HOLY F*CK, Claude Code just built a 30 m walking machine with 160 parts and a 104-bone rig, from one Python script
15 repos to build, rig, motion-capture and animate any 3D model with Opus 5.5, 111K+ stars combined, every link checked
Build. Rig. Capture. Generate. Animate. Ship.
01 AnyMachine → an agent builds machines that move in Blender
github.com/apekshik/anymachi…
02 Blender MCP → control Blender with any LLM
github.com/ahujasid/mcp-for-…
03 UniRig → SIGGRAPH 2025, one model rigs them all
github.com/VAST-AI-Research/…
04 Puppeteer → NeurIPS 2025, auto rigging and animation
github.com/Seed3D/Puppeteer
05 Make-It-Animatable → CVPR 2025, characters ready to animate
github.com/jasongzy/Make-It-…
06 Mesh2Motion → auto-assigns animations to your model
github.com/Mesh2Motion/mesh2…
07 FreeMoCap → free motion capture for everyone
github.com/freemocap/freemoc…
08 MediaPipe → Google's pose, hand and face tracking
github.com/google-ai-edge/me…
09 Kalidokit → turns tracking into avatar rig motion
github.com/yeemachine/kalido…
10 GVHMR → SIGGRAPH Asia 2024, 3D motion from video
github.com/zju3dv/GVHMR
11 4D-Humans → reconstructs and tracks people in video
github.com/shubham-goel/4D-H…
12 HY-Motion 1.0 → Tencent's text to 3D character motion
github.com/Tencent-Hunyuan/H…
13 MDM → motion diffusion, text to motion
github.com/GuyTevet/motion-d…
14 MoMask → CVPR 2024, masked motion generation
github.com/EricGuo5513/momas…
15 three-vrm → VRM avatars in three.js
github.com/pixiv/three-vrm
Model → skeleton → mocap or text → motion → three.js
Bookmark this. How Opus 5.5 turns motion design into writing, in the article below ↓
It is becoming f*cking dangerous: ~90% of the Claude Code team's code is written and reviewed by Claude, the trick is making it prove its work, 7 repos give your agent the same proof loop
1. Drive the real browser from a verify skill, the agent clicks through the user flow via Playwright's accessibility tree before it says done: github.com/microsoft/playwri…
2. Run the same check on the phone, Mobile MCP drives iOS and Android apps on simulators, emulators and real devices: github.com/mobile-next/mobil…
3. Write one rule that must always hold, fast-check throws random inputs at the code and shrinks any failure to the smallest case: github.com/dubzzz/fast-check
4. Test the tests with StrykerJS, it plants small bugs in your code and every mutant that survives is a bug your tests would miss: github.com/stryker-mutator/s…
5. Block production code until a failing test exists, Probity hooks into Claude Code and tells the agent why it was blocked and what to do next: github.com/nizos/probity
6. Turn a mistake the agent keeps repeating into a YAML lint rule with ast-grep, it matches code structure and runs in CI on every diff: github.com/ast-grep/ast-grep
7. Run /hookify from Anthropic's own plugin, it reads the chat, finds what you corrected and writes a rule that warns or blocks: github.com/anthropics/claude…
In the team's internal tests, a separate grader checking Claude's work against a rubric lifted task success by up to 10 points, most on the hardest problems
On the Claude Code team, a check that still fails after 3 rounds stops Claude and goes to a human
Full guide in the article below
HOLY F*CK, a full token diet for your AI agent in 15 repos
15 repos that make your agent read less, say less and pay less, 880K+ stars combined, every link checked
Count. Squeeze. Skip. Shorten. Cache. Local.
01 tiktoken → count tokens before you send
github.com/openai/tiktoken
02 code2prompt → codebase into one prompt, tokens counted
github.com/mufeedvh/code2pro…
03 LLMLingua → compresses prompts up to 20x
github.com/microsoft/LLMLing…
04 TOON → JSON for prompts, fewer tokens
github.com/toon-format/toon
05 MarkItDown → files and Office docs to Markdown
github.com/microsoft/markitd…
06 Jina Reader → any URL to LLM-friendly text
github.com/jina-ai/reader
07 Repomix → a whole repo in one file
github.com/yamadashy/repomix
08 Serena → reads symbols, not whole files
github.com/oraios/serena
09 Claude Context → code search, ~40% fewer tokens
github.com/zilliztech/claude…
10 caveman → talks like a caveman, cuts 65% of tokens
github.com/JuliusBrussee/cav…
11 RTK → 60-90% fewer tokens on dev commands
github.com/rtk-ai/rtk
12 GPTCache → a repeat question hits the cache
github.com/zilliztech/GPTCac…
13 Semantic Router → decides by meaning, no LLM call
github.com/aurelio-labs/sema…
14 Ollama → open models on your own machine
github.com/ollama/ollama
15 llama.cpp → local LLM inference in C/C++
github.com/ggml-org/llama.cp…
Count → squeeze → read less → say less → cache → run local
Bookmark this. How to run an AI employee for almost nothing, in the article below ↓
HOLY F*CK, a full video editing room for Claude Code in 14 repos
14 skills and tools that cut your video while you're away from the laptop, 160K+ stars combined, every link checked
Watch. Cut. Caption. Grade. Render.
01 video-use → raw clips in, final.mp4 out
github.com/browser-use/video…
02 ffmpeg-skill → 42 local FFmpeg tools, no cloud
github.com/kajisho5/ffmpeg-s…
03 Claude Eyes → agents that watch your footage
github.com/moritzkremb/claud…
04 WhisperX → word-level timestamps and speakers
github.com/m-bain/whisperX
05 Silero VAD → finds where speech starts and stops
github.com/snakers4/silero-v…
06 auto-editor → cuts every silent gap automatically
github.com/WyattBlue/auto-ed…
07 PySceneDetect → finds scene cuts and transitions
github.com/Breakthrough/PySc…
08 editly → a whole edit in one spec file
github.com/mifi/editly
09 OpenTimelineIO → open interchange for timelines
github.com/AcademySoftwareFo…
10 DaVinci Resolve MCP → Claude controls Resolve Studio
github.com/samuelgursky/davi…
11 Uncut → macOS editor with Claude Code inside
github.com/blazejkustra/uncu…
12 Yap Studio → agent cuts dead air, adds B-roll
github.com/cathrynlavery/yap…
13 Take One → one line in, a launch video out
github.com/talsiach/take-one
14 FFmpeg → the engine under almost every tool here
github.com/FFmpeg/FFmpeg
Footage → silence cut → word timing → scene cuts → grade → render
Bookmark this. How Opus 5.5 turned motion design into a writing skill, layer by layer, in the article below ↓
HOLY F*CK, a full 3D machine studio for Claude Code in 13 repos
13 skills and tools to use with Claude Code, 310K+ stars combined, every link checked
Model. Explode. Animate. Score. Render.
01 Open 3D Skills → scenes, exploded views, x-ray cuts
github.com/Moh4696/open-3d-s…
02 Machine Skill → one line in, an interactive 3D machine out
github.com/AirsupHQ/machine-…
03 Anatomy → explains ideas by building working objects
github.com/wheresryan22/anat…
04 Blender Skills → 28 pro Blender workflows as skills
github.com/EL4CTEO/blender-s…
05 Manifold → clean solid 3D booleans
github.com/elalish/manifold
06 three.js → the 3D library most of these run on
github.com/mrdoob/three.js
07 Claude Animated Shorts → 2D characters in a 3D world
github.com/mromidev/claude-a…
08 Cinematic Skill → films drawn by code, one continuous camera
github.com/Moh4696/cinematic…
09 reanimate → copies an animation, proves it moves the same
github.com/NiazMorshed2007/r…
10 Motion2Code → motion video into editable SVG + GSAP
github.com/lovisdotio/motion…
11 opus-sound-layer → 1,139 CC0 sounds and a mixer
github.com/Bodila51/opus-sou…
12 HyperFrames → write HTML, get a video, built for agents
github.com/heygen-com/hyperf…
13 Remotion → make videos with React
github.com/remotion-dev/remo…
Model → take apart → animate → score → render
Bookmark this. The 7 layers behind viral Opus 5.5 videos are in the article below ↓
It is becoming f*cking dangerous: Anthropic merges 12x more code changes a day than in May, and its Claude Code team says your old CLAUDE.md can make Claude worse
Lydia Hallie, Claude Code team at Anthropic:
"Anthropic went from about 2500 changes a day in May to more than 30,000 a day in September"
82 sec from her AI Engineer talk, 200K views in 13 hours
"We've now essentially become managers for dozens of sessions"
"You may still have things like IMPORTANT or NEVER DO THIS… in your CLAUDE.md file. But with today's models, you really don't need that and it can actually make the model go off track"
"Only about 14% of people caught a clearly dangerous command. Auto mode blocked 89% of them… But about one in 10 of those commands still went through"
She just reminded everyone to run /doctor prompt-audit. This clip shows why
Full talk in the reply
unavailable
plus retweeted
Introducing Wave.
The world's first AI that reads your customers' mind.
$1.47 trillion will be spent on software this year, but only a fraction of software features are used by customers.
AI can build anything you want, but it doesn't know what your customers need.
Wave fixes that.
plus retweeted
In 2015, I built an AI that detects early cancer.
Today, we’re introducing ARC by Rayline: the world’s first unified model for routing AI.
> 97.5% of Opus 5.5 quality
> Up to 1/10th of the cost
> Fully open-source
It picks the right model and the right amount of thinking for every step of your agent’s work.
Try it here and read our technical report below:
rayline.ai
It is becoming f*cking dangerous: Fable 5.2 is already inside Claude Code, and it stays off until you turn it on
1. Update Claude Code to v2.1.321 or newer, the first version with Fable 5.2
2. Type /model fable, or start with claude --model fable
3. Check the Fable row in /model: if it says Requires usage credits, Fable spends credits instead of your plan limits
4. Hand it the ambiguous problems: a root cause hunt, an outage, an architecture call
5. Describe the result you want and let it work out the steps
6. Stop adding "verify your work" to prompts, it checks its own work more often than smaller models
7. Give it a whole task you would normally split into pieces
Claude Code docs say it suits tasks larger than a single sitting
1M context window by default, medium effort to start, API price $8 / $40 per million tokens, down from $10 / $50 for Fable 5.1
Full guide in the article below
It is becoming f*cking dangerous: Apple shipped its own plugin for Claude Code that builds, tests and runs your iPhone app
1. Update to Xcode 27 on your Mac, the plugin ships inside Xcode itself
2. Open Claude Code in your project and run /plugin install xcode@claude-plugins-official
3. Open the project in Xcode and allow the agent when Xcode asks the first time
4. Ask Claude to build the app and launch it on the iPhone simulator
5. Let it drive the simulator and check how the app behaves
6. Ask for SwiftUI previews and the build log, it reads the errors and fixes them
7. Run Apple's own skills: VoiceOver and Dynamic Type specialists, XCTest to Swift Testing
Author on the listing: Apple Inc., added to Anthropic's official marketplace on Nov 6
Claude Code pulls the plugin from your local Xcode every session, so it updates with Xcode
Full guide in the article below
This is getting out of hand: everyone is making showreels with Opus 5.5 and sleeping on the ~450 ready-made motion blocks that make them look pro.
HeyGen dropped HyperFrames on GitHub, a set of skills that plug straight into Claude:
-> github.com/heygen-com/hyperf…
Why are you still asking AI to draw every frame from scratch? Stop treating Claude as a script writer. Start turning it into a full motion studio.
Here is the workflow: install the core set once, describe the video in one sentence, and Claude assembles it from ready blocks (kinetic type, captions, 3D camera moves). It works like a LEGO box for video: you never build a brick yourself, you just snap them together.
Step-by-step setup:
1. Run npx skills add heygen-com/hyperframes and pick "Core Skills"
2. Type /hyperframes plus your idea, like a 10-second product intro with a fade-in title
3. Add /media-use for music, voiceover and color grading, or paste a Figma link into /figma
4. Move old Remotion, Lottie or Three.js projects over with /remotion-to-hyperframes
5. Run /reverse-any-video on any clip and instantly get it back as an editable project, frame by frame
The whole install is one command, and you will never animate a title by hand again.
Full step-by-step guide in the article below
Make sure to save this post for later
Motion designers are cooked. Opus 5.5 doesn't make videos. it writes the code that draws them
I rewound every viral Opus 5.5 video. the same 7 layers sit under all of them:
THE 7 LAYERS
┣ 01 picture · 900 still frames = 15 seconds of video
┣ 02 code · one function draws any moment: seek(t)
┣ 03 sound · every big moment lands on a beat
┣ 04 rules · a CLAUDE.md that bans the AI look by name
┣ 05 reference · cut a video you love into frames. steal the style, never the story
┣ 06 words · one line, one paragraph or one night (9,500 characters)
┗ 07 second look · claude scores its own frames, fixes the 3 worst, repeats until 8+
setup takes 10 minutes:
/npm i -D playwright && npx playwright install chromium
this video is one HTML file claude wrote. every frame drawn in JavaScript, every sound generated in python. 0 cameras, 0 After Effects
6 copy-paste prompts + the full rewind are in the article below ↓
bookmark it before you make your first one
I don't understand why everyone isn't doing this yet: Anthropic's own Claude Code guide lists 7 habits you can stop
1. Stop typing think hard, it's plain text, effort is the control: set /effort or type ultrathink for one turn
2. Delete IMPORTANT and ALL CAPS, shouted rules get over-applied, turn hard rules into a hook or deny rule
3. Cut architecture and file trees from CLAUDE.md, Claude reads the repo, keep only what isn't in there
4. Drop keep going and don't be lazy, it's more likely to cause over-work, give Claude a check to pass, /goal for unattended runs
5. Quit spawning subagents just to save context, most models have a 1M window, keep them for parallel work
6. Stop handing a task to a subagent for a better answer, a fresh reviewer helps because it hasn't seen your reasoning
7. Stop rationing context, it compacts on its own, still /clear between tasks for better, cheaper answers
Opus 5.5 runs at medium effort by default, /effort status shows yours
In the terminal ultrathink turns rainbow when it can raise that turn to xhigh, v2.1.311+
Full guide in the article below
It's pretty wild that people still run Claude Code bare when these 7 repos have over 1,000,000 GitHub stars combined
1. Load 4 coding rules drawn from Karpathy's post (230K): /plugin marketplace add forrestchang/andrej-karpathy-skills
2. Make Claude write the least code that works (186K): /plugin marketplace add DietrichGebert/ponytail
3. Add 7 design skills with 192 palettes (146K): /plugin marketplace add nextlevelbuilder/ui-ux-pro-max-skill
4. Run Garry Tan's own setup, /office-hours to /ship (141K): paste the install line from github.com/garrytan/gstack, needs Bun
5. Cut Claude's replies to caveman length, code stays exact (118K): /plugin marketplace add JuliusBrussee/caveman
6. Add Addy Osmani's 25 skills, spec to ship (114K): /plugin marketplace add addyosmani/agent-skills
7. Shrink the shell output Claude reads (86K): brew install rtk, then rtk init -g
After each marketplace add, run /plugin and install it from the Discover tab
gstack adds a Stop hook and Codex reviews of your diffs, rtk patches settings.json: read both READMEs first
Full guide in the article below
Everyone is obsessed with making AI smarter. Almost nobody is talking about the fact that their AI keeps forgetting everything they teach it.
Claude Code can become something much more interesting than a coding assistant.
A second brain that remembers your projects, decisions, people and notes across sessions.
Here's how to build it:
1. Create a folder called SECOND BRAIN
2. Open Claude Code inside it
3. Add four folders: MEMORY/, notes/, people/, projects/
4. Create a CLAUDE.md that tells Claude how to organize and retrieve important information
5. Use @-imports to load your most important memory files at startup
6. Tell Claude to save key decisions as Markdown files, one topic per file
7. Create a custom /brain skill to search your notes and bring relevant context back
Now watch the video.
Every branch represents stored knowledge. The glowing line visualizes a memory being retrieved.
Something you decided weeks ago can become useful context in a completely new session.
No vector database. No expensive memory app. Just Markdown files and Claude Code.
The crazy part? Your knowledge doesn't have to disappear when the conversation ends.
Full SECOND BRAIN setup in the article below.
I still don't get why everyone isn't fixing this: Claude Code silently deletes your chats after 30 days, one line changes it
1. Open ~/.claude/settings.json and add "cleanupPeriodDays": 3650, 0 fails validation
2. Keep secrets out, transcripts sit in ~/.claude/projects as plain text, check the size: du -sh ~/.claude/projects
3. Name work while you still remember it: /rename auth-fix, later in the same repo claude --resume auth-fix
4. Query an old chat without touching it: claude -p --resume <session-id> --fork-session "list the decisions we made" >> decisions.md
5. Export the open chat: /export auth-fix.txt
6. Ask Claude to add a SessionEnd hook that copies transcript_path into your second brain folder, every chat lands there when it exits normally
7. Run /insights: an HTML report on what you work on and where you hit friction, it spends your tokens
Plans, checkpoints, subagent logs and /insights reports go in the same sweep, no message shown
Transcripts of sessions started or last continued in the desktop app or Cowork are exempt by default
Full guide in the article below
It is becoming f*cking dangerous: one command in Claude Code turns your codebase into a map, 141K stars in 7 months
1. Install with uv, the package has two y's: uv tool install --python 3.13 graphifyy
2. Add the skill to Claude Code: graphify install
3. List tests, docs and images in .graphifyignore: test factories took 6 of the top 10 spots in a repo Graphify reviewed, each image gets its own subagent
4. In a repo you trust, start Claude Code at the root, type /graphify . (tree-sitter parses code with no LLM, Claude still spends tokens running it)
5. Open graphify-out/graph.html and read GRAPH_REPORT.md: most connected nodes first, then surprising connections
6. Ask with words from your code: graphify query "auth token", or trace a path: graphify path "A" "B" --undirected
7. Remind Claude to check the graph: graphify claude install, rebuild the code part on every commit: graphify hook install
Every edge is tagged EXTRACTED, INFERRED or AMBIGUOUS: check the source before you trust the last two
Official repo: github.com/Graphify-Labs/gra…
Full guide in the article below
It is hard to believe, but 66,000 GitHub stars went to a way of running Claude Code without paying Anthropic
1. Open github.com/Alishahryar1/free…, on Mac or Linux run the curl installer from the README
2. Run fcc-server on Linux or open the Free Claude Code app on Mac and Windows, it opens the Admin UI
3. Turn on Proxy Authentication with your own token, by default the proxy is open to your whole network
4. Create an API key at build.nvidia.com, paste it under Providers, NVIDIA NIM, Add key, Save, then click Apply
5. Start the agent with fcc-claude
6. Switch in the native /model picker among providers you set up, 61 in the catalog, free and paid, DeepSeek, Kimi, Groq, Ollama
7. Add Fallback Models so a dead provider rolls to the next one, a failed request can consume usage from more than one
Free routes mean other models, not Claude, Anthropic does not support this, and your code goes to the provider unless it is local
The README claims 1.3B+ free tokens a month, availability and limits are set by each provider and can change
Full guide in the article below
Whoever built this will put video editors out of work: raw footage in a folder, Claude Code hands back final.mp4
1. Paste the setup prompt from github.com/browser-use/video… into Claude Code, it installs itself
2. Export your ElevenLabs key first, the tool is free and open source, you pay for transcription and Claude usage
3. Drop your raw takes into one folder, cd there and run claude, transcription starts right away so test on one short clip
4. Type: edit these into a launch video
5. Read the plan it proposes and say OK, it does not touch the cut before that
6. Wait while it removes umms and dead space, ask for color grade, subtitles and overlays too, each overlay gets its own subagent
7. Approve the preview to get edit/final.mp4, every cut is checked first, up to 3 fix passes
30K+ GitHub stars, MIT license, works for talking heads, montages, tutorials, travel, interviews
Claude reads a ~12KB transcript plus PNG filmstrips, frame by frame would be about 45M tokens
Full guide in the article below








