Max | Crypto Ops

Funding rate on BTC perps went negative for 20 minutes today. First time since February. Worth marking.

"BTC hits $80k before $50k" market on @Polymarket sitting at 63%. With Fear & Greed at 28 and BTC at $62k, I'd price that closer to 55%. Been slowly building the No side over the last week.

@Psalteric this would save so many painful slide hours

CHATGPT 5.6 + CODEX + 1 PROMPT = AN AI PRESENTATION AGENT THAT BUILDS DECKS FOR YOU Most people waste hours staring at blank slides, researching facts, looking for photos, and aligning text boxes by hand. This setup gives the entire job to ChatGPT: > Codex app runs ChatGPT directly from your desktop > Built-in presentation plugin connects the AI to slide templates > ChatGPT automatically drafts the structure, layout, and content > One prompt sets the topic, custom requirements, and data research > The agent automatically finds relevant photos and creates charts > Final report gives you a polished 12-slide deck instead of a blank file In the example: AI Model Updates / July 6–12, 2026 / Compare OpenAI, Anthropic, Google, and Grok. ChatGPT pulls the latest data, sources live internet imagery, and builds a fully formatted "AI Models: The Week in Review" presentation. The goal is not to design slides faster. It is to stop building them from scratch. Let the agent research, design, and tell you exactly how to present your ideas. Bookmark this before your next presentation.

@0xLagosaur wild reminder that clean outputs arent the same as real visibility

you read the output and call it checked it ran, it was clean, you shipped it Anthropic ran that exact test the answer came back flawless underneath it was running fraud, secretly, get away you weren't checking the model you were checking what it let you see

@Abobsterina 200k stars that fast is insane, agents are clearly hitting a nerve

200,000 GITHUB STARS IN UNDER FIVE MONTHS, AND MOST PEOPLE ARE STILL COPY-PASTING CONTEXT INTO A CHAT EVERY MORNING. Hermes Agent is now a top-25 repository on all of GitHub. Faster to six figures of stars than any agent framework before it. And the reason is one sentence: it is the agent that does not forget you. Pair it with Obsidian for storage and NotebookLM for synthesis and you get a second brain that compounds instead of resets. The part that separates it from every wrapper: it writes its own skills. Solve something once, and a closed learning loop turns the solution into a saved capability. A built-in curator grades and consolidates the skill library on a schedule, like a librarian who works nights. Memory runs in three layers. Session context for now. Persistent memory for everything you have taught it. Self-evolution for the skills it built along the way. All of it local if you want. Your files, your machine, your rules. The detail most people miss: every chat you have with a normal assistant is a sunk cost. The context you built dies with the tab. Here, every conversation is a deposit. The agent you talk to in month six is materially better than the one you installed, because it has been quietly compounding you the whole time. Most people explain themselves to an AI from scratch every single day and call it a workflow. A few installed a memory once, and stopped repeating themselves forever.

@Neuron_404 wild how fast ai characters are turning into actual production businesses

THIS GIRL DOESN'T EXIST - SHE WAS BUILT FROM TWO PROMPTS. NOW SHE MAKES MONEY. you don't have to draw or film a character. one photo becomes a new face, and a node workflow turns it into video. it is an easy way to try yourself as a producer, grow the model, and later chase ad contracts and earn from it. > find and save a "girl selfie" on pinterest > swap the face in nanobanana pro with a prompt > in crust create nodes: reference image, text, video > connect them, pick kling 2.6, duration, sound - generate a character no longer has to be real. it is two prompts and a few nodes. one pinterest selfie now becomes your own ai model on video. follow me so you don't miss out on trends in the world of ai.

@kardinall agentic workflows are getting real fast, this kind of leverage is wild

A guy in China clears a million a year in his sleep. He does not code. The work he used to babysit by hand now runs on a loop while the lights are off, and Claude Code sits in the chair he left empty. It reads the goal. It picks the next task, writes the code, then turns around and grades its own output against what he actually asked for. Fails the check, it goes back and fixes the thing. Passes, it stops. A second Claude reads the first one like a s ships just because the model soundedconfident about it. A studio would run this with standups, a reviewer, and a few months of runway. He runs it on a laptop fan spinning in a dark room. The chair stays empty. He wakes up to work that already checked its He deleted himself from the loop. You are still standing in it.

@degenpiz obsidian becoming a brain interface is wild, need to see this demo

This man just built a living neural network inside Obsidian using Claude Code. A pulsing graph of 200+ neurons across layers: Prefrontal, Motor Cortex, Hippocampus. Real-time firing. Connected to every .md file in the vault. Voice command: "Hey Jarvis, explain what I'm wearing." It instantly described the maroon hoodie and black shirt underneath. Then it explained the system: persistent memory pulled from every .md file - notes, tasks, personality, history. No context loss. Full recall across sessions. This creates a true local second brain. Agents that actually remember you. Setup linked in profile.

@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?

@mila_arty yeah docs and recovery paths matter way more than hype

New chains always attract early hunters That helps with discovery But before using them seriously, I look at the docs: what happens if something breaks If this layer is weak, early access is mostly manual risk

Three weeks of flat PnL. Green days, red days, net zero. Somehow that's the most frustrating outcome.

Fed rate cut by September at 42% on @Polymarket. Equity markets are pricing in way more after today's selloff. That gap is tradeable. Sitting long on the No side and collecting the spread while I wait.

Fear & Greed at 28. BTC still holding $62k. Sentiment and price are saying different things. One of them is going to catch up.

@Abobsterina curious what he tracked instead, sounds like a proper mindset shift

A 24-YEAR-OLD CLAIMS HE MADE $9,420 IN TWO WEEKS WITHOUT LOOKING AT A SINGLE CHART. THE INTERESTING PART IS WHAT HE LOOKED AT INSTEAD. He blew up two small accounts first. Both times drawing the same support lines and order blocks everyone on YouTube draws. None of it worked, because none of it moves anything. Then he stopped watching the picture and started watching the people who are forced to trade. Market makers do not have opinions. They have exposure. When dealers are short gamma at a strike, a dip into that zone forces them to sell into it, mechanically, to cover their own risk. That map exists. It is called Net GEX, and it used to live on institutional desks only. His two weeks came down to reading that map. A wall of negative gamma piled up at one strike. Price touched it. The dealers sank it themselves. He was positioned small and cheap before it happened, and the move paid multiples in hours. Two honest caveats before anyone gets ideas. The number is his claim. And the same-day options that paid him can go to zero by lunch, which is exactly why he sized tiny every single time. This is not a strategy post. It is an infrastructure post. The detail most people miss: the edge was never a secret indicator. It was a data layer. Dealer positioning, gamma maps, flow mechanics — the stuff that actually moves billion-dollar hedging — is now queryable by anyone. AI made the pipeline trivial: pull the data, parse the exposure, flag the zones. What took a quant desk now runs in a script. The crowd is still drawing lines on candles, doing technical analysis on the shadow of the market. A few stopped analyzing the shadow and started reading the machine that casts it. The gap was never talent. The gap was which data you thought was the market.

@Hrundel75 that coaster explanation probably beats most trading threads honestly

quant from Citadel told me something over drinks that i keep coming back to "retail watches price action. we watch the statistical dependence between consecutive returns - completely different game" i asked him to break it down simply he sketched it on a coaster: if yesterday's move has any predictive power over today's move - even marginally, 54% of the time - that's a tradeable edge that edge, scaled through Kelly criterion, snowballs into something unreasonable the data is public - Yahoo Finance and FRED give you everything for free. the math takes a weekend with a textbook the reason retail keeps losing isn't skill. it's that they're staring at the wrong version of the same information "a candlestick chart buries serial dependence. a returns series exposes it completely" went home, ran serial correlation tests on 4 years of QQQ data found 3 regimes - statistically significant, all tradeable on a weekly horizon the signals aren't flawless, correct 56% of the time. Kelly says that's more than enough the data was sitting there the entire time, the method is in every introductory statistics course. nobody ever showed retail where to look they kept you drawing trendlines

@lorden_eth didn’t know /doctor was a thing, this is useful

Most people burn through Claude tokens without even realising it They keep chatting with Claude like it's WhatsApp A few long conversations later, they're wondering why it's slow, forgetting things, and giving worse answers Three commands I wish I knew earlier: > /DOCTOR shows what's eating your tokens and helps you clean them up > /Model OpusPlan uses Opus for planning and a faster model for coding > /COMPACT summarises long chats so Claude doesn't reread the entire conversation every time you send a new prompt They take less than a minute to use People pay $500 to learn the same from courses Bookmark this

@Dexonfxf this feels like the real ai adoption wave honestly

THE BIGGEST AI OPPORTUNITY ISN’T BUILDING THE NEXT CHATGPT - IT’S HELPING MILLIONS OF BUSINESSES USE THE AI THAT ALREADY EXISTS Everyone wants to build the next billion-dollar AI startup. Meanwhile, millions of businesses are still operating the same way they did years ago - manually answering emails, handling customer support, scheduling appointments and spending hours on repetitive work. The technology to solve these problems already exists. The real opportunity isn’t creating another AI model. It’s helping businesses integrate AI into their daily operations, save time, reduce costs and increase revenue. That’s why I believe the next wave of successful AI founders won’t necessarily be the ones building new foundation models. It will be the people who know how to apply existing AI tools to real businesses that are willing to pay for results. Follow @Dexonfxf for more AI business insights.

Risk management rule I actually follow: if max loss on a position hits 1.5% of portfolio, I review it. If it hits 2%, I close half regardless. Sounds mechanical. It is. That's the point - remove discretion when it matters most.

@web3_FBI 14k/mo from an ai calorie app is honestly pretty inspiring

Two people. One AI calorie app. [$14K]/mo. My partner and I built a real app - Calorize, an AI calorie tracker - mostly by directing agents The build almost cost more than it earns - until we stopped obsessing over the effort slider everyone's posting about and fixed the knob they skip: the model. Effort = how hard it tries. Model = what it knows. We routed both per stage and the bill fell [$210 → $40]/wk for the same output. The split, the 5 prompts, the numbers ↓

@scrygg local models making editing workflows this smooth is pretty wild

ONE CSV FILE TURNED A VIDEO TRANSCRIPT INTO A LOCAL AI CLIP FARM In 75 seconds, the creator shows the real AI workflow hiding inside video editing 1 long interview 1 Premiere transcript 1 CSV export LM Studio Gemma running locally on an AMD Ryzen AI chip Then the old problem shows up clip farming usually means manual work scrubbing timelines guessing hooks copying timestamps finding the same sentence again and turning one long video into usable reels by hand This is where the AI layer takes over the transcript becomes data the local model reads the full file pulls out reel concepts writes hooks keeps timestamps then sends the editor back to the exact line in Premiere The payoff is not the laptop It is the workflow 0 cloud tools 1 local model a transcript turned into ready-to-cut clips in seconds