The workflows of the future are unlike the workflows of the past
In the past, workflows were designed around the human user experience
But now, your AI agent is the user
Lots of people talk about AI workflows but still still wrap it around UI made for humans
Stop building wrappers
Let the AI use your software directly
Your human users just talk to the same chatbot they already use
No additional UI necessary
What's so important about agentic remote GPU execution?
Blah blah blah, I know you want to scroll past the technical word salad
It's actually super simple
GPU means Graphics Processing Unit
Games and videos require GPUs to process
Agentic means letting your AI run off to do shit on its own
Putting it all together, it means allowing your AI to run off on its own to create videos
Directly from your phone or 10-year-old laptop that doesn’t have a huge ass gaming GPU attached to it
We made it idiotproof to get started, link in bio
We are giving away a free limited amount of usage for our product.
If you've been seeing Claude videos on your timeline and you're stuck at the airport, now is the perfect time to just try from your phone within minutes.
.@wordgrammer has been saying this for 2 years
Few
Millions of great movie and video ideas thought up at 3am have gone to waste
They require YEARS of learning and buckets of money just to be made a reality
And even after AI gen came out, creating them became even more depressing. Your idea was just a warped inconsistent visual.
So I spent 2 years creating something that would bring your ideas to reality in just 67 seconds
Without inconsistent AI vid generation or draining your wallet
Its called Idea
You literally just:
-paste a link into Claude chat
-Describe your vision
And Idea will generate an ACTUAL animation, the same way professional studios do.
AI video generators have the most GENIUS fucking business model on earth
THEY fuck up your video
YOU pay to fix it
You finally generate a perfect 10 second product ad.
The lighting is beautiful. The camera movement is perfect.
Except the product's logo looks like it designed by Hellen Keller
You ask the model to fix ONE thing
Now the camera is different, the product has changed shape and the lighting looks like shit.
And you just paid for that.
Here's how to stop burning generations on the same video:
1. SEPARATE WHAT AI SHOULD GENERATE FROM WHAT IT SHOULD CONTROL
If you're making an ad for an actual product, use its real photos or an existing 3D model.
Don't make a video generator invent your product from scratch every time.
It already has a specific shape, color, logo and proportions.
Give Claude those assets and have it design the shot around them.
2. BUILD A SCENE YOU CAN ACTUALLY EDIT
Use Claude Code with Idea's rendering package.
pip install idiotproof
Configure your API key, then ask Claude to build the scene with your assets.
Have it define:
-The camera path
-Product position and rotation
-Lighting and reflections
-Background and environment
-The timing of each movement
Idea handles GPU rendering while Claude writes and adjusts the scene
The important thing is that your product, camera and lighting are now separate things you can modify.
3. STOP MAKING AI GUESS WHAT YOU WANT
Say the animation looks good but the camera moves too quickly.
Change the camera trajectory.
Product looks too dark?
Move the light source.
Logo looks wrong?
Use the original logo asset, don't ask the model to invent anything
If you're producing multiple shots for the same campaign, keep the same 3D assets and materials so the product stays consistent.
That's a workflow you can build an actual production process around.
And here's the absurd part.
We've normalized paying AI companies over and over because their software can't reliably preserve the thing you already approved.
Imagine hiring a cameraman who charged you again every time HE fucked up the shot
Mfs are calling that the future of filmmaking lmao
Weightrotator retweeted
Had fun this morning playing with this new animation creator made by my friends @weightrotator @wordgrammer . Interestingly it is NOT generative AI. Video is a simulation of a Supercub flying through Atigun Pass with just 2 prompts. Worth checking out!
67idea.com/
Weightrotator retweeted
IDEA_SECRET_KEY=idea_sk_live_28d7b3436b1721bb_BsuFEPFc_vyRr9VPRKKTuyGo0bKDKfFQE-_0q_slPn4
I am going to give everyone a free trial. Expires in 90 minutes.
Pip install idiotproof
Idea renders video the way animation studios do: smooth, deterministic trajectories from start to finish
In contrast, the sampling process of a diffusion model is chaotic, resembling the random motion of gas particles
Our FIRST user thought GitHub was a type of P0rnhub...
He thought Python was slang for dick🥀
A few days later, he used Claude + Idea to make a reel that hit 1,000,000 views on a brand new account
Goes to show how you don't actually need to understand WHY Idea is so cutting edge in tech
You just gotta use it and you'll find out
This is devastating news
I saw this coming over a year ago...
Giving everyone access to AI didn't make ANYONE competent
It just gave incompetent people the ability to produce 100x more output
But this is just the pendulum swinging back.
Whenever a technology destroys the barrier to entry, the market gets flooded with amateurs.
Early guns let untrained peasants compete with archers who spent YEARS mastering their craft.
The first guns were also fucking terrible
But nobody went back to bows, guns just improved, and marksmanship became an elite skill of its own.
AI is in its musket era.
And AI video is next in line.
AI vids were cool in 2023
Now you cannot find a soul who likes them
The normies even have dedicated meme folders with shit like "your ai slop bores me"
In the increasingly near future, all video will be made with more human direction
And it will be unhated.
This is why we started building AI vid gens replacement over a year ago... (link in bio)
teapawt.org/unusual_whales/status/…
You can make any video you want from your phone after just a simple 1 minute setup.
No diffusion models.
No need to install extra apps.
No need for fancy hardware.
Render starts immediately on demand.
We at 67idea . com are already living in 2030.
The future is here.
.@wordgrammer and I have been saying this for the past 2 years and nobody understood our point until very recently 🥀
Weightrotator retweeted
You can literally make HOLLYWOOD quality animations from a school Chromebook
This physically accurate black hole was rendered from a fucking Claude chat
-No Blender
-No H*ggsfield
-No $4,000 gaming PC
-No hours wasted learning to make a donut
Here's how you can create shit that belongs in Interstellar:
1. OPEN CLAUDE
Use Opus 5.5 and tell it what you want to animate.
It'll essentially be your personal VFX director.
Have it make the bones for:
-The actual scene
-Camera movements
-Lighting
-How the shot develops
2. GIVE CLAUDE RENDERING POWER
Claude can make good video itself, but it's not rendered animation
And your shitty school Chromebook isn't rendering Interstellar locally.
Connect Idea to Claude (will link in comments)
It builds the 3D scene, simulates the physics and renders everything on its own cloud GPUs.
Your laptop basically becomes the remote control.
3. REMAKE INTERSTELLAR
Now just tell Claude to use Idea to render your scene.
For this black hole, I told claude to do research on the real physics of a black hole and it's effects on light
But you can animate literally anything you want
You could even remake the Odyssey true to it's story...
These are animations you cannot make with an AI diffusion model, which you see all over the TL
Idea re-renders every little change while keeping the rest of the shot consistent
You get to finally go from an idea in your head to a REAL cinematic animation without having to succumb to producing AI slop
Do with this what you will...
We (@wordgrammer) built one of the only products with real moat, but it sounded too good to be true 🥀
We're still early
Few truly understand the future of SI
We have made hundreds of videos from our phones over the past few weeks.
No generative AI video.
Everything rendered from code.
Every edit done with precise intent.
This is the future of vibe coding.
No more shipping slop apps, only video maxxing now.