cesar ai

@onchainmilady car office setups are getting next level

This guy has a complete AI set up in his Tesla > MacMini > Portable monitor > Tesla monitor > PowerBank You might think that he is stupid, before you understand that MacMini set up is 3x cheaper than MacBook Now imagine what opens to him if he runs local LLMs wherever he goes and he has 0 dependance on any subscription Bookmark the full guide below

@RitOnchain the train on synthetic test on real part is super clever

Senior Quant Researchers just revealed how they use Generative AI for trading and asset management. 60-minutes. free. By Quant PHD Researchers. here's what they cover: • why LLMs hallucinate financial logic (like misapplying the square root of 252) • agentic AI frameworks for parsing alternative data (10-Ks, Fed speeches) • evaluating synthetic data via "train on synthetic, test on real" protocols • the myth of explainability vs 7-billion parameter trading models worth more than any $900 Quant course.

@0xwhrrari this sounds like a must watch

Andrew Ng and AMD CEO revealed the plan How anyone can become a $40k AI engineer from scratch: 00:00 - one engineer with AI can replace 5 employees 14:15 - one day with AI can replace 3 months of work 23:28 - how to build a profitable company for $20 This free 26-minute talk is worth more than a $90,000 Stanford AI degree Bookmark and watch it today Then read the article below

@TouchlineX 0: 2 ez for Spain

📸 - THIS LITTLE BOY IS CRYING A LOT NOW!

@Yumzlef mcp integrations are changing the game completely

Fanvue just shipped an MCP connector for Claude - and it's a clean example of where creator tooling is heading MCP lets an assistant read a platform's real data through a defined interface. So Claude can pull your actual analytics, message history, and pricing signals, then draft replies in your own voice - instead of you copy-pasting screenshots into a chat. The shift isn't "AI writes your posts." It's your tools becoming things an agent can operate directly. Bookmark this before every creator platform ships one.

@elg_oleksandr absolute dream workstation right here

HE INVESTED $14,000 IN A PERSONAL AI SERVER THAT RUNS EVERY MODEL LOCALLY WITHOUT SENDING A SINGLE REQUEST TO THE CLOUD. Not built for gaming. Not built for crypto. Built to run advanced AI models entirely on hardware sitting in his own office. The system is powered by a 96-core Threadripper PRO, backed by 512GB of memory and four GPUs with a combined 96GB of VRAM. Every component is carefully assembled. One small mistake could ruin parts worth thousands of dollars. But the hardware is only half the story. The real advantage is complete local inference. No cloud APIs. No recurring subscriptions. No data leaving the machine. He opens a terminal, enters a prompt, and the GPUs immediately spin up. Moments later, the model generates a response—all without connecting to external servers. Everything stays private. Everything runs offline. Everything remains under his control. Most people rent AI compute every month and send their prompts to someone else's infrastructure. He chose a different path. One upfront investment. Unlimited local AI. His machine isn't just another high-end PC. It's a self-owned AI workstation built to operate independently.

@Marko_Poly having that tracking layer is a lifesaver

a trader is up $11 million in 30 days. his last 7 days say you shouldn't copy him. both numbers are on the same profile - and that's the point. > swisstony on Stride: +$11.06M over 30 days, $1.63B in volume, biggest win $1.17M > that's a 0.68% edge per dollar traded - enormous size, microscopic margin > last 7 days: 304 bets, 55% win rate, ROI −4%, edge over market −12% > translation: this week the market priced his entries better than he did > 46% soccer, 43% tennis, avg entry 71¢ - hundreds of simultaneous positions on favorites this is the difference between a P&L screenshot and a track record. the $11M headline says "follow him anywhere." the 7-day edge number says "his current entries are losing to the market." most copy-trading sells you the first number and hides the second. the reason I'll actually use Stride's data: it shows edge over market and what copying would have returned after slippage - the numbers that decide whether size means skill or just size. $11M traders exist. copyable weeks are rarer. learn to tell them apart before you tap anything. Join Stride - [betonstride.com/r/q38rvmx]

@Atenov_D @prophetmarketai fading the machine sounds incredibly fun

Every prediction market before this needed a stranger to bet against you. @prophetmarketai just deleted the other trader. A consensus of 5 AI models prices any Yes/No question in under 30 seconds and takes the opposite side of your trade itself. No liquidity to wait for. No dead market. You're not fading the crowd anymore. You're fading the machine. How it works: → Write any verifiable Yes/No question with a resolution date. → 5 models price it in ~30 seconds. → You pick Yes or No. The AI is always the counterparty. → Odds reprice live as news moves. → AI panel settles it against primary sources. Sports, crypto, macro, esports, weather, anime, streamers - the entire long tail of markets that never had a matching trader before. One niche you read better than a general model = a standing edge against the machine. You don't need to be right about everything. You need one corner nobody at an AI lab is watching

@crytonbuton packaging is where the real value is

$41,000 from a free GitHub upload. The creator was 13 years old. Most people downloaded the script. Mason built a business. He handed the project to Fable 5 with clear rules. The AI planned every update. Cheap models handled the coding. Fable reviewed every change before shipping. Setup cost: $120. Launch time: one weekend. Monthly subscription: $49. 842 customers. Revenue: $41,258. Everyone is chasing new ideas. The money is in packaging existing ones. Save this.

@kadsxr scaling personality is wild

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.

@Nekt_0 protecting the loop is everything

Karpathy sat with Lex and explained why 3AM is his best working hour. No meetings. No messages. No errands. No humans pulling the problem out of his head. He said hard work needs days of uninterrupted context. You load the problem into mental RAM, keep it there, and let the answer compound instead of restarting from zero every morning. That was the whole productivity system. Not a calendar hack. Not a morning routine. Just protecting the loop long enough for the work to finish. The article is pointing at the same thing from the AI side. Agents only become useful when they stop acting like one-off chats and start running in loops: take context, act, check, update, repeat. Same pattern, different machine. Karpathy does it with attention. AI agents do it with context windows, tools and memory. Prophet does it with markets. @prophetmarketai turns a future event into a yes/no market, then lets an AI price it and take the other side. You are not waiting for some random person to match your view. You create the market, make the forecast, and the system gives it a live interface. That is the interesting part. Opinions are cheap until they are forced into a structure. Will this happen or not? What price is the AI giving it? Is your read sharper than the market? Deep work needs a clean loop. Conviction needs a clean market. Prophet is not available in the US. Predicting carries risk app.prophetmarket.ai/?ref=nOUbD3yOM…

@scrygg strix halo mini pcs are absolutely insane for local ai

THIS MINI PC HIDES A LOCAL AI WORKSTATION INSIDE A CONSOLE-SIZED BOX In 58 seconds, the creator pushes the GMKtec EVO-X2 beyond the spec sheet. One Ryzen AI Max+ 395, Radeon 8060S graphics, a 2TB SSD and a built-in local AI assistant inside a box that barely takes up desk space Then the doubt appears AI branding is easy mini PCs can throttle and expensive hardware means nothing if real performance collapses So he moves straight into the test ultra settings 50% render scale 165 FPS at one point then a stable 133 FPS under gameplay That is where the machine stops looking like another mini PC. The same hardware can handle games, daily workflows and local AI without turning the desk into a GPU tower The payoff sits under the hood 128GB of LPDDR5X memory Not a cloud demo Not a gaming gimmick A compact AI machine with enough memory to change what can run locally

@ArchitectHappy_ notebooklm is easily one of the best free tools out there

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 ↓

@insomnia_vip @sixsevenapp finally a mini app that is actually fun

TELEGRAM FEELS FUN AGAIN For a while every new mini app started to feel exactly the same Then I came across @sixsevenapp Instead of endless tapping, everything revolves around one meme, your Telegram account and competing with other people The leaderboard, MOG Arena and all the small interactions make it surprisingly fun to come back Best communities arent built by forcing engagement They grow because people genuinely enjoy being part of them Feels like 67 is doing exactly that

@Abobsterina packaged engineering discipline is an incredible resource

THE DIRECTOR WHO RAN GEMINI’S DEVELOPER EXPERIENCE AT GOOGLE OPEN-SOURCED HIS PERSONAL CLAUDE SETUP. IT JUST PASSED 77,000 STARS. Addy Osmani. Fourteen years at Google. Chrome DevTools, Lighthouse, Core Web Vitals — if you build for the web, you have used something he shipped. His last role: Director at Google Cloud AI, leading Gemini’s developer experience. He left the company weeks ago. His personal loadout stayed public. agent-skills is what it sounds like: the production-grade skill pack he ran himself. Testing, code review, refactoring, debugging — 15 years of Google engineering discipline, packaged as drop-in skills for any coding agent. MIT license. One install. github.com/addyosmani/age… Sit with the irony for a second. The man responsible for making Gemini pleasant for developers maintained a public toolkit that plugs straight into Claude Code. Not a leak. Not a scandal. Just an engineer who works where the work is, publishing what actually helped him. The detail most people miss: this is what separates a senior setup from a default one. Not secret prompts. Not a better model. Encoded judgment — what to check before declaring something done, how to review, when to refactor, written down once so the agent stops guessing at standards. 77,480 stars say the market noticed. Most people’s agent setup is a raw install and vibes. One of the most respected engineers of the modern web spent years distilling his into files, and handed it over for free. Installing it takes a minute. Ignoring it is a choice.

@helicerat0x bandwidth really is the silent bottleneck

NVIDIA CEO, Jensen Huang gave Stanford 72 minutes on how compute actually works: 11:33 - co-design: 1,000,000x in 10 years, Moore's law gave 100x 17:59 - NVIDIA uses more Anthropic and OpenAI tokens than just about anybody 28:16 - MFU, the metric every lab tracks, is simply wrong 30:09 - H100 price is bandwidth, not flops the bottleneck that shaped NVIDIA's multi-billion dollar racks is the same one in your $700 GPU watch it, then read the breakdown of 19 GPUs ranked by bandwidth, not VRAM, below

A 24-YEAR-OLD CREATOR TURNED AN AI MODEL PERSONA INTO $3.8K/MONTH ON FANVUE The mistake is thinking the face is the product. The product is the character people can recognize, remember, and pay to keep seeing. The reel gives her a simple wrapper: a plain intro, a sudden character switch, a short caption, and an ending that can loop. Then she rebuilds the format around an AI model instead of making her real identity the asset. It is not a realistic clone, a fake celebrity, or a character with age ambiguity. It is a fictional adult-only persona with one clear lane: cowgirl aesthetic, private photo drops, and paid channel access. The video does the attention job, but the business happens after the scroll. The path is short reel traffic to profile curiosity, then profile curiosity to Fanvue, then Fanvue to paid tier, then paid tier to private channel upsell. If 40,000 viewers create 400 profile visits, 40 clicks, and 8 subscribers at $19 to $49, the first win is not viral fame. The first win is proof that the character has buying intent. After that, the work becomes operational. She can make 20 safe visual variations, keep the same persona rules, test captions, track which scenes drive clicks, and cut anything that feels too close to a real-person likeness. The hard constraint is not image quality anymore. The hard constraint is trust. If the audience does not understand what they are subscribing to, refunds and platform risk eat the margin. That is why AI models are not just replacing creators here. They are turning creators into small media owners with characters, funnels, and retention math.

@refudgio definitely trying this setup tonight

OBSIDIAN'S CEO AND ANDREJ KARPATHY JUST QUIETLY KILLED THE SECOND BRAIN INDUSTRY it started as a single gist. 21 million people read it before anyone noticed what it actually meant the idea: stop treating your notes like a diary and start treating them like a codebase. Obsidian is the editor. Claude Code is the engineer working inside it. your notes are the thing being built three commands run the whole thing: one drops in anything, a video, a PDF, a random thought and splits it into small linked pieces that plug straight into what you already have written one lets you ask a question and get an answer pulled from your own words, your own pages, not a guess from a model that's never met you one runs quietly every week, hunts down the notes that contradict each other, deletes what's gone stale, and reconnects the orphaned pages nobody linked back then the Obsidian CEO shipped something that wasn't an "AI button" - five small skill files that teach Claude to speak Obsidian's own language natively. no broken links, no messy imports that repo hit 41,000 stars in weeks Karpathy pointed the whole thing at 100 of his own articles, 400,000 words, and just let it run. it links itself. it audits itself. it never sleeps no database. no subscription. just folders of text and an agent that does the one job humans always quit halfway through: keeping the filing system honest most vaults are graveyards. this one reads itself before you wake up

@Skaly__Bull solving context amnesia changes the whole game

Five models. 95% accurate each Stack them with nobody watching and that's a 23% failure rate Nobody catches it Nobody remembers what already broke Watch an agent forget its own plan mid-task - it doesn't get dumber, it goes blind The fix isn't a smarter model It isn't a fancier vector database either It's one controller holding the whole run in its head - every subtask, every failure, every fix - memory as plain files, not a black box Claude in that seat plans the work, routes it to Grok, GPT and Llama and lets nothing fail quietly Intelligence was never the bottleneck Amnesia was Full breakdown below

@noisyb0y1 one prompt replacing a sales team is wild

I PASTED ONE PROMPT INTO CLAUDE - AND IT BUILT ME A FULL CLIENT ACQUISITION SYSTEM IN MINUTES no more agencies, no more sales team - one prompt and the system finds clients itself Claude analyzed the niche, found where clients hang out, wrote outreach scripts and built the funnel from zero what agencies charge $3,000/month to do manually - done automatically in one session the system runs 24/7, finds leads, qualifies them and passes only the ones ready to buy one prompt replaced an entire sales department - and costs $20/month instead of $3,000