Wemoo

THIS HARDWARE STARTUP JUST BARELY TOUCHED THE FUTURE: A PHYSICAL HOLOGRAPHIC LENS THAT LETS YOU LITERALLY SLICE THROUGH A 3D BRAIN SCAN IN REAL-TIME. Most medical visualization tools force you to rotate complex 3D assets with a clunky mouse or uncomfortable VR headsets. This engineering team built an analog interactive layer that acts as a physical slicing tool. The execution looks straight out of a sci-fi movie: as the user moves a tracked glass frame back and forth, the screen's rendering engine instantly projects the corresponding depth layer of the neurological scan directly onto the glass. The architecture completely bridges the gap between digital data depth and real-world spatial positioning: → Zero-Latency Volumetric Slicing: Operators can intuitively slice through complex data arrays without clicking through infinite UI menus. → Physical Depth Feedback: By mapping the monitor's projection to the tracking points, the asset maintains perfect parallax alignment from any angle. → Hardware Interface Revolution: The system removes the abstraction of digital tracking models, allowing natural human interaction with 3D space. Stop limiting multi-dimensional data models to flat, dead 2D screens. The future of spatial analytics requires physical depth control. Save this.

@thesupermanmx This is a cherry-picked research demo. The second you try running this open-source pipeline on regular video footage with fast motion blur or complex reflections, the tracking will break and your 3D mesh will look like a distorted glitch.

China open-sourced a model that reconstructs any scene in 3D from a regular video, in real-time. one camera. no LiDAR. 10,000+ frames without falling apart. just walk around with your camera and watch the entire world get rebuilt in 3D at 20 fps. → runs at ~20 FPS on a single GPU → Stable over 10,000+ frames → Beats optimization-based methods on benchmarks → Works on drone footage, driving videos, indoor walkthroughs 100% open source.

This looks incredible on a visual graph view, but a massive web of 4,000 interlinked nodes is functionally unreadable for daily work. The second you try to navigate that massive spiderweb, you'll realize a clean nested folder and a basic search query are ten times faster than clicking through random floating dots.

A 29-YEAR-OLD SYSTEMS ARCHITECT FED CLAUDE 9 YEARS OF UNSTRUCTURED NOTES. IT BUILT A KNOWLEDGE GRAPH THAT HOLDS 4,102 FILES AND NEARLY A DECADE OF MEMORY. Not a messy folder. Not a graveyard of text files. Something else. He took the full scope of his digital life — daily logs, system designs, deep research — and mapped it as a graph instead of a flat directory. Every thought sits as a node. Every piece of context links back to the exact idea it supports. The detail most people miss: it is not just organized, it is automated. The semantic connections were calculated across the whole graph by an AI, not manually tagged by a human. Most people hoard notes in a folder and hope they eventually remember to read them. This treats human memory like a dependency graph. You can see exactly which thought connects to the next one, and exactly how an idea from five years ago shapes what he is building today. That is the real unlock. A second brain stops being a pile of pages and starts being a system you can actually see the shape of.

THIEVES ARE USING WI-FI JAMMERS TO EVADE DETECTION FROM HOME SECURITY CAMERAS… …BUT THE NEXT GENERATION OF INFRASTRUCTURE IS ALREADY CUTTING THE SIGNAL. Most homeowners think their wireless security setup makes them untouchable. A group of tech-savvy intruders just proved that a $20 signal jammer from the dark web can blind your entire cloud-connected camera system in under 3 seconds. The footage shows real-world execution: as the intruders approach a suburban home, their portable RF jammer floods the 2.4GHz and 5GHz bands, causing the wireless security feeds to instantly glitch, pixelate, and disconnect. By the time the system attempts to re-authenticate, the data stream is compromised. While basic consumers are left completely vulnerable to network signal interference, advanced infrastructure architects are already looking at decentralization and physical failsafes. If your security design depends entirely on standard Wi-Fi, you are running a critical single point of failure. The immediate systemic defenses against wireless jamming are straightforward: Ethernet Overrides (PoE): Transitioning critical perimeter monitoring nodes to hardwired Power over Ethernet cables completely neutralizes RF jamming attempts. Localized Storage Arrays: Deploying cameras with local micro-SD storage ensures data captures continuously, even if the wireless link to the router is broken. Cellular and Mesh Backups: Incorporating secondary, multi-band communication channels that trigger an immediate localized alert the moment a Wi-Fi signal drop is detected. Stop relying on basic cloud subscriptions and consumer-grade wireless networks to guard your perimeter. Real security requires absolute structural independence. Save this.

@0xkkai This architecture completely overpromises on the autonomous maintenance loop. Running a local cron job that forces an LLM to sweep a massive markdown vault every single morning will quickly bottleneck your hardware and quietly run up a massive API token bill while you sleep.

ONE OF CLAUDE'S DEVELOPERS SHOWED HIS SETUP. HE DOESN'T USE CLAUDE THE WAY YOU DO he didn't post this on twitter. didn't make a thread. just showed his setup on a call and someone recorded it and when i saw what was inside, i understood why he never starts a conversation with claude from scratch the first file in his vault is called CLAUDE.md. it's not a note. it's a full profile: who he is, how he thinks, where he gets stuck, what projects he's running, how he wants to be talked to. claude loads this file automatically every session meaning the ai doesn't ask "how can i help?" he opens a conversation and claude is already in context. knows his goals. remembers what he postponed two months ago. sees where he's contradicting himself and that's just the beginning every project is its own folder with a clean structure: inputs, process, outputs, feedback. when he works on something, he opens only that folder. claude sees one project, not the chaos of his entire life anything he does more than once is saved as a skill. email a client, break down a call, prep a document. one line and claude does it his way and at 7am every day claude walks through the entire vault on its own. files new stuff. links it. flags what's gone stale. writes him 3 lines: what changed overnight he wakes up and the brain already worked and here's what got me the most: it's all just text files. no cloud. no lock-in. if a better model drops tomorrow, he points it at the same folder and everything keeps working he basically uploaded himself into a file system. and now ai doesn't guess who he is. ai knows most people use claude like google with manners. this guy built himself an external memory that grows every day i break down finds like this every day - follow so you don't miss the next one

@0xObssnnn This math completely ignores real enterprise overhead. You are freezing your cloud renewal for a $97,000 box, but the moment you factor in specialized local cooling infrastructure, high-voltage facility power lines, and human server maintenance, your "savings" evaporate.

Nvidia put a $97,000 computer on a desk and the cloud math stopped working. A friend who signs GPU invoices at a fintech, around $60K a month, sent me this breakdown with one line: "we froze the AWS renewal." Jensen announced the box in front of 30,000 people in Taipei. Datacenter memory, desk footprint. The rental comparison is brutal. At AWS rates for equivalent compute, the box pays for itself in 2 to 3 months. Then it keeps running on your power bill instead of your cloud bill. But the reveal is in the fine print. The trillion-parameter claim on Nvidia's slide has an asterisk, and the presenter reads it out loud around the 6-minute mark. There's also 1 thing missing from the spec sheet, near minute 4. Every $500 PC has it. This box doesn't. He ends with which machine he'd actually buy. It's not this one. Renting compute was a phase. Nvidia just named the end date.

This approach will absolutely obliterate your wallet on high-volume context-window costs. Forcing a local CLI developer agent to crawl through 8,294 full text files for a minor weekly link audit will hit your Anthropic token rate limits within minutes and cost you hundreds of dollars in API calls.

Karpathy method + Obsidian + Claude Code = a second brain that forgets nothing. Every second brain dies the same way you save, you tag, you never come back. 8,294 notes and 11.3GB of highlights turn into a graveyard with good folder structure. The fix came from 1 Karpathy line: you understand nothing until you can build it from scratch. Reading is not learning and storage is not memory. So flip the frame. Obsidian is the codebase, Claude Code is the engineer, and your notes are legacy code waiting for a refactor. Put the engineer inside the vault open a terminal in your Obsidian folder and run Claude Code. It reads every note as plain markdown, with no plugins, no exports and no API glue. Give it a constitution 1 CLAUDEmd file with 30 lines: your naming rules, your folder logic and what you’re building toward. Every session boots with your brain’s operating manual. Then run the Karpathy loop. After reading anything, don’t summarize it rebuild the idea from memory in a fresh note, then 1 command: check my version against the source and mark what I got wrong. The gap between your version and the real one is the part your brain keeps. Every Sunday, 1 prompt: find orphan notes, merge duplicates, link what belongs together and flag what I never touch. 15 minutes, and the vault compounds instead of rotting. And ask it real questions. Not “what did I save about agents” but “what do my March notes claim about agents, and where do they contradict each other.” It cites your own files back at you, so notes from 2024 start answering questions from 2026. 8,294 files went from dead weight to a working memory in 1 weekend of setup. You never needed more notes you needed a librarian who never sleeps.

using a single file avoids folder sprawl but creates a massive merge conflict nightmare if you sync across multiple devices. without a git-backed history pipeline or an automated timestamp parsing script, the agent will inevitably choke on formatting syntax when rewriting your top-level headers.

1 Neural Network + Obsidian + Karpathy’s 1-file method = the most unhinged second brain build of 2026. It remembers everything you’ve ever done, and it costs $0 on top of what you already pay. The base is Karpathy’s append and review: 1 giant note, new thoughts stack on top, old ones sink, every few days you reread and pull the survivors back up. No folders, no tags, no plugins the rereading IS the system, because review is what turns storage into thinking. The flaw: past 10,000 lines, no human rereads anything. That’s where the neural network takes over. You keep the note in Obsidian 1 vault, everything dumps to the top: ideas, links, meeting fragments, half-thoughts. You never organize, you only dump. It all lives as plain markdown on your own disk, and that detail is the whole trick. Because now you point Claude Code at the vault folder, and it reads every line you’ve ever written. “What did I think about pricing in March.” “Find the 3 ideas I keep circling.” “What did I drop that deserves a second look.” It answers from YOUR notes, with quotes, in 15 seconds. Then once a week, 1 prompt closes the loop: read the last 7 days, surface the 5 entries worth pulling back up, flag anything that contradicts what I wrote a month ago. The model does the sinking and surfacing Karpathy did by hand, and the note stays alive instead of turning into a graveyard. Week 1 feels like nothing. Week 4 you hit the first “I already solved this in January.” Month 3 you consult your past self more than Google. Most second brains die in 11 days under 40 plugins and 200 folders. This one is 1 file and a loop, and it compounds because dumping takes 0 discipline. Notion stores what you thought. This thing argues back.

@KanikaBK This is a total exaggeration of the onboarding timeline. You are not going to OCR, format, tag, and link twenty-four years of messy, handwritten machine shop notebooks into a functional digital vault over a single Saturday afternoon without hiring a data-entry department.

A 58-year-old machinist put 24 years of paper notebooks into Obsidian over one weekend and.... the Obsidian graph that came out stopped his daughter mid-sentence when she saw it. He searched a 2009 repair problem and got the answer in two seconds with three linked notes he had never connected on paper. I know this is a crazy story! My article covers the 12 Obsidian tricks that make that kind of organization feel effortless for normal people. Full guide below.

@chesny This setup is a total administrative trap. The second your folder expands past a few dozen raw sources, your agent will waste your entire API budget crawling local markdown files and checking links for simple queries you could have found with a basic text search.

Andrej Karpathy, el CEO de Obsidian y Claude Code acaban de construir el segundo cerebro más inteligente de la Tierra. Todo empezó con un gist de una página que leyeron 21 millones de personas. Karpathy le da la vuelta a todo lo que sabes sobre tomar notas: Obsidian es el IDE, Claude Code es el programador y tus notas son el código fuente. No le haces preguntas a la IA que olvidará para mañana, haces que mantenga una wiki viva. Tres comandos ejecutan todo el sistema: Ingest: sueltas un artículo, un podcast o un PDF, y Claude lo divide en páginas atómicas enlazadas a todo lo que ya sabes. Query: preguntas cualquier cosa y responde desde tus propias notas, con tus propias palabras, citando tus propias páginas en lugar de adivinar a partir de datos de entrenamiento. Lint: una vez a la semana, Claude recorre todo el vault, marca contradicciones, elimina afirmaciones obsoletas y vuelve a conectar las notas huérfanas. Entonces Steph Ango hizo su movimiento. El CEO de Obsidian no añadió un simple botón de "Preguntar a la IA" a la aplicación, sino que publicó 5 archivos de habilidades que le enseñan a Claude a escribir en el lenguaje nativo de Obsidian: wikilinks, Canvas, Bases y la CLI. El repositorio superó las 13 900 estrellas en semanas y ahora ya tiene 41 000. Karpathy lo ejecuta con sus propias lecturas: 100 artículos y 400 000 palabras, con enlaces cruzados y mantenidos mientras duerme. Sin base de datos vectorial, sin embeddings, sin aplicaciones de memoria de $20 al mes, solo una carpeta de markdown plano y un agente que nunca se cansa de la parte aburrida: enlazar, archivar y realizar el mantenimiento que ha arruinado a todos los Zettelkasten desde 1965. Tu vault tiene 3000 notas que nadie volverá a abrir jamás. El suyo se ha leído a sí mismo por completo antes del desayuno. Todas las aplicaciones prometían un segundo cerebro, pero este es el primero que piensa.

@semichenkko This setup is an absolute thermal trap. Running a continuous local inference model on a desktop Mac Mini stuffed inside a sealed nylon backpack will choke the system's fans, cause severe thermal throttling within twenty minutes, and permanently degrade your hardware.

This guy from China is making $12,500 a month just by walking around major crypto conferences with a backpack. 🤯 He hooked up a Mac Mini running OpenClaw to a compact CUKTECH 10 Mini battery. But why would anyone need a portable AI server? At massive events, the Wi-Fi is always terrible and power outlets are nonexistent. While other creators wait until the evening to reach their hotel and upload data to the cloud, his pocket AI works entirely locally. It processes keynote audio on the go, extracts alpha insights, and instantly posts viral threads on X. Total monopoly on speed and content generation without being tethered to a wall. How do you build this exact pocket-sized money printer? Breakdown below 👇

ARGENTINAAAAA 🇦🇷

@Dexonfxf This is a recipe for complete operational stagnation. You are going to freeze your hiring pipeline while managers waste months trying to force half-baked AI agents to do complex human jobs, only to end up with broken code and zero productivity.

SHOPIFY JUST CHANGED HIRING FOREVER - BEFORE YOU HIRE A HUMAN, YOU MUST PROVE AI CAN’T DO THE JOB Imagine applying for a job… But your biggest competitor isn’t another person. It’s AI. That’s exactly the direction Shopify is taking. CEO Tobi Lütke introduced a new rule across the company: Every team must prove that AI cannot complete the work before they’re allowed to hire another employee. Let that sink in. This isn’t about replacing everyone with AI. It’s about changing the default. Instead of asking: “Who should we hire?” Companies will now ask: “Can AI handle this first?” And in many cases… The answer is yes. Writing. Research. Coding. Customer support. Marketing. Data analysis. Internal operations. AI can already complete many of these tasks 24/7, at a fraction of the cost, while continuously improving. This doesn’t mean human talent becomes less valuable. It means the value shifts. The most valuable people won’t be those who simply execute tasks. They’ll be the ones who know how to use AI to multiply their output. Shopify is one of the biggest tech companies in the world. When companies of this size change how they hire, others usually follow. This isn’t just another AI update. It could become the blueprint for how companies build teams over the next decade. The biggest career skill of the next 5 years might not be coding. It might simply be knowing how to work with AI better than everyone else. If you want to stay ahead of where AI is taking business, startups, and careers… Follow @Dexonfxf for real AI stories, business cases, and the biggest AI opportunities every day 🚀

@cyrilXBT This is a recipe for massive context-window inflation. The moment you ask a simple query, your local agent is going to parse dozen of interlinked Markdown files, running up a massive Anthropic API bill for information you could have easily found with a basic desktop search.

I genuinely don't understand why everyone isn't using this yet. Andrej Karpathy, OpenAI co-founder, posted a simple idea that went massively viral: Stop using AI to write code. Use it to build a second brain. You point Claude Code at a folder. Drop in any source: an article, a transcript, a PDF. Claude reads it, links it, files it into a living wiki of everything you know. It compounds like interest. The more you feed it, the smarter it gets. Here's the whole thing: 1) Install Obsidian 2) Create a vault 3) Open it in Claude Code 4) Paste Karpathy's wiki idea and tell Claude to build it 5) Claude makes three folders: - raw (for sources) - wiki (for its pages) - CLAUDE. md (that runs it) 6) Drop any source into raw and say: "ingest this" 7) Ask questions across everything, forever Five minutes to set up and you never start from a blank chat again. Full step by step guide below.

@polydao This is just a rehashed version of basic software cron jobs wrapped in new AI jargon. You are going to waste hours trying to construct these delicate loops, only for a minor API structural shift to completely break your automation pipeline and drain your API budget in minutes.

THE LOOP ENGINEERING SETUP THAT LETS CLAUDE RUN YOUR WORK FOR HOURS WITHOUT YOU while you babysit it prompt by prompt, other people set the loop up once and let Claude plan, act, verify, and fix its own work with no one at the keyboard that gap is the whole skill of loop engineering here are the 3 resources that actually teach it, in the order i'd read them: > Claude's "Getting started with loops" - the cleanest entry point, straight from the team that built the tooling x.com/ClaudeDevs/sta… > the 14-step roadmap from prompter to loop designer - the practical progression once the idea clicks x.com/0xCodez/status… > the Loop Engineering orange book - the deep conceptual breakdown for the full mental model github.com/alchaincyf/loo… learn this and the work that eats your whole day starts running without you save this 👇

@Roman9078963816 A single RTX 5090 with 64GB of system RAM cannot run a 200B+ parameter model at usable token speeds. The model will instantly overflow your GPU's VRAM, offload layers to your system DDR5 RAM, and throttle your generations to a painful one token per second.

HE SPENT $4,000 ON ONE PC… …AND STOPPED RENTING AI COMPUTE FOREVER. Most developers still pay every month to run AI. He decided to own the hardware instead. Here’s what’s inside: → AMD Ryzen 7 9800X3D → RTX 5090 → 64GB DDR5 → 4TB Samsung 9100 Pro SSD → Silent liquid cooling → Built for 24/7 AI workloads What does it unlock? • Run 200B+ open models locally • Build AI agents without API limits • Fine-tune models on your own machine • Keep sensitive data off the cloud • Zero hourly GPU rental fees The hardest part wasn’t the price. It was the build. Cable routing. Cooling. Thermal paste. Hours of testing before everything worked perfectly. Now the machine quietly runs in the background: AI agents. Coding assistants. Private RAG. Automation workflows. While others keep paying monthly subscriptions… …he owns the infrastructure. The next AI advantage won’t come from a better model. It will come from owning the compute that runs it. Bookmark this before local AI becomes the new standard.

MILITARY POLICE JUST RAIDED A RESIDENTIAL APARTMENT AND UNCOVERED A MASSIVE YOUTUBE VIEW FARM WITH HUNDREDS OF PHONES. While you are spending thousands of dollars trying to optimize your SEO and praying for the algorithm to notice your content, underground operations are quiet-shipping industrial-grade view loops. During a recent raid, military police uncovered an underground view-farming operation running directly out of a residential house. The setup is pure physical execution: - Entire walls of actual, physical mobile devices. - Direct cellular IP distribution to simulate natural, geographic human access. - Automated scripts playing music videos, liking, and commenting on loop. They aren't using software emulators. Emulators get flagged instantly. Instead, they use real silicon, real batteries, and real touch controllers to trick the world's most advanced behavioral bot-detection engines. Look at the scale of the operation: hundreds of screens perfectly synced to boost views for music artists who want to look like overnight sensations. Stop relying on organic luck. Understand the physical infrastructure behind the numbers. Save this.

@Hrundel75 The math is only half the battle. If your C++ knowledge is limited to basic syntax and you don't understand cache locality, memory alignment, or template metaprogramming, your code will fail the performance benchmarks in the very first technical round.

the finance kids everyone respected in college made $130k at morgan stanley the quiet physics kids nobody paid attention to made $700k at firms most people have never heard of jane street. citadel. two sigma. hudson river trading they don't hire from career fairs. they recruit from math competitions and algorithmic programming contests the screening is 4 areas: combinatorics, coding, quantitative reasoning, strategic thinking same problem types recycled every cycle, same format every hiring round the preparation path is fully public, costs nothing, and takes 6-9 months a kid from a no-name school who grinds this for 9 months has outperformed ivy league grads who walked in expecting their resume to carry them - pedigree matters less than the problem-solving ability most people spend 4 years networking their way into banking for $130k the people who looked in a different direction spent 9 months and landed 5x the compensation the resources to do this have been publicly available for years save this and start training your mind the difference isn't intelligence it's that nobody showed you where the real opportunity was

THIS DEVS FROM SHENZHEN BUILT A PHYSICAL PHONE FARM IN HIS APARTMENT. HE ACCIDENTALLY LEAKED HIS LIVE CRYPTO CLICKER BALANCE OF $642,180. While everyone is manually tapping their screens like monkeys for $5 a week, he is quiet-shipping a hardware clicker empire from his bedroom. He didn't build a software emulator. Emulators get instantly flagged and banned. Instead, he built a physical wall of 80 real Android devices, running native touch events on real cellular IPs. The setup is pure industrial-grade bypass: Zero Emulator Flags: The apps run on actual physical silicon. The platforms see zero botting signatures. The Master Mirror: One swipe on his laptop mirrors the exact touch coordinates across 80 screens in microsecond sync. Continuous Air-Drops: While he sleeps, his scripts farm daily check-ins, auto-swipes, and clicker quests across 480 accounts. Look closely at the screen right next to his laptop. It reveals the real story: $642,180 cleared and fully verified. Stop tapping your screen. Build the hardware loop. Save this.

@z0rynx This is a prepper fantasy. You are going to spend $2,600 on loud enterprise hardware that heats up your bedroom just so you can search Wikipedia offline during a rare 6-hour power outage instead of just using your cellular data.

A guy on Reddit spent $2,600 on a server rack so the internet can die without him noticing. The rack sits in his bedroom. Netgear switch, 4 yellow cables, a small panel glowing green in the dark. Inside it: a 70B model running fully offline. Next to it: all 6.9M Wikipedia articles 24GB through Kiwix, searchable in 0.3 seconds, no connection needed. The model runs at 12 tokens a second. Slower than ChatGPT. But it answers at 3 AM when the ISP is down. It answers when the API changes terms. It answers when the $20 subscription becomes $200. Operating cost: 380 watts. About $34 a month in electricity. That's the full price of a private AI plus the sum of human knowledge. Here's the part people miss. Every prompt you send to a cloud model gets logged. His prompts never leave the room. Medical, legal, financial ask anything, and the only witness is the plant behind the rack. People called it overkill. Then a regional outage killed internet for 6 hours, and the same thread filled with 200 comments asking for his parts list. The cloud is just someone else's computer. His is 3 feet from the bed.