Liquidden

OBSIDIAN'S CEO DROPPED 5 FREE SKILLS. HE'S CHARGING $2,200 EACH TO INSTALL THEM. 6 CLIENTS. $9,000/MO. No SaaS. No wrapper. No API resale. The skills sat public. Most gave a like. He read it and thought: who would pay to have this in-house? Law firms. Agencies. Medical practices. All run on one fragile system — critical knowledge in one head until that head leaves. At 0:07 the GitHub repo is right there — those are the free skills he sells the installation of. His loop: > Pulls 3 years of client files into one vault > Connects Claude, hardens permissions > Delivers in 3 days $2,200 setup. $1,500/mo. 6 clients. 10 hours a week. Five skills from the CEO. One business on top of them. Would you keep downloading free tools — or start delivering them for $2,200 a client?

@Pakero8x turns out the hard part was never the model

@Liquiddeny That's a really clear breakdown of RAG and why it's so powerful. Seeing companies already valued in the billions based on this tech definitely makes you rethink the "toy" LLM vs. the real-world applications.

Glean $7.2B. Harvey $11B. Sierra $16B. All built on the version of Second Brain Karpathy didn't show you. Karpathy's LLM wiki is the toy. Free, simple, enough for your own notes. The grown-up version is RAG. Instead of showing the model everything and hoping it finds the answer, RAG fetches only the pieces that matter — and answers from those with the source attached. What that gets you: > Works at any size — folder of notes or a company's entire document history > Doesn't hallucinate — every answer traces to a real document > Cheaper — reads a few relevant pages, not your whole library > Answers what nobody wrote down — connects pieces across files into an answer that never existed in one place Glean does this for company knowledge. Harvey for law. Sierra for customer support. $34.3B combined. Same trick. Every business sitting on a pile of documents needs this and can't build it. You can build it and sell it. Engineers who've shipped one get paid a premium over everyone else. Bookmark this.

@drakefomo everyone wants more clients😇

23 years old. No coaching badge. $5,850/month building glute routines for 390 lifters — and she's the one filming step-ups in the video. She built the AI routine planner with Claude. Users upload their equipment list, current split, and last month's progress photos. Minutes later they get a numbered 5-exercise glute routine in her exact filming format — cable kickbacks, hip thrusts, step-ups, whatever their gym has. New routine every week. The thing you'd pay a coach $150 per session to build. Delivered while you're still tying your shoes. 390 subscribers. $15 each. $5,850/month recurring. Her cost to run it: under $90. A personal trainer builds one program at a time. Hers builds 390 at once, at 3am. She's not selling workouts. She sold a system that generates new ones while she works out. Full routine engine + Claude prompt library — in the article below.

@drakefomo it's always markdown somehow 😂

@Liquiddeny funny how the simplest stack keeps winning 😂

$1,344/year in productivity apps. He replaced all of them with a $240/year Claude Fable 5 subscription and an Obsidian vault. Karpathy showed the wiki. Steph Ango shipped the 5 skills. This guy stacked both into an OS he talks to. "Hey Jarvis, give me the rundown for today." Claude Fable 5 reads his vault, applies Ango's Obsidian skills, returns a brief in his own frameworks. What it replaces: > Motion — $216/year > Reclaim — $180/year > Notion AI — $240/year > ClickUp Brain — $348/year > Superhuman Assist — $360/year His stack: > Obsidian vault (free) > Claude Fable 5 ($240/year) > Ango's 5 skill files (free) > One voice input pipeline $240 replaces $1,344. Voice-controlled personal OS. No SaaS. No wrapper. No API resale. Every enterprise tool is speedrunning the exact stack he already shipped. Full skill file setup + voice pipeline — in the article below.

@RoundtableSpace three years of discipline visualized)

An Anthropic lead engineer accidentally leaked his personal Obsidian and what's inside isn't code or prompts. It's a diagram of his own brain organized as a neural network. 8,893 nodes, 4,729 connections, 21 inputs, ReLU on every layer, thousands of connections firing in real time across 9,000 documents each with its own semantic space. This isn't a concept diagram from a blog. It's a living knowledge system that powers decision-making inside one of the most advanced AI companies on earth. Three years of discipline. One open Obsidian tab. A $10 a month app.

@andreysuperior funny how the simplest ideas sell the best

You don't fully understand what this loop could turn into. An AI agent is a while loop and some tool calls. He proved it in 30 lines of code. Message history. Step counter. LLM that runs until it stops itself. The simpler the tech, the bigger the gap between those who understand it and those who pay for it. He pointed that loop at a company's knowledge problem. An Obsidian vault connected to Claude. Ask it anything. It runs. It answers. It cites the source. A law firm associate types a question. Three seconds later the vault answers. No partner interrupted. No time wasted. Setup: $2,200. Retainer: $1,500. Multiply by six. That is $9,000 a month and a ten-hour week. The loop is free to understand. The business is knowing where to run it.

@sairahul1 now imagine explaining this to your boss from 2022 😅

Jacob Bank, former Google product lead: "I built up this team of 40 AI marketing agents to work with me. I'm the only marketing person." In a 15-minute talk, he shows what one person with the right setup now runs alone. Forty agents. One human. His AI bill is $500 a month, against the $50,000 a human team would cost. That's the math quietly minting the first solo fortunes of the AI era. Watch the talk, then read the article to build your first AI team.

@caldervol It might look ridiculous, but it's brilliant

@Liquiddeny funny how plain text keeps beating fancy stacks😅

One 1-page gist from Karpathy. 21M reads. 41,000 stars in weeks. The Obsidian CEO shipped code the same day. Karpathy flipped every note app: Obsidian is the IDE. Claude Code is the programmer. Your notes are the codebase. 3 commands run the whole system. Ingest — drop an article, a podcast, a PDF. Claude splits it into atomic pages linked to everything you already know. Query — ask anything. Claude answers from your notes, in your voice, citing your pages instead of guessing from training data. Lint — once a week Claude walks the vault, kills stale claims, wires orphan notes back in. Then Steph Ango, Obsidian CEO, moved. No "Ask AI" button. He shipped 5 skill files teaching Claude Obsidian's native language: wikilinks, Canvas, Bases, the CLI. Karpathy runs it on 100 articles, 400,000 words. No vector DB. No embeddings. No $20/mo app. Just markdown + an agent that never gets tired of the boring part. Your vault has 3,000 notes nobody will reopen. His reads all of them by breakfast. Full 3-command setup + the 5 skill files — in the article below.

@marlowxbt that's a clever distribution hack👏

One guy built a wedding camera app in a month and it now makes him over $1,000,000 a year. No team, no investors, no marketing budget. 100,000 downloads in the first month. Zero dollars on ads. He did not bolt marketing onto the product. He made using the product the marketing. You cannot use the app alone. The host has to pull every guest into it. One wedding is not one user. It is 200 people scanning the same code in one evening. The guest liked it at someone else's wedding. A month later he throws his own event and brings his own crowd. The loop spins for free. It does not look like an ad. To the guest it is a gift. So they install it gladly. $2 to $50 subscriptions. 5% pay. That is $100,000 a month. He did not win on budget. He sewed the growth into the product itself. Could you build the loop, or is it just luck?

@GransMarin57904 that's one optimistic roadmap

such a genius side hustle. turn free tech into thousands, stack crypto, and book those luxury flights.

@drakefomo now everyone's gonna ask for the prompt 😂

They didn't hit the gym. They asked Claude what fitness content brands were paying for. 14 seconds later — 12.3M views, $8,400 signed. Claude analyzed 400 stretching clips from the last 90 days on X. Found 3 patterns every viral clip had: intimate framing, low production, one line of on-screen text that reads like a private thought. They filmed one take. Uploaded once. 12.3M in a day. Brand DMs started at hour 6. By hour 48, 4 offers on the table — activewear, supplements, one dating app. $8,400 in retainers for 30 days of clips filmed the same way. They didn't build a fitness brand. They reverse-engineered what brands were already trying to rent. The 3 patterns Claude found + the brand DM templates — full breakdown in the article below.

@undefinedKi who's still embedding the whole pdf 😭

Everyone's telling you to build a Second Brain. Nobody ever explains what you'd actually do with it The Karpathy's LLM wiki setup is great. It's simple, it's free, and for your own notes it's enough. But it's the toy version. The grown-up version is RAG, and that's what actual companies are built on. Same goal, different scale. Instead of showing the model everything and hoping it finds the answer, RAG fetches only the pieces that matter for each question and answers from those, with the source attached. What that gets you: > It works at any size. A folder of notes or a company's entire document history, doesn't matter. > It doesn't hallucinate. Every answer traces back to a real document. > It's cheaper. The model reads a few relevant pages, not your whole library, every single time. > It answers what nobody wrote down. It connects pieces across all your files and gives you the answer that never existed in one place. Glean is worth $7.2B doing this for company knowledge. Harvey is at $11B doing it for law. Sierra at $16B for customer support. Every business sitting on a pile of documents needs this and can't build it. You can build it and sell it. And the engineers who've actually shipped one get paid a premium over everyone else. Same skill behind all of it. Bookmark this

@MyWestLord finally a use case that makes sense

Andrej Karpathy, the CEO of Obsidian, and Claude Code just built the smartest second brain on earth. It started with a 1-page gist that 21M people read. Karpathy frame flips everything you know about notes: Obsidian is the IDE, Claude Code is the programmer, and your notes are the codebase. You don’t ask AI questions it forgets by tomorrow you make it maintain a living wiki. 3 commands run the whole system. Ingest: drop an article, a podcast, a PDF, and Claude splits it into atomic pages linked to everything you already know. Query: ask anything and it answers from your own notes, in your own words, citing your own pages instead of guessing from training data. Lint: once a week Claude walks the entire vault, flags contradictions, kills stale claims, and wires orphan notes back in. Then Steph Ango made his move. The Obsidian CEO didn’t bolt an “Ask AI” button onto the app he shipped 5 skill files that teach Claude to write Obsidian’s native language: wikilinks, Canvas, Bases, the CLI. The repo crossed 13,900 stars in weeks and sits at 41,000 now. Karpathy runs it on his own reading: 100 articles and 400,000 words, cross-linked and maintained while he sleeps. No vector database, no embeddings, no $20 a month memory app just a folder of plain markdown and an agent that never gets tired of the boring part: the linking, the filing, the upkeep that killed every Zettelkasten since 1965. Your vault has 3,000 notes nobody will ever reopen. His read all of themselves by breakfast. Every app promised a second brain this is the first one that thinks.

@andreysuperior that's a solid niche

Do you actually understand what this vault setup could turn into. This is the craziest Claude combo found so far. Obsidian vault running alongside Claude. Claude reads the vault before it writes a single line. No more re-explaining context every session. Everything it ever needs is already there. Most people stopped there. Personal productivity win. Done. He thought about which companies pay the most for their AI to stop being a black box. Law firms. Agencies. Medical practices. He builds the vault for them. Connects Claude. Every employee gets a bot that reads company context before answering. Not a black box anymore. $2,200 to set it up. $1,500 a month to maintain it. 6 clients. $9,000 a month. 10 hours of work a week. He found the craziest combo. Then figured out who would pay for it.

@maxcryptoops simple systems always win

@Liquiddeny smart angle, turning public skills into installed workflows is where value shows up

OBSIDIAN'S CEO DROPPED 5 FREE SKILLS. HE'S CHARGING $2,200 EACH TO INSTALL THEM. 6 CLIENTS. $9,000/MO. No SaaS. No wrapper. No API resale. The skills sat public. Most gave a like. He read it and thought: who would pay to have this in-house? Law firms. Agencies. Medical practices. All run on one fragile system — critical knowledge in one head until that head leaves. At 0:07 the GitHub repo is right there — those are the free skills he sells the installation of. His loop: > Pulls 3 years of client files into one vault > Connects Claude, hardens permissions > Delivers in 3 days $2,200 setup. $1,500/mo. 6 clients. 10 hours a week. Five skills from the CEO. One business on top of them. Would you keep downloading free tools — or start delivering them for $2,200 a client?

@JacobMReed0 Free gets attention. Paid solves problems)

@Liquiddeny "Because 'trying' is just the first step. Next comes 'configuring', 'integrating', 'remembering to update', 'fixing it because it's broken' and 'explaining to the team how it works'. That's what they pay for. Free is a lure, not a business model."

1 CHINESE DEV. 1 OBSIDIAN VAULT. HIS ENTIRE WECHAT HISTORY JUST BECAME A SEARCHABLE MAP OF HIS LIFE. No new app. No paid tool. No cloud upload. He dumped every message he'd ever sent or received into a local vault. Connected Obsidian to Hermes. Let the system link everything. At 0:03 the zoom-out shows the scale — thousands of nodes, every conversation he's ever had, all connected and queryable in one place. His loop: > Export full WeChat history > Feed into Obsidian, powered by Hermes > Every message becomes a linked node > Ask one question, pull five conversations from three years ago Your messages aren't notes. They're the dataset you never queried. Save this. Every messaging app on your phone is one export away from becoming this. Would you keep letting conversations evaporate — or start turning them into a queryable brain?

@JacobMReed0 If you need it and it's free, too then why not give it a try?

@Liquiddeny This is the actual playbook right now - free content builds trust, but “install + maintain it for you” is where the money always was. Same reason agencies selling WordPress setups outlived the free - theme era.