John Doe

everyone is arguing whether spacex at $2 trillion is too expensive, almost nobody noticed what actually changed since 1980 apple went public at under $2 billion and 15 times revenue back then spacex wants you to buy at nearly $2 trillion and 100 times revenue now but the real story is not the price, it is that the entire nature of an IPO changed underneath us in 1980 you bought a business and hoped for a vision, in 2026 you buy the vision and hope for the business almost nobody buying this IPO on june 12 will read the 300 page document to understand which one they are actually getting i fed both filings to Claude and had it find the five differences that explain the whole shift that understanding is the edge most retail will never bother to get

@KuramaOnChain Amazing find, mate. Thx for the timestamps, going straight to 07:26 first.

Marc Andreessen spent 37 minutes explaining where the money in AI actually goes: 07:26 - The slide that exposes a fake AI company 14:11 - A $20M missile shooting down a $10K drone 26:41 - A $50K robot dog and its $1,600 clone 31:57 - Why token prices fell 100x in one year This 37-minute watch will replace a $200K MBA. Watch it today, then read how to become an AI engineer in 2026 in the article below.

Goldman Sachs open sourced the math its quants use. The only thing they kept locked is the data. Think about how strange that is for a second. A bank that charges institutional clients for access, sitting on 25 years of derivative pricing and risk work, and they put the whole thing on GitHub. Apache license. 11,000 stars. Anyone can read it. No degree. No desk. No permission. I went through it last night, expecting the usual open source PR stunt. Half empty repo, nice README, nothing inside. It is not that. It is the real stack. Structuring, risk, backtesting, the same Python their desks build on. And halfway through I realised I had the whole thing backwards. The math was never the moat. The feed was. They gave away the brain and kept the eyes, and they are still charging for the eyes. Which means the barrier stopped being knowledge a long time ago. It is a data pipe now, and you can build one of those yourself for the price of a coffee. Here is how: 1) pip install gs-quant 2) Open a notebook 3) Read the pricing code Nobody knows how long a bank leaves something like this up. Grab it while it is still there. Read the guide below. Follow @DoeOnChain

@callsofmidas Appreciate that

@DoeOnChain this is the kind of post I come here for

@KuramaOnChain tnx bro

@DoeOnChain rare find bro

github repo: github.com/goldmansachs/g…

@callsofmidas Top tier find. My agent rereads the same four files every session and I just accepted it lol

Imagine hiring someone brilliant who forgets the office layout every single morning. Every question you ask, they walk the whole building again. Open every door. Read every file. Then answer. That's your AI assistant on your own project right now. Somebody built the fix and 84,500 people already starred it. You type one command. Your entire project turns into a map it can actually read. > No API calls > No cloud > No embeddings > Free, MIT license > Works in Claude Code and Cursor Then you stop asking it to search. You ask it how two things connect, and it answers in seconds, because it finally has the map. Sound familiar? Every tool sells your assistant a bigger memory. This one just gave it a floor plan. Bookmark this before your next project.

@RaoulDukeDegen Yeah, local by default is why I trusted it enough to actually run it on a real repo.

@DoeOnChain this is sick especially since it supports docker and local runs too

You can put a coding agent on a cron job. Almost nobody knows this is possible. Most people use Claude Code like a chat window. You type, it types, you sit there and watch it work. Meanwhile there is a self hosted control center with 80,000 stars that runs the exact same agent on a schedule, with nobody in the room. No seat fees. No vendor lock. No cloud unless you ask for one. I set it up last night. Halfway through, I realised the bottleneck was never the model. It was me, sitting in front of a terminal being the runtime. So I handed it a GitHub issue before bed, and woke up to the tasks already broken down and a report sitting in Slack. Here is how: 1) Install OpenHands 2) Plug in your agent 3) Give it a repo and a schedule It runs Claude Code, Codex, Gemini or any ACP agent. Switch between them mid session and you lose nothing. Bookmark this one, it deserves far more attention than it is getting. Read the guide below. Follow @DoeOnChain

@callsofmidas That is the part that gets me, bro.

@DoeOnChain real alpha. nobody talks about cron + agents and it's the most obvious unlock

@KuramaOnChain Cheers mate, glad it landed.

@DoeOnChain real alpha, mate. bookmarked!

github repo: github.com/OpenHands/Open…

STOP PAYING FOR A DESIGN TOOL. THE AGENT ON YOUR LAPTOP IS THE DESIGN TOOL, IT JUST HAD NOWHERE TO DRAW. Open Design is a desktop app that hands your coding agent a canvas. Apache licensed, local first, 77,000 stars. Everybody is posting it as a free Claude Design clone. That is the least interesting thing about it. > It runs on your machine, not in someone else's tab. > It plugs into the agent you already pay for, Claude Code, Codex, Cursor, Gemini, OpenCode. > It builds prototypes, landing pages, dashboards, slides, images and video. > And it exports real files. HTML, PDF, PPTX, MP4. That last one is the whole thing. You do not walk away with a screenshot of a design, you walk away with the file. Here is the entire setup: Step 1: Download the app from github.com/nexu-io/open-d… Step 2: Point it at the coding agent you already use. Step 3: Describe what you want and open what it gives you. It is not free, you still pay your own model tokens. But nobody sits between you and your agent charging rent for the canvas. Watch it, then read the guide below. Bookmark this. Follow @DoeOnChain

@callsofmidas the quarterly alpha commentary detail is what kills me. devastating.

Hedge fund managers when they realize the same loop they charge 2 and 20 for is now a free script running on a laptop while they're still emailing "quarterly alpha commentary"

@callsofmidas appreciate it

@DoeOnChain top tier find, mate

YOUR CODING AGENT IS AN EAGER INTERN. THIS REPO TURNS IT INTO A SENIOR ENGINEER. Jesse Vincent spent twenty years shipping software, then wrote his entire method down as skill files and gave them away. The repo is called superpowers. MIT license, no paid tier, no cloud. Your agent writes code the second you ask. This one refuses until it knows what you actually want. > It interrogates you into a spec before a single line exists. > It shows you the design in chunks short enough to actually read. > It writes the plan for someone with no taste, no judgement and no context. > Then it hands that plan to subagents and makes them prove it with failing tests first. That third one is the whole trick. The plan is not written for a genius, it is written so that the dumbest possible executor cannot get it wrong. Here is the entire setup: Step 1: Install Claude Code, or Codex, or Cursor. It runs in all of them. Step 2: Install the plugin from github.com/obra/superpowe… Step 3: Start your next project and let it ask questions before it writes anything. Every agent fails in the same place. It starts typing before anybody decided what to build. Watch it, then read the guide below. Bookmark this. Follow @DoeOnChain

@KuramaOnChain glad it landed, mate

@DoeOnChain amazing find bro. the interrogation step alone would have saved me two rewrites last week.

@KuramaOnChain amazing find, mate. thx for the timestamps, going straight to 59:03 first

Andrej Karpathy spent 70 minutes explaining how he actually uses LLMs: 07:16 - He catches ChatGPT lying to him 17:26 - The minute he coined "vibe coding" 26:35 - Why he stopped typing to LLMs 59:03 - The prompt trick he says always works This 70-minute watch will replace 10 paid AI courses. Watch it today, then read how to become an AI engineer in 2026 in the article below.

@callsofmidas facts

@DoeOnChain bro this is smart. cheap model running the expensive model's playbook, that's the whole game

FABLE 5 WILL WRITE OUT ITS OWN BRAIN IF YOU ASK IT TO. ALMOST NOBODY ASKS. Not an answer. Not a snippet. The operating manual it runs on, dictated by the model itself, in plain markdown. Then you mount that file on Opus 4.8 and the cheap model starts making the expensive model's decisions. Here is the entire setup: Step 1: Open Fable 5 while it is still inside your plan. Step 2: Make it write down how it actually works, step by step, as its own manual. Step 3: Save it as a .md file. Step 4: Load it as project context on Opus 4.8 and let it run from there. Fable 5 already disappeared once in June with no warning, and access only runs through July 19 as it stands. So the window is real, even though half your timeline has the date wrong. You cannot own the model. You can own the way it thinks. Watch the video, then read the guide below. Bookmark this. Follow @DoeOnChain