what a hassle - i fw @rakebitcom
Dm for collabs 👉
Joined August 2024
- Tweets68.4K
- Following576
- Followers13.1K
- Likes115K
S 🫧 retweeted
THIS IS INSANE
most "ai + infra" demos stop at an api call
@linq_ai linqprotocol mcp is different: your client gets a control plane over real k8s compute
- deploy / restart services
- logs, metrics, diagnostics
- networking + vm controls
the agent prepares the deployment or renewal handoff. you still sign the escrow in wallet
that's agents operating infra, not chatting about it
🤝 Paid partnership
CLAUDE + REA CAN REBUILD SOFTWARE FEATURES WITHOUT COPYING A SINGLE LINE OF CODE.
Opus 5.5 uses REA to analyze how a program works, then recreates its features in a completely separate clean-room environment.
Here’s how the workflow works:
Run npx rea-agents setup to connect REA with Claude Code.
Choose a feature and let REA investigate how it works. Save its behavior, logic, and rules in a spec file.
Open a fresh Claude session with only that spec. Rebuild the feature without accessing the original source code.
Use /goal to run the tests and check 3 intentionally broken implementations. All 3 must fail.
One project reportedly recovered 697 functions in just 12 days, with 696 exact matches.
The real power isn’t just reverse engineering.
It’s creating tests that catch mistakes before they make it into the final software.
F**KING INSANE.
A FREE Claude skill just turned a single picture into one of the most detailed 3D hummingbird models I’ve ever seen.
And it only took ONE prompt.
I still can’t believe people aren’t charging for this.
Here’s what it built:
The original picture provided the bird’s appearance, while Three.js handled the 3D model: 300 tiny parts beneath the outer plates and 7 glass vanes on each wing
Opus compared its first render against the reference image and only started the animation once they matched
At 1.7 seconds, the wing unlocks at its root and flips in mid-air as the entire scene slows down to 0.4x speed
The eye separates into 4 rings, revealing a 10-blade iris underneath
The beak sleeve slides back, exposing a 1.8 mm probe that extends 24 mm
A new component opens every 2 seconds, while the exploded view spins in reverse and rises into an overhead pass
This is what a single image and the right prompt can produce.
I’ve got an even crazier render coming next. I’ll share whatever it creates.
The skill, prompts, and live simulator are in the article below.
S 🫧 retweeted
data must be booming if micro1 has an AI agent literally calling companies and offering them $5M+ just to access their internal data.
Meet Kai, micro1's AI partnership specialist.
His job is to get companies paid for their data, fast.
In his first month alone, Kai has:
- Closed $3M+ in data licensing deals
- Conducted 836 partnership calls
- Saved nearly 400 hours of meetings
Calls with Kai take just 7–10 minutes, and eligible companies can go from that conversation to a six- or seven-figure licensing offer in as little as 36 hours.
These partnerships help us build RL environments where frontier AI learns from the work happening inside real companies.
Learn more in comments below.
S 🫧 retweeted
THIS VIDEO COST HIM $0 TO EDIT
No Claude.
No ChatGPT.
No credits.
No editor.
A free open model did the entire edit.
→ Install OpenCode (free agent that reads the same skill folders as Claude Code)
→ Download the Editor V1 skill (built once in Claude, then reused for free)
→ Set the model to GLM 5.3 Flash Free
→ Paste the setup prompt so it installs FFmpeg, Python, and the rest
→ Drag in your raw video and type “edit this video with editor V1”
→ It cuts, adds transitions, graphics, captions, and sound all on your laptop
1.6 million tokens.
$0.00 spent.
No account.
No card.
Bookmark this before you pay for another editing tool.
S 🫧 retweeted
🤯 WHAT IF YOU COULD BUILD AN ENTIRE MEDIEVAL CITY WITH AI?
Someone is using Claude Opus 5.5 + Blender + Unreal Engine to create a fully explorable walled city.
And the crazy part?
Every street is walkable. Every district is planned. Even the smallest details have a place.
Here's how the workflow works:
01. Connect the tools
Run Opus 5.5 on high effort and connect Unreal Engine and Blender through MCP.
02. Design the city first
Map out the walls, gates, main roads, districts, and central plaza before building anything.
03. Build a modular asset library
Use Blender to create reusable walls, towers, rooftops, houses, and market stalls.
04. Generate architectural variations
Create 10–15 different house designs using the same modular kit, with different materials, roof colors, and structures.
05. Populate the districts
Fill the streets with houses, gardens, orchards, and market areas.
06. Place the landmarks
Add the cathedral, fountains, and marketplace exactly where they belong on the original map.
07. Create the player character
Generate a character using Tripo AI and bring it into Unreal Engine.
08. Explore and refine
Switch to third-person mode, walk through every street, identify problems, and fix anything that looks wrong up close.
BUT HERE'S THE MOST IMPORTANT PART:
Don't start by building houses.
Start by planning the roads.
The streets determine where everything belongs. Once the layout is established, the buildings, districts, and landmarks can fit naturally into the city.
Claude handles much of the construction.
You decide the city's vision, choose the important landmarks, and judge whether the final world actually feels alive.
We're getting closer to a future where describing a world is enough to start building it.
AI-assisted game development is getting seriously wild.
🔖 Bookmark this if you're experimenting with Unreal Engine, Blender, or AI-powered worldbuilding.
THIS IS F**KING INSANE.
CONSOLE GAMES ARE NOW RUNNING NATIVELY ON PC WITHOUT EMULATORS.
Recomp Store just launched its PC section, bringing native ports of games from PlayStation, Nintendo, and Xbox to PC.
Here’s what you get:
→ Native performance with higher FPS and modern resolutions
→ No emulator required
→ Support for mods
→ One place to access all Recomp projects
→ A dedicated desktop launcher is coming soon
Gaming on PC just got a whole lot crazier.
Link below 👇
S 🫧 retweeted
THIS IS F**KING INSANE!
No camera. No mic. No face on screen. The whole pipeline can run on AI.
My version of the workflow:
Pick a niche
Research what people already search for.
Have Claude draft the hook, script, retention beats, and outro
Turn the script into AI narration
Add generated B-roll and visuals
Edit, publish, repeat
The first video isn't the win. The win is what comes after: you find a format that gets views, then make variations and let the analytics tell you what to improve.
You stop being a creator and start running a content engine.!!
S 🫧 retweeted
THIS IS F**KING INSANE
THIS IS HOW PEOPLE ARE QUIETLY BUILDING FACeless KIDS CHANNELS WITH AI.
the videos look like they came straight out of an animation studio.
but the workflow can be almost entirely AI.
start on YouTube.
find a kids channel already pulling millions of views and study what keeps viewers watching:
→ bright characters
→ simple songs
→ repetitive stories
→ looping animations
→ recognizable formats
then turn those observations into original concepts with an AI animation tool.
generate the characters.
build the backgrounds.
animate the scenes.
create the little dance loops.
without drawing a single frame.
an AI voice can handle the nursery rhyme, counting song, narration, or whatever format you choose.
then stitch everything together and publish on a consistent schedule.
no camera.
no drawing tablet.
no traditional animation studio.
just a repeatable production pipeline running from a laptop.
and because kids content can generate significant repeat viewing, a strong format can keep bringing viewers back long after the first upload.
one person can build an entire content library without ever appearing on camera.
just remember: study successful formats, don’t copy another creator’s characters, stories, or visual identity.
full guide + pipeline in the article below.
S 🫧 retweeted
Claude Opus 5.5 generated 100 different motion effects on a single character in just 30 seconds.
and here’s the interesting part: every frame shows a different effect.
you can pause the animation anywhere and see how the character transforms through different visual styles, effects, and motion treatments.
think about what this means for motion designers.
creating 100 individual effects manually could take hours of work. this workflow generates them directly from code.
here’s how to try it:
→ download the PDF containing the motion design guide.
→ drop it into Claude Opus 5.5.
→ ask it to follow the instructions and generate the effects.
→ experiment with the prompts and customize the results.
the real opportunity isn’t just generating animations faster. it’s using AI to explore visual ideas that would take much longer to create manually.
read the full motion design guide in the article below.
save this if you’re experimenting with AI-powered motion design.
S 🫧 retweeted
BNB Chain has memes. It has games. It has tokenized gold.
But what happens when you bring all three together?
I've been looking into $BUBU, and I think the concept is worth paying attention to.
BUBU is building a gaming ecosystem around a golden bull, combining Tamagotchi-style progression, virtual homes, characters, fashion shows, kart racing and collectibles
But the part that really caught my attention is the gacha mechanic.
Through Golden Bull 金牛, players can open collectible NFT packs, chase different rarities and discover what's inside and these aren't just ordinary NFTs
Powered by ERC-6551, each card can have its own onchain account capable of holding assets like $BUBU and XAUt (Tether Gold)
That's what makes this particularly interesting:
the possibility of discovering real tokenized gold inside a collectible card, rather than just receiving another in game currency or digital image
You can collect cards, keep them sealed, trade them or open them to access their contents.
The bigger vision is to combine gaming, collecting and onchain asset ownership into one ecosystem.
My thesis is pretty simple:
- Memes bring attention
- Games create engagement
- Collectibles give people something to chase
- tokenized gold adds an entirely different dimension
Crypto gaming needs more interesting ideas, and BUBU is exploring one worth watching.
Gaming meets gacha. Gacha meets gold.
bubu.gold (bubu.gold/)
$BUBU | BNB Chain
THIS HAS TO BE ILLEGAL.
Claude Opus 5.5 just built a fully interactive 3D robot scorpion simulator from a single image.
I gave it one render of a robot scorpion and told it to recreate every single part in code, then make the entire thing move.
Here’s what it built:
8 legs moving in a scuttle gait at 2.7 steps per second, with every foot landing in its own position
2 claws with blade like fingers that snap shut at 100 N on the grip gauge
A 5-segment tail that unlocks its blade and whips around to 167° during a strike
300 parts and hundreds of bolts beneath 7 armour plates, with something coming apart every 2–3 seconds
A full 360° turntable animation where the entire machine breaks apart and snaps back together before the rotation ends
The final result is a 26-second sequence that ends with the stinger just an inch away from the camera lens.
All of this started with ONE image.
The skill, prompt, and live simulator are below.
ISN’T THIS JUST INSANE?
And no, it wasn’t generated by AI.
I had Opus 5.5 write the entire thing in code.
Three.js is seriously amazing.
I used to make a living building websites, but I honestly can’t imagine creating something like this from scratch.
The entire webpage is built almost entirely with JavaScript.
Imagine how painful it would be to write all that crazy code by hand.
The original post includes instructions on how to build it, so if you’re interested, check it out and give it a try.
Pepe’s stepbrother is making moves
$DUCHI launches november 5th.
5 PM UTC | 12 PM NYC | 9 AM LA
TG gets the CA first.
TG: t.me/realduchi
X: @duchimeme
S 🫧 retweeted
every freelancer knows the pain of clients asking for just one more thing before launch.
@_MonoDesk built an ai co worker to handle the business side of freelancing.
→ checks requests against agreed project scope
→ flags extra work and drafts client replies
→ creates proposals with phase 2 quotes
→ manages projects through a shared client portal
→ works with claude and chatgpt through mcp
less admin. fewer scope headaches. more time for creative work.
and it’s free.
🤝 Paid partnership
