OLK ≡BushiAGI @olkthink

@paulg decent aura🦦fire for any aspirational musicians; but an authentic Italian HQ is rizzling people🍝

@Entrepreneur Chat, farm your aura🦥start looksmaxxing🦥speak like a philosophy chad alpha-6.7🦥no more skibidi ohio L

yo Chat, it’s no cap, but a hype-mixed 🧒 A real Meta’s new alpha app in Japanese Market is Threads (look at the ranking, kinda consistent trend)☝️Plus, Sam aka ChatGPT has been the most highkey influential LLM in Japan. A recent study says 78.8% of Japanese Gen Z students use ∧ prefer ChatGPT🦥🍵Lowkey Gemini is gaining popularity here fr

@CellCellPress 🦦 Early gut microbiome access hits different…is that haematopoietic stem cell-maxxing?

@paulg 150M USD ARR reflects its scale. It's not cash collected. Faster scaling itself doesn't mean it's worth investing in🦦My inner Buffett isn't convinced rn. But you can save hair-drying time with a faster Dyson ∵ it's physics🧒 *bought additional Coca-Cola stocks

@MIT Mais je suis un futur Français

@alexandr_wang authentic geek IQ test here, bro🦦🍰

Corp AI productivity gap: Need training from a peer-reviewed truth Yo🧒fam. SV uncs are smoking Zen incense believing it’s a productivity hack. But enhanced productivity in your corp comes from the way you guide your employees in this Pre-AGI time. A peer-reviewed truth is here☝️ In the study, educated workers show caution in using ChatGPT. That’s not ∵ Sam is bad💀but ∵ they lack basic technical fluency; e.g., prompt engineering, how to extract meaningful output, how to discern hallucinated outputs, etc. Tho the famous Jensen Huang’s claim "Everyone is a programmer" explicitly states high approachability to LLMs among non-tech nodes🦦But educated workers still seem to need firmware updates to enhance your corp’s productivity. One plausible explanation of the gap in AI-enhanced productivity among industries is employees’ general AI literacy ∧ prompt engineering skills. Lvl 1. An employee needs to understand each task ∧ required criteria Lvl 2. They need to know an optimal process to reach their task-oriented goal Lvl 3. They need to utilise AI tools for faster output Lvl 4. They should be able to discern what cannot be done by AI tools to push the outcome quality If any employees meet the Lvl 4☝️🧒they’re productivity brute-forcers; real assets for your corp. But the study spots many educated workers sit between Lvl 2-3. This isn’t a transitional phase for workers but for corps. This AI productivity race necessarily causes industrial natural selection, removing maladaptive players. You’ve already seen dead PhD credentials in the labour market. The next natural selection is already happening among industrial layers🦦☕️Need apex prompt engineering for your workers? DM me, sir

@WSJ 1. Solve suffering by delaying ageing☝️🧒AGI is an additional turbo brain. Many unhappy problems can be solved with humans ∧ LLMs + incoming JEPA☝️I like Sir Demis. I'm gonna use his AGI🦦🍵But ageing takes the peak 1st priority ∵ you decay.

@clankenluben My voiceover + my music (Zero AI)🦦🍵

Octavian (aka Zuck) updated his aura🦦☕️🧒Tim ∧ a new Apple CEO’s Steve Jobs cosplay is now dead meta, showing a funeral outfit vibe. So Zuck has shifted to more GenZ-approachable outfits. But many problems are present☝️ 👍: Zuck Octavian actively learns how to physically present himself by himself. He bypasses a corporate PR team, rather disliking those mandate-nerfed bureaucratic visuals. 🛑: Older GenZ’s taste is already unc-coded. Zuck’s visual presentation is showing mixed millennial rizzful dad ∧ older GenZ chill bro vibes, which is much better than Timmy’s funeral outfits. What worked for Steve is what never works for others🦦 so Zuck’s new orientation is valid. 👍: This SS is from a telephoto video uploaded by Muse. Telephoto photography ∧ videography give you an accurate image of a subject. He has an actor-tier head-to-body ratio, which is W fr🔥 🛑: The current meta eyewear impairs eye contact, which is crucial to nonverbal communication. Homo sapiens heavily rely on facial nonverbal signals to communicate (≈ 50%) depending on gender ∧ other individual factors. 🛑: "BUILDING IS MY LOVE LANGUAGE" this text on a T-shirt highlights two catastrophic problems. 1st, it overpowers Zuck’s facial presence with aesthetic but lengthy bold placement. 2nd, screaming a personal statement is what Millennials ∧ GenZ uncs have loved to do. Fr🦦 this social behavioral pattern is now a dead meta. People today craft a private mood board, communicate with their sim-cluster (iykyk) ∴ a bold general personal statement has zero aura weight. I re-designed his T-shirt. To be a cultic charisma in the frontier iykyk wearable tech is a move. 👍: Text ∧ pants colors reflect 2026’s earthy authoritative color palette. Zuck W here. 🛑: Fits are lowkey concerning. Bulky shirt, pants, ∧ dad shoes create a massive visual mass under his neck. But his facial presence must be prioritized to interact with his audience☝️ 🦦: ah…huh? Why Zuck Octavian wears a dead Balenciaga dad sneakers meta. One piece of advice to adults is that what Balenciaga made viral is a small hurricane that never joins the legacy standard. Products look fire when Balenciaga launches; but the stuff does not hold long-lasting rizz like the iconic industrial subversive visual counterculture Saint Laurent has created (a brand that belongs to the same company Kering S.A.). But his presence is much more rizzful ∧ living compared to Funeral Suits by AAPL🐆rate 7/10💨

Unlike GPT, Gemini, Claude, I observed Muse explicitly ask for context for each ambiguous prompt. This allows Muse to employ short CoT, which exhibits lower hallucination rates than lengthy chains due to reduced compounding drift, irrelevant token noise, ∧ over-reliance on language priors☝️🦦

Agentic AI causes declining wages, employing automation fr yo chat🧒enjoying pre-AGI time chaos? me? I love the vibe🔥But AGI isn’t a sole trigger of human labour’s obsolescence in multiple fields. Let’s update your firmware rn☝️ A false premise: AGI will cause catastrophic layoffs in the near future. Correction by labour functioning ∧ underlying mechanisms: The degree of AI automation negatively affects wages, causing the abundance of labour. tl;dr: AGI is not necessary to spawn the AI dystopia society. 1. LLMs can sufficiently reduce the demand of human labour when an advanced actor model is combined. Due to cognitive rigidity, LLMs are unable to adopt complex work flow in mega corps. But, bro, Agentic LLMs can meaningfully reduce human employee counts in small-medium business cases. Imagine this ongoing case in the 2020s: Howley Goodman & Family Law now can optimise their workflow, using a legal LLM. Howley gives one prompt to the AI once a week, not a junior legal assistant Joe, who was fired a few months ago. Howley is quite satisfied with the better output quality by his AI ∧ reduced labour costs. Now his son goes to the cello lesson weekly, with money saved from firing Joe. Howley family’s wellbeing has increased🦦❤️ 2. Advanced robotics + JEPA world model automates physical labour Accumulated technical developments, you can expect many blue-collar jobs are gonna disappear. Before 2040s, thanks to advanced robotics, you’re likely to see Japanese Damascus steel knife by tosho (means a professional Katana smith) produced in China. Imagine this emerging case in the 2030s: You go to Costco with your wife ∧ kids once a month. Your wife is looking for a new blender, saying "it’s low-key pricey." One staff member heard ∧ responded to her. "This new Ninja Lite has the same blades ∧ 70 bucks cheaper. The outcome is comparable. But lacking app support." He, no, a Costco robo went back to his job, stacking the boxes of Italian Espresso machines. "Honey, do you believe the machine? I feel I’m gonna buy Ninja Maxx." You wondered if you’re the next to be replaced👨... This pre-AGI time is massively altering human behaviour through reconstructing pre-existing economies. Need more strategies? DM me, sir🍀

@clankenluben 🧒he sees himself as an Einstein

People don't know🧒Yann is just farming engagement on X, replying to Elon. That's his adaptation to SV hype cycles. The problem is, this CS unc is too rude toward real research engineers who achieved this sci-fi car demo. He employs the same logic that burned the city hall in Limoges, but with a different execution layer. To me, LeWorldModel is just transitional applied science no one will remember in the 2100s. But humans will still drive Porsche ∧ Mercedes in the 22nd C☝️His achievement is remarkable but not noteworthy for humanity, unlike Valses Nobles et Sentimentales by Ravel

Yann LeCun @ylecun

a week ago

@spiroskaye @elonmusk I'm definitely not anti-achievement and never criticized real achievements, including by Elon. I even own two Teslas. I do expose marketing stunts that, while cool on the surface, are not the achievements they are claimed to be. Compressed air thrusters might give you

290 843 8K 648K 537

@alexandr_wang Muse...🦦sip it gradually....careful...bro

@SamCKx @ylecun nah🦦chat. That's Doc's strategy to farm engagement on X. His adaptation to SV hype culture ☝️🧒 nothing more

@clankenluben 🦦Latent Space

yo🧒Reject a myth that AI is mere software, as humans’ advanced cognition has originated from our hardware’s evolution. Our ancestors have formulated the software layer as formal education. This hardware logic can support (if there is no optimal hardware for an AGI model, AGI doesn’t spawn)☝️ Some may argue this kid is mixing biology ∧ CS; nah that’s an optimistic take🦦Do you remember brute-forcing ML in ChatGPT just has marginally improved Sam’s frontier model? Software maxxing hasn’t landed on autonomous AI models until today. Plus, building an AGI computing centre is always physical ∧ natural-scientific, employing advanced engineering ∴ Human brain ∧ AGI share the same foundational point, which is hardware dictates how the software layer can run. But, let’s restart AI analysis from JEPA Yann’s LeWorldModel🍰bro tl;dr JEPA meaningfully gives us very accurate ∧ limited brains for cyborgs ∧ still an AGI model is missing Here you can see skeletal essences of Yann’s LeWorldModel (JEPA)🐆 LeWorldModel can provide advanced psychomotor skills to cyborgs ∵ 1. JEPA’s first function is state abstraction 2. the encoder maps each frame observation into a compact latent representation 3. the model works in a low-dimensional space instead of pixels 4. It’s a physics engine to predict what’s coming next from observable hidden cues rn🧒for example, LLMs🤖: (seeing CR7 is running)…He’s Cristiano Ronaldo. The picture is taken in the 2010s. (searching…the web) Yes he GOATed in this match again🐐 JEPA🤖: (seeing CR7 is running)…the bro built different…(analysing the player’s movement ∧ ball’s trajectory…) the subject is going to GOAT; likelihood 91% On the surface, LLMs can look smarter ∵ they are cheating💀Mechanistically, JEPA provides advanced physical cognition with cyborgs that can delete most human blue-collar nodes from the labour grid. Mass-scale economy reconstruction is emerging🦦 What JEPA lacks: ☝️ Meta-awareness mimicry to do agentic self-correction *Consciousness itself is not necessary for AI ah, why? ∵ AI is based on (Si)licon while we humans are based on (C)arbon-rich structures🦦Consciousness in primates serves as their self-correcting system, not a mythic spiritual supremacy🧦 ✌️Mathematical rigour in reasoning: JEPA employs an encoder that maps images ∧ a predictor that autoregressively forecasts the next state. So no CoT here🦦Due to the nature of predicting the next physical reality with likelihood, JEPA can fail to surpass LLMs’ greedy deductive reasoning. Being open to the world amplifies noises in deductive reasoning. 🚘Tesla’s FSD (narrowed world model) is struggling to recognise kids in costumes (Halloween), which suggests JEPA may struggle with processing a vast amount of known data IRL. This world model’s cognitive limit meaningfully negates LLMs’ obsolescence. 🍀Need more strategies? DM me, sir

oh dear, this text signals…🦦You fed Claude my image, LLM-tier AI can’t process my reasoning fr. (If anyone has no access to optimal knowledge(x), their execution phase must produce mid outcomes) ☝️ Your AI’s criticism is superficial; chronologically ∧ mechanistically, my line always happens first. Imagine a surveillance camera, chat. My line highlights the foundational clue to spot a suspect, your line nitpicks how to labour cops. Without reliable clues, your cops can’t work accordingly. Kinda lowkey elementary-tier logical inference 🍵🦥🍰

@alexandr_wang Zuck's read is logically correct as documented in multiple peer-reviewed studies ∴ Alex's perception of Zuck is true too🦦alpha feels alpha