BACKWOOD

@igor_mikerin @PolymarketTrade @Polymarket good job igor'

Turning $10 into 80x in a single day. Yes, you read that right. Polymarket trader I’ve talked about several times is still using his weather market strategy. Today, in just 24 hours, he grew his account from $10 to $800. It’s an incredible phenomenon that we’re all watching. According to his profile: > @opopv?via=igor-mikerin" target="_blank" rel="nofollow noopener">polymarket.com/ru/@opopv?via=… What markets does he trade the most? “Will the high temperature in Taipei or Shenzhen reach X°C?” Can you name at least one reason not to look for an edge in weather markets?

@sandy4kad will try it later for sure thx buddy

HIS CLAUDE.md MAKES EVERY TOPIC GO THROUGH THE SAME 4 STAGES. NOTHING GETS FILED UNTIL IT'S BEEN TAKEN APART 4 TIMES no thread. no writeup. just a terminal window next to a graph view, someone recorded the screen mid-session zoom into the graph and the pattern jumps out immediately. every subject repeats itself 4 times. "fiscal policy quick reference." "fiscal policy checklist." "fiscal policy case study." "fiscal policy deep dive." same thing for cognitive dissonance, for the french revolution, for churn rate, for a hundred other topics scattered across the screen that's not obsidian doing that. that's the instruction sitting in his CLAUDE.md. one rule: nothing gets marked "understood" until it passes through all 4 stages. quick reference for the raw fact. checklist for how to apply it. case study for where it actually played out. deep dive for the mechanism underneath he said most people learn something once, feel like they get it, and move on. the CLAUDE.md is built to stop him from doing that to himself. claude won't treat a topic as closed until the fourth note exists the terminal running behind the graph is the automation. he doesn't manually create 4 notes per topic. he drops one idea in and claude spins up the other 3 versions on its own, each one forcing a different angle on the same fact hundreds of nodes on that screen. almost none of them are one-offs. every one has 3 siblings he never would have written by hand most people write a CLAUDE.md to sound smarter to the model this guy wrote one that won't let him fake understanding anything i post builds like this most days - follow before the next one disappears

@Gyome1_ yeah that's a very clear explanation

@0xBackwood boris always on point, love these conferences with him

@Dexonfxf thx for great analyze fam

MOST PEOPLE USE CLAUDE LIKE A CHATBOT. A FEW ARE USING IT AS A SECOND BRAIN THAT NEVER FORGETS Every day, most people repeat themselves to AI. Same context, same background, same "here's who I am" - over and over. One builder connected Claude directly to Obsidian. Now every note, every idea, every project update gets read and written automatically. Ask a question today, get an answer built on everything you've ever told it. No re-explaining. No lost threads. Save this before it's buried in your feed.

@0xzynex thanks Zynex already bookmarked

A free model just did the one thing every paid tool still charges you for. Almost nobody noticed. Kimi K3 dropped open source this week, and it doesn't just talk anymore. It hands you finished files. Not a draft. Not a "here's how you could do it." The actual done work, ready to ship. The part that breaks people is the price. It's still free. Fully open, no subscription, no waitlist, no "upgrade to unlock." Companies are billing $200 a month for tools that do less than this and lock the output behind a paywall. Kimi K3 opened the door and let anyone walk in. Same result. Zero cost. The gap everyone said would take years to close just closed in one release. Everyone's still paying for the old thing. The upgrade is sitting right there, free, and barely anyone pressed play.

@sairahul1 no way man this is real gem

MOONSHOT JUST CLONED CLAUDE CODE AND MADE IT FREE it's called Kimi Code CLI. open source. MIT licensed. built by the team behind Kimi K3. and it already does things Claude Code doesn't. → drop a screen recording directly into chat → built-in planner, coder and explorer agents → plan mode before touching a single file → automatic MCP configuration → works across VS Code, JetBrains and Zed → one binary. starts in milliseconds. the CLI costs nothing. K3 starts at around $3 per million tokens. worth using before everyone switches.

@Gyome1_ great speech here

Stephen Chin, Head of Developer Relations at Neo4j: “Most AI agents lose the context behind decisions across Slack, CRM, emails, support tickets, and internal databases.” Foundation Capital says context graphs could become a $3 trillion startup opportunity. In this talk, Stephen explains how to build agents that remember past interactions, retrieve the relationships normal RAG misses, and preserve the reasoning behind every decision. The system connects three layers: short-term memory long-term memory reasoning traces Every new run reads from the graph, makes a grounded decision, then writes the new context back for the next agent. Bookmark and then read the article below on how to build context graphs that turn disconnected company data into memory for autonomous agents.

@OxVelnox very useful skills thank u fren

60 SKILLS. ONE CLAUDE SUBSCRIPTION. SOMEONE JUST TURNED IT INTO AN ENTIRE BUSINESS. Marketing demand acquisition. Campaign analytics. Content creator. Social media analyzer. App store optimization. C-level advisor. Every folder has a SKILL md inside. Every skill tells Claude exactly how to do that job. Not prompts. Not templates. A system where Claude knows your standards before it writes a single word. Most people use Claude as a chatbot. This guy built a business on top of it. The gap between those two things is a markdown file in a folder.

@N01ennn amazin material thx man

Stop negotiating against yourself Anthropic's Claude Code team just explained why the old software rules are dead and why they now build the things they used to call impossible: 03:51 - product taste now matters more than implementation 05:09 - rewrites went from forbidden to encouraged 37:53 - why implementation getting cheaper means you should be more ambitious 41:50 - Fable editing a full conference video in one shot their co-founders' rule: we do not negotiate against ourselves. stop talking yourself out of the ambitious thing writing code was the constraint. ambition is the new one Watch it, then build something bigger with the guide below

A 21 YEAR OLD JUST REVEALED HER EMERGENCY SURVIVAL PLAN. IF SHE WENT BROKE TODAY, SHE WOULD LET CHATGPT RUN A FULLY AUTOMATED YOUTUBE CHANNEL. She posts a video breaking down her exact workflow. She says it is the easiest $690 she has ever made in a single week. She opens ChatGPT. She goes into the developer settings and connects a custom tool called Composio MCP. She does not write a script. She does not open an editing software. She types one sentence into the chat: "Generate a viral AI bunny video with Composio and post it to a YouTube Short." ChatGPT takes over. It generates the idea. It creates the video file. Then, because it is connected to Composio, it bypasses her entirely. The AI logs into her YouTube account, writes the title, and hits publish. She just watches the progress bar on her screen. Her dashboard shows $690 in pure revenue for a single week of doing absolutely nothing. The traditional creator economy is obsessed with editing, thumbnails, and uploading schedules. The ghost economy realizes that if an AI can make the video, it can also post it. A traditional YouTuber spends three hours editing and uploading. She spends five seconds typing a sentence and closes her laptop. Most people will watch her video and call it lazy. A few will open a new tab.

The real alpha in this architect's approach is shifting from static storage to a dynamic dependency graph. Using an LLM to auto-generate the semantic layer is a massive unlock. It essentially turns a messy "second brain" into a compounding asset. Instead of fighting context collapse, you’re building a bespoke oracle for your own thoughts.

A SYSTEMS ARCHITECT FED CLAUDE NINE YEARS OF NOTES AND TURNED 4,102 FILES INTO A KNOWLEDGE GRAPH. Daily logs. System designs. Deep research. Half-finished ideas from years ago. Not a flat folder. A map of how his thinking developed over time. The useful part is not the visual. It is the semantic layer. Instead of manually tagging every note, AI proposes which ideas are related, what supports what, and where an old thought connects to current work. That turns personal knowledge into something closer to a dependency graph. You can trace a decision back to its source. You can see which old research shaped a new project. You can find the notes that keep appearing around the same problem. Most people save information and hope future-them remembers the right search term. A graph changes the question. Not "where did I write that?" But "what does this idea depend on, and what else does it affect?" The vault stops being a library. It becomes a visible model of how you think.

@irongiantXBT The Kimi K3 is still more affordable than сlaude, for example, especially when you take account of account suspensions

PROS ARE BUILDING ENTIRE GAMES WITH CLAUDE WHILE LABS PANIC AND THROW $100 CREDITS AT ANGRY USERS amateurs are distracted by Kimi K3 and GPT-5.6 Sol fighting for leaderboard status. meanwhile the frontier labs are terrified of churn and bundling Fable 5 into paid tiers to stop mass cancellations. the reality of the current ai pricing war: > Kimi K3 steals the top ranking from established labs > panicked executives throw $100 retention credits at pro users > developers ignore the drama and use Claude to generate fully functional games from scratch the unedited 30-second gameplay footage 👇

@0xNoryxx staff ai replace is currently at a high level rn it’s terrifying to think how many people might lose their jobs. they’re simply being made redundant, and we’re not talking about 50 here, but 500 or even thousands – for example, if it’s a large company or organisation.

Andrew Ng studied AI agents for 14 years - and shared the main secrets in 1 hour that courses sell for $10,000 agentic AI is not a chatbot that responds - it is a system that plans, acts, checks itself and repeats until the task is done most developers build agents that fail on the first error - the right agent fixes itself and keeps going reflection is when the agent rereads its own response and finds mistakes before you see them - one function that multiplies quality multi-agent systems are not one smart agent but a team of specialists where each does their part better than everyone else combined tool use is the difference between an agent that thinks and an agent that acts - browser, code, API, files - these are the hands of the agent he said agentic AI will make one developer more productive than a team of 50 people - and it is already happening save this and watch today - 1 hour that replaces a $10,000 course

@Serantych thx for sharing already bookmarked

Anthropic engineer just showed how he plans, builds, and runs loops with Claude Code in 40 minutes: "as the models get smarter they need less direction - we cut Claude Code's system prompt by 80%." here's how he actually works: → /goal - give Claude an exit condition so it powers through instead of stopping to ask → workflows - one sub-agent builds, a separate one verifies. a model grading itself goes easy → plan by exploring first: make Claude map a tool's edge cases before you build bookmark this ↓

@0xCodez yeah they used it cause haven't any limits in claude lmaoo

Anthropic engineer: "At Anthropic, >90% of our engineers are building with self-improving loops. In 4-6 months, it will be 100%. my agentic loops can run for days without spending hundreds of dollars." in this 40-minute podcast, an Anthropic engineer reveals how to build effective agents from scratch. Agent → harness → loops → memory = modern agent This one video will replace 10 paid courses on vibe-coding. Watch it today, then explore the same setup in the article below.

@0xMovez wow i need to check it

HIGGSFIELD JUST MADE THE LLM INSIDE THEIR SUPERCOMPUTER FREE. Before, every text message burned credits. Brainstorming, scripting, prompt refining - all cost money. Now the "thinking" part is zero. → Brainstorm content concepts → Write and refine scripts → Perfect your prompts until they're exactly right Iterate on your ideas for free and only pay when you generate an image or video. This changes how you use the platform ↓

@Gyome1_ the second brain is not a myth anymore

A CHINESE DEVELOPER SPENT 6 YEARS BUILDING A SECOND BRAIN. THEN HE TURNED IT ON AND WATCHED HIS ENTIRE LIFE APPEAR AS A MAP the clip was buried on Chinese server dark Obsidian window. thousands of notes. one graph spreading across the screen like a nervous system at first it looks like a pretty visualization then he opens a math formula the graph immediately highlights every theorem, diagram and old note connected to it something he wrote three years ago appears beside what he is studying today that is when you realize this is not a folder of notes anymore it is a model of how he thinks every subject is a district every note is a memory the lines show which ideas keep touching each other, even when the person who wrote them forgot the connection then an AI agent enters the vault it can trace the graph, find missing links, rebuild outdated notes and surface the exact page needed for the current problem the developer no longer searches through thousands of files the brain brings the right memory to him people spend $200 a month on smarter models and still explain themselves from zero in every chat this guy built one system the models can enter save this before your Obsidian becomes another digital graveyard the full second-brain setup is in the article below

@Nekt_0 linus the legend of tech world for sure

32 years after creating Linux, Linus Torvalds was asked if AI-written code worries him: “Developers already write an enormous number of bugs every day.” In 4 minutes he explains why LLMs are another step in programming’s evolution, from machine code to assembly, C, Rust and now natural language. His most practical use case is not replacing developers. It is reviewing code and catching stupid patterns that compilers miss. Most people debate whether AI will hallucinate. Linus is more interested in whether it can catch the mistakes humans already ship.

@aiedge_ stop scrolling just build your billion dollar companies lmao

If you're 18-30 y/o, I want you to watch this clip from Sam Altman. I've seen it 10+ times, and it never stops hitting. Sam Altman: "We're going to see 10-person billion dollar companies pretty soon... if I were 22 right now, I'd feel like the luckiest kid in history." It's easy to get caught up in AI dommerism. But the reality is, we are truly living through the greatest technological shift ever, and it should excite you. The average person now has dozens of real, viable ways to build generational wealth through this AI gold rush. Truly a great time to be alive. F*ck the doomers.

@Skaly__Bull the loops is our near future tbh

Your AI's real problem isn't intelligence It's impulse control He handed Claude Code a 3-page prompt and expected magic Got 100 hotfixes instead - each one punching two new holes through the walls it just patched Fix the sub-agents, break the chat Fix the chat, wreck the UI Classic whack-a-mole, except the mole kept multiplying Turns out the AI wasn't a bad coder It was a hyperactive junior dev that starts typing before it finishes listening One skill killed the chaos: freeze the AI, interrogate the request, force an approved written spec before it touches a single file Full breakdown below