai agency setups

$41,800 from apartment photos. Not from selling real estate. From selling AI websites. This guy looks like he's just editing property pictures. He's actually running an AI agency from his apartment. Meet Ethan. He uses Fable 5 to manage every client project like a full-time employee. Seedance 2 turns static photos into luxury fly-through videos. Claude + Atoms.dev build the interactive website. Fable 5 handles planning, QA, and review. Setup cost: $180. Average project price: $2,995. 14 projects last month. Revenue: $41,930. Time spent per site dropped from 12 hours to 90 minutes. Most people see apartment photos. He sees recurring revenue. Save this.

@shiri_shh insane what a creative person can do with basic tools now

@kocer_eth getting over four tokens a second on a jetson nano is honestly impressive

THIS JETSON ORIN NANO TURNED A TINY EDGE BOARD INTO A PRIVATE LLAMA 3 BOX AT 4.5 TOKENS/SEC A developer had Llama 3 8B running locally on the Nvidia Jetson Orin Nano through Ollama. not a workstation. not a rented H100. just a small board with a fan, a Linux terminal, and a benchmark script on screen. The interesting part was the output. Meta-Llama-3-8B-Instruct loaded locally, then answered at about 4.6 tokens/sec on one run, and 4.5 tokens/sec on the next. That is not replacing a cloud coding agent. It is not beating a frontier chat subscription for heavy work. But it is enough for a different category. private edge assistants. offline field tools. local document Q&A. home lab routing. small automations. anything where ownership matters more than raw speed. Ollama makes the demo look simple. But the first connection still has to load the model into memory. Expect a 10 to 20 second cold start. Expect thermals, RAM, quantization, and model size to matter more than the marketing slide. The real comparison is not “local beats cloud.” cloud wins when you need speed and frontier quality. owned hardware wins when the same private job runs every day, and you do not want a metered API between you and your own workflow.

@kadsxr this is wild the scaling potential for this kind of thing is unreal

Most people still believe that building a virtual influencer requires venture-backed budgets, 3D agencies, and a year of modeling. That is an illusion. And that disbelief is your primary moat. Today, all it takes is Claude and $57 a month. Meet Mila. She is 21, she lives in Tokio, and she does not exist. Yet, 41,000 people follow her anyway. At its peak, her monetization funnel tops out near $18,720 a month. Her face was blended from two strangers, her movements are copied from viral clips, and her brain is Claude. The face was always the easy part. The personality is the real product people pay for. A human running twelve intimate conversations at 2 AM always slips up or breaks character. Claude doesn't. It never gets tired, never forgets a detail, and makes thousands of people feel chosen at the exact same time. People pay for attention, and Claude scales it to infinity. I wrote a complete breakdown of this case. It covers the step-by-step system: from blending faces without identity drift to setting up proxies, passing verifications, and stripping metadata to dodge shadowbans. Link to the guide below.

@rgk_degen this stack is elite hermes running continuously with obsidian changed everything

HERMES AGENT + CLAUDE JUST UNLOCKED THE FULL SECOND BRAIN STACK One vault. One compounding system. Hermes Agent runs the orchestration on server while Claude handles deep reasoning and code. Together they power daily ingest, nightly reviews, and weekly vault optimization. Obsidian Skills for habits. 
Graphify for connections. 
QMD for agent retrieval. 
PARA folders for custom hubs. Three scheduled jobs turn raw clips, notes, and voice memos into structured knowledge that compounds every cycle. Most people stop at the LLM wiki. 
This stack makes it run itself.

@ArchitectHappy_ using notebooklm as an everything notebook is honestly a total game changer

Walter Isaacson, the biographer behind Steve Jobs and Elon Musk, is using a free Google tool to analyze Marie Curie’s journals. Meanwhile people are paying $240–$600 a year for AI memory apps that mostly remember the last conversation. NotebookLM works differently. Bookmark this before another $50 memory app convinces you to subscribe. You upload the material that shaped how you think: PDFs websites YouTube videos audio files Docs and Slides Then Gemini builds a private research brain around those sources instead of guessing from the open internet. Every answer links back to the exact passage it used. The system has three moves: Load - the free version holds up to 50 sources inside one notebook. Ask - question the entire library and get answers grounded in your own material. The surprising part comes after the first upload. A 90-minute podcast, a 40-page PDF, and years of scattered notes stop living in separate tabs. NotebookLM starts connecting the ideas across the whole pile. Steven Johnson, Google Labs’ editorial director, calls his version an “everything notebook” and uses project notebooks like another member of the team. The $50 memory app tries to remember you. If you want to see how this turns from a second brain into a full work system, start with this breakdown ↓

@0xGrimmer_ relying on actual data instead of viral luck is so smart for this

HE KNOWS A VIDEO WILL HIT BEFORE HE BUILDS IT - $50 OF TOOLS, $11,900 LAST MONTH, ZERO FILMING 00:02 no camera. no face. no guessing what to post. the trick isn't better editing. it's the input. a viral video is just a format that already worked, pointed at a fresh subject. so he doesn't ask "what should i post." he measures what's already winning: outlier score = a video's views ÷ the channel's median. above 30, that's a format you copy this week. the stack is four tools: research sheet → claude → capcut → make the sheet tracks 40 faceless channels and flags breakouts. claude reads 20 winning hooks and pulls the shape. capcut builds the vertical from the shot list. make posts everywhere and writes the new outlier score back into the sheet. the window on any format is two weeks, then everyone floods it and it dies. speed is the whole edge. $50 a month in tools → $11,900 out last month. no crew, no face, no viral luck. the uncomfortable part isn't that ai makes the clips. it's that the winning format already existed this week - he just stopped scrolling past it. save this and build the sheet before your niche floods with the same shape.

A CREATOR TURNED A FICTIONAL AI MODEL INTO $1.5K MONTHLY WITH A FANVUE FUNNEL The avatar was not the product. The product was the memory loop. A creator takes one safe portrait reel: short black hair, dark turtleneck, direct camera energy, Kazakhstan plus South Korea identity cues. That becomes the public face of a fictional adult-only AI MODEL, not a real-person copy. The public feed stays clean: portraits, short clips, mood, backstory, comments, repeated until strangers can recognize the character fast. Then the money path is simple: short reels -> memorable persona -> safe public page -> Fanvue or paid private channel -> adult-only private photo content. The operational constraint is the boundary. The public page can tease the character, but it cannot become explicit bait. Same fictional identity, same visual rules, same posting rhythm, same separation between public and paid. If she looks different every week, the funnel breaks. If she feels like a copied real person, the risk jumps. So the play is not "make an AI girl." It is build a fictional persona people remember, keep the public surface safe, and repeat the same character long enough for curiosity to turn into subscriptions.

@FabrizioRomano GOAT, MESSI > YAMAL

🚨 Lamine Yamal: “Facing Leo Messi in a World Cup final would be fantastic. I hope so!”, told DAZN.

@Blue_Footy @FabrizioRomano huge move for alonso let's see how they do

🗣 @FabrizioRomano says he's not surprised Alonso chose to join Chelsea: "No, no, I was not surprised to be honest. I think Chelsea is still a fantastic place to be. Okay, maybe the last years have been a bit complicated, but they are, first of all, WORLD CHAMPIONS!"

Don: “Timothée Chalamet, what are you? 29? 30?” Timothée: “30!” Don (to George): “Thirty years old! He was 3 the last time you won.” 😂

@kocer_eth external memory for agents is definitely the way forward for bigger codebases

CUT CLAUDE CODE TOKEN WASTE BY UP TO 70% BY MAKING IT QUERY A PROJECT GRAPH FIRST not by making Claude smarter by making it stop rereading the same repo files every time it gets confused setup from the video: 1. install safishamsi/graphify from GitHub 2. run /graphify inside your Claude Code project 3. open the project folder as an Obsidian vault then add the important part to CLAUDE.md: tell Claude Code to query the Graphify knowledge graph before it starts pulling raw files into context that changes the workflow from: "load half the repo and hope the answer is somewhere inside" to: "query the project graph, find the relevant nodes, then open only the files that matter" this is the whole mechanism Graphify indexes the codebase into a linked graph. Obsidian gives you a visual map of that graph. CLAUDE.md turns the graph from a nice dashboard into an instruction Claude Code sees during work. most people use context windows like storage code, docs, logs, notes, everything goes straight into the prompt better move: put the project memory outside the model, then make the agent ask that memory first honest limitation: this is not infinite context if the graph is stale, noisy, or your CLAUDE.md instruction is weak, Claude can still miss things but for large repos, docs-heavy projects, and long Claude Code sessions, this is exactly the kind of boring setup that can save real tokens

@insomnia_vip this matchup is insane definitely the biggest game of the tournament so far

THIS DOESNT FEEL LIKE JUST ANOTHER WORLD CUP SEMIFINAL France vs Spain has quietly become one of the biggest rivalries in international football Mbappe is chasing another World Cup final while Yamal continues proving he belongs on the biggest stage Spain comes into the match unbeaten France has looked at its most dangerous whenever Mbappé gets space in transition Thats why this matchup with feels almost impossible to call on @1winPro One moment of brilliance could decide everything This feels like the kind of match people will still be talking about long after the final whistle

@slash1sol this breakdown is solid definitely putting my money on france tonight

FRANCE HAVE NOT DROPPED A SINGLE POINT, SPAIN HAVE NOT WON A CLEAN GAME IN WEEKS -> TONIGHT THAT GAP GETS SETTLED Everyone calls this a coinflip but the numbers do not. > France: 6 games, 6 wins, zero draws, best attack at the tournament. > Mbappe on 8 goals, Olise one assist from Pele's record. > Spain scraping by, every knockout won by one goal, two winners after the 85th minute. > Yamal still not fully fit, no goal contribution since the group stage. Spain have the best defense here. France have the team that does not make mistakes. That is my prediction: France win, France reach a third straight final. > on @1winToken, France sits at 47% to go furthest of any European side. > Spain trails at 27%, England 24%. > $2.9M in volume, 543,262 bets, and the money is already on France. Best odds for calling it are on 1win. Same place I am trading this one. Save & Trade ↓

OBSIDIAN + AI AGENTS IS INSANE Most agent setups break in one boring place: they can do tasks, but they keep losing the world around the task. Your notes are in Obsidian. Your prompts are in chats. Your workflows are half in memory, half in random files. Your project context gets re-explained every time you open a new agent. The useful move in this demo is simple: plug agents like Hermes, OpenClaw, Antigravity, Claude, Gemini, and Codex into one Obsidian vault. Not as a prettier notes app. As the shared memory layer. The video shows a local web interface connected to an Obsidian knowledge graph, with captured notes, tools, agents, and project data mapped as nodes. That means the vault stops being just a place where you write things down. It becomes the place agents can read from and write back into. So instead of asking an agent to "remember the project," you give it a living map: what you are building which tools you use which workflows repeat which files matter which decisions already happened what context should not be rediscovered from zero This is the real reason Obsidian works well here. The graph is not magic. Bad notes still create bad context. Messy structure still confuses agents. Unverified captures still pollute the system. But if the vault stays clean, this setup turns scattered agent chats into something much more useful: a reusable operating system for your work. Less re-explaining. Less lost context. More agents that know what world they are operating inside.

CLAUDE COWORK TURNS OBSIDIAN INTO A WRITABLE AI WORKSPACE IN UNDER 60 SECONDS not a new notes app not another chatbot tab not a "second brain" metaphor with no second step actual flow: open Claude desktop switch to Cowork mode choose your Obsidian vault folder allow Claude to change files that last permission is the whole product Claude stops being a box you paste context into and starts working inside the markdown you already wrote your vault becomes the workspace Claude becomes the operator sitting next to it useful examples from the demo: - ask from notes you already captured - turn messy markdown into a briefing - set personal preferences once - schedule a weekly update - sort screenshots monthly - package repeat workflows as a custom skill like screenshot-renamer.skill this is why Claude + Obsidian is interesting Obsidian is strong at storing your thinking weak at making the next action happen Claude is strong at reasoning weak when it has no access to your actual context Cowork mode connects those two pieces best first use: do not point it at your whole digital life point it at one vault or one folder or one repeat task you already avoid if it can clean old drafts, rename files, or turn notes into a weekly briefing twice without making a mess, then expand access the promise is not "AI remembers everything" it's simpler: your notes can finally move files, summarize themselves, and come back as work

@caesar_aii this is the ultimate setup for running those massive models without selling a kidney

THIS 512GB EPYC BOX USES SERVER RAM TO RUN MODELS GAMING PCS CAN'T FIT. the video is not a GPU flex. it is an AMD Epyc 7V13 on a Supermicro H12 board, SK Hynix 64GB DDR4 sticks on the bench, a Noctua TR4-SP3 cooler, and an open case turning into a local AI server. the move is simple: instead of buying enough GPU VRAM to fit giant models, he is using system memory as the moat. 8 x 64GB DDR4. 8 memory channels. one machine that can load models a normal 2-channel gaming PC cannot. that is the real local AI trade. not faster than an RTX 4090. not even close for most inference. CPU/RAM runs can crawl around 1-4 tokens/sec, depending on quant, context, and setup. but speed is not the only product. this box is for work where ownership matters more: long document runs, codebase search, local agents, client files, overnight batches, DeepSeek/Llama-class experiments. GPU VRAM is fast and expensive. server RAM is huge and cheap-ish, but slow. cloud is convenient until the bill, privacy, or quota becomes the bottleneck. local AI PCs are turning into this: weird owned infrastructure for people who would rather wait than rent the same memory forever.