Yanik

claude rated its own 3d cartoon 4.2/10. then a critic agent pushed the same short to 7.9 for free it's called make-short. one skill, one command: /make-short <your idea> 1. measure -> breaks a short you love into numbers: shots, colour, brightness, story steps 2. story -> writes a paradox hook, escalation, payoff and twist, fact-checks the real history, waits for your approve 3. voice -> times every word of your own recording for free (or elevenlabs). the voice is the clock 4. build -> writes the shots, acting and blender scenes in code. zero blender skills needed 5. gate -> renders one still per shot and checks it by numbers and by eye before the full render 6. critic -> a fresh agent scores the render with the exact second of every lost point. first render: 4.2/10 7. fix -> fixes only what the critic confirmed and renders again. same short: 7.9/10 no 3d animator. no editor. no story writer. just claude code, blender and one skill bookmark this before you storyboard your next short

Everyone is talking about Fable 5.2 and Mythos 5.2. The models are real, they dropped yesterday, and Anthropic launched its IPO roadshow the same morning Whether the benchmark gains translate to your actual work is a different question What they actually do: -> Fable 5.2 is Anthropic's new coding and reasoning model, priced at $8 / $40 per million input/output tokens -> On high effort, Fable 5.2 nears Opus 5.5 Max on the FrontierCode 2.0 benchmark at roughly half the task cost -> Mythos 5.2 is the frontier model, built for the hardest work in coding, finance, and the sciences The limits: Anthropic hasn't published granular benchmark tables yet. The FrontierCode 2.0 result comes from third-party analysis, not an official leaderboard. "Nearing" is not "matching" - especially at the tail of hard problems. And both models only launched yesterday, so the community hasn't stress-tested them at scale. What to watch: the effort setting. Anthropic built flexible tiers into both models - the benchmark result is for high-effort mode. Run a few tasks on medium effort first. If the output quality holds, you just halved your costs. The video below shows the benchmark comparison, the pricing breakdown, and how to switch models in Claude Code on one screen. So don't judge them by the announcement. Judge them by your next hard task. Paste this into Claude Code: "Switch me to Fable 5.2 and run the last three coding tasks you helped me with today, one at a time. For each one, show me the output and score your own confidence out of 10. Then tell me honestly: on which one would you have done better on Opus 5.5?" Did your results change, or did the model just get cheaper?

holy f*ck, claude code finally has a memory that actually works it's called claude-mem [github.com/thedotmack/claude…]. one install, 115k stars: 1. observer agent -> runs beside yours, captures every file edit and command, compresses it with AI into short observations 2. /learn-codebase -> front-loads your whole repo in about five minutes so your agent knows the codebase before you even start 3. mem-search -> finds anything from your project history in three steps: index, timeline, details. tokens only on what it opens 4. <private> tags -> wrap anything sensitive and the observer never stores it. your secrets stay on your machine 5. My Work -> live dashboard in your browser showing requests, progress, and a saved review queue no re-reading the same files. no re-asking the same questions. no lost context between sessions. just claude code and one install bookmark this before you start your next project

you don't need a youtube team anymore. claude code just became one. for free it's called youtube agent skill. one install, 11 skills: /yt-script -> writes your script with 21 hook formulas and scores how strong your hook is /yt-edit -> reads your transcript, finds pauses, filler words and repeats, gives you a cut list with timestamps /yt-comment -> drafts replies in your own tone, sorts questions from praise, ignores trolls /yt-plan -> plans your week: one big video, one simpler one, three shorts /yt-package -> builds title + thumbnail as one package /yt-viral -> scans your niche, finds videos that blew up way above the channel's usual, tells you which hook formula worked no editor. no scriptwriter. no social media manager. just claude code and one plugin bookmark this before you film your next video

holy f*ck, claude code can now run your youtube channel. for free it's called youtube agent skill. one install, 11 skills: 1. /yt-script -> writes your script with 21 hook formulas and scores how strong your hook is 2. /yt-edit -> reads your transcript, finds pauses, filler words and repeats, gives you a cut list with timestamps 3. /yt-comment -> drafts replies in your own tone, sorts questions from praise, ignores trolls 4. /yt-plan -> plans your week: one big video, one simpler one, three shorts 5. /yt-package -> builds title + thumbnail as one package 6. /yt-viral -> scans your niche, finds videos that blew up way above the channel's usual, tells you which hook formula worked no editor. no scriptwriter. no social media manager. just claude code and one plugin bookmark this before you film your next video

cartoons like this pull hundreds of thousands of views on TikTok some accounts make $1,000+ a week posting them I'm not an animator. my Claude skill made all 5 episodes episode 5 just dropped. a full minute this time. how the skill built it: > grandpa's "IMPACT PROTOCOL" opens the episode, with 12 AUG 1943 circled in red > Max knocks on every wall of the bunker. "SOLID", "SOLID", "SOLID". then he stomps on the floor and it answers "TONK" > under the old 1943 crates sits a hatch marked B-2, locked with a 4-digit code > knocks on the hatch, the light pours in, and Leo slides down the ladder, terrified > he says nothing. he walks straight to the hatch and types the code without a single wrong key and this is still the rough version ! part 6 is coming. find out what Leo knows the skill and every prompt are in the article 👇

this is pure f*cking gold for anyone who can't draw but wants a cartoon channel a guy finds a secret radio lab behind a jar shelf in his grandpa's 1943 bunker, and every frame of it is code Claude Opus 5.5 wrote cartoon accounts like this pull 742K and 1.8M views a video on TikTok TikTok pays about $0.40-1.00 per 1,000 qualified views, so 1M views is $400-1,000, and my whole stack costs $26 a month how I'd copy it: > give Claude one line: a hero + a clock ("a meteor hits in 5 hours, Max needs a bunker") > it writes the timeline first, then draws every shot on an HTML canvas > mouths move from voice loudness, captions pop word by word > record, convert with ffmpeg, post it as Part 1, 2, 3... part 4 is below. every command is in my article

Insane that over all this time Polymarket users have burned money on fees totaling $270.50M every day Polymarket earns roughly $300K in fees a crazy figure, but if you look deeper it gets more interesting total Payers - 1,543,639, with Avg / payer only $175 one user burned on Fees - $5.09M with a profit of $23.5M that is, the fees for the AVG user are not that big, but for huge "Taker" whales - it's really not little His profile: polymarket.com/profile/0x204… You can see all the data yourself on @PolyScalping : polyscalping.org/stats/fees

this guy made $55.6k on Polymarket by betting against the world falling apart cat-dog joined in June 2026 and has traded 218 markets with $615k in volume no crazy moonshots, his PnL chart is just a slow, steady climb some of his biggest wins: > $28.8k on MicroStrategy NOT selling Bitcoin by May 31 > $19.3k on Xi Jinping NOT being out before 2027 > $15k on China NOT invading Taiwan by end of 2026 >$6.6k on Xi NOT being out in 2025 check his profile - polymarket.com/profile/0x705… what’s interesting is that he often buys “No” at 80-97¢ on dramatic headlines instead of chasing longshots he’ll put $200k+ on a single market to make 7-9%, picking up small, boring wins while everyone else bets on chaos with $42.5k in open positions right now, i’m really curious to see what he fades next getting paid to bet that nothing ever happens is pretty insane ngl

Claude just became a full-on motion designer, and almost nobody’s talking about it it makes motion graphics & edits video like a pro but the sound effects it adds? still terrible here’s the fix: 3 FREE tools that put pro-level sound on your timeline, on their own the right sound, at the right moment, at the right volume, like a real editor did it > EARSHOT: can’t find the sound you need? it generates it from scratch with code. you can even build a whole custom sound pack for your project > REALSFX: it scans your timeline, finds every moment that needs a sound & places each effect on the exact frame > OPENVERSE API: gives the AI access to 5M+ high-quality sound effects. just say “find 3 sharp whooshes & download them” and it does it no sound library subscription. no hours of manual syncing. $0 save this before your next edit!

cartoons like this pull hundreds of thousands of views on TikTok. some accounts make $1,000+ a week posting them I'm not an animator. my Claude skill made all 3 episodes episode 3 just dropped. how the skill built it: > the thing they forgot in part 2 finally gets revealed: the red censor bar wipes off and shows "SOLAR PANELS" > the plan is drawn live on a chalkboard: a route from the bunker to the garage, "120 steps, night only", days 1 to 7 circled as they talk > the old photo of grandpa is all code, down to the scratches, the torn corner and the handwriting on the back: "Viktor knows the other way in." > two voices, real lip sync, and every line has its own emotion: a whisper, panic, a laugh at a childhood memory > it ends with 5 knocks on the hatch, the dog barking, and someone else's fingers on the edge of the lid and this is still the rough version follow the guide from the article and the same cartoon comes out 100x better: a locked cast, real lip sync, numbered scenes you fix in plain words, a render ready to post ! part 4 is coming. find out who knocked the skill and every prompt are in the article 👇

my Claude skill just made episode 2 of a cartoon series. and it ends with the end of the world part 1 got you asking what was waiting in the bunker. part 2 answers with a meteor. how the skill built it: > two characters, two different voices, and their mouths move to the actual loudness of the audio, not a fake loop > every emotion is code: brows drop when Leo yells, sweat runs when Max panics, they jump on the loud words > the stray dog they save on the street is a full rig: floppy ear, wagging tail, shaking when it's scared > the checklist ticks in sync with Max's voice, until the last line turns red. that's the thing they forgot > the meteor hit is layered math: flash, shockwaves, a fireball, a mushroom cloud, debris and trees bending in the blast and this is still the rough way of doing it. follow the guide from the article and the same cartoon comes out 100x better: a locked cast, real lip sync, numbered scenes you fix in plain words, a render ready for TikTok part 3 is coming. you'll find out what they forgot the skill and every prompt are in the article 👇

TikTok accounts make $600–3,000/month on cartoons like this. I made mine with one prompt with Claude "a meteor hits in 5 hours, Max needs a bunker." that was the whole prompt how it built it: > no image generator anywhere: every frame is code, down to the price tags on the empty shelves and the "1943" stamped on the hatch > the story is a timeline before anything gets drawn: 10 shots, each 2–4 s, and a countdown that ticks from 05:00:00 to 00:39 across all of them > the panic scene has two layers of people, flickering lights, a guy carrying a tower of toilet paper and cans rolling across the floor > the captions light up word by word, in sync with the narrator > the last shot goes down the ladder into a WWII bunker, and two eyes open behind the ammo crates that's what ONE prompt gets you. with the full guide from the article, the same cartoon comes out 100x better: a locked hero, lip sync, numbered clips you edit in plain words, and a final render ready for TikTok the skill and every prompt are in the article 👇

his is f**king insane.. One developer spent 10 months building the ultimate Claude Code setup Then he won an Anthropic hackathon with it. And instead of selling it, he gave it away for free 286 skills. 68 subagents. Memory. Security Claude stops being one assistant and works like a full dev team: a coder, a tester and a security reviewer. It also has a skill that learns from your past sessions, so the setup keeps getting better And it’s not locked to Claude Code. It runs in Cursor, Codex and Gemini too 200K+ devs have already starred it on GitHub. Most people are still prompting AI one message at a time. Bookmark this before your feed buries it

guy who won Anthropic’s hackathon just gave away his entire Claude Code setup for free → github.com/affaan-m/ECC 286K stars. top-20 on all of GitHub: → 71 agents, each with its own job → a planner breaks your task into steps → an architect decides how to build it → a reviewer hunts for vulnerabilities → a separate agent fixes your broken build → 311 skills + 97 commands the wild part: it watches what you repeat and turns it into new skills on its own the longer you use it, the better it knows you. one command to install.

Сlaude sound designer bills you for every minute of video 3 open-source tools for $0 you edit videos with Claude - they look like an Apple keynote and sound like a silent movie. the reason is simple: Claude can’t hear sound, it only fakes it -> layer 1: the library > openverse - 5M+ free sounds, Claude can search them even without an API key openverse.org -> layer 2: the placement > real sfx - a free Claude Code skill: finds the moments in your video that need sound and drops in a real recording, frame-perfect > ships with only 78 UI sounds, so start with openverse -> layer 3: the ears > earshot - almost 3,000 GitHub stars. compares your edit with pro launch videos and creates sounds right in code github.com/Leonxlnx/earshot making videos with Claude? save this