Building Mixar @mixie3D | IITD '21
- Tweets 221
- Following 133
- Followers 27
- Likes 1,284
You can make your movie as you visualise it
Steve Jobs explains why Pixar had to “edit the movie before making it” because animation was too expensive to get wrong.
Mixar is pushing that logic further: if agents can work inside 3D, you don’t just make planning cheaper.
You make production cheaper to iterate too.
17
You would love to see this
Disney would've loved this in 1995.
Today we're launching @mixie3D. (And yes, we made this entire film in it.)
Making 3D has always meant one person doing a dozen jobs. Now a team of AI agents does them at once.
Tell Mixar what you want. Agents model, texture and light the scene at the same time, and you can step in and change anything.
4
107
Natural way to interact with 3D scenes
asked Astra to build an addon that tracks your hand movements through your webcam so you can grab and throw objects inside a 3D viewport, live, with actual physics.
no mouse. no clicking gizmos. just... reach in and grab it.
first time it worked, my brain just hurt.
65
Very fun video to watch. Go check it out
@MondRubberduck Posted an amazing video on how artists can save time using Mixar.
Go check it out!
piped.video/GuhdArwAR_o?si=4QL4…
51
This is what I think the Apple iPhone Ultra would look like. Visualised in 3D using Astra in Mixar .
1
4
251
You can't run the engine of a Ferrari on a Toyota Corolla chassis. You need the right harness to tame the power of that engine.
Frontier LLMs are like a powerful engine.
Harness engineering is very underrated. Having the right harness can unlock the full capabilities of a model.
Harnesses often get dismissed as just scaffolding, just prompt engineering, and not real research. But that couldn't be farther from the truth. The same model weights that score 30% on ARC-AGI score 95% with a better harness.
So we gathered a group of researchers and founders working at the frontier to do a deep dive into the state of harnesses.
We cover how we got to this point, the case for making your harness as expressive as possible, and what YC learned building an agent for every employee in the company.
00:00 - @FrancoisChauba1: Why harnesses matter
04:27 - Building an auto-researcher by accident
07:13 - A five minute history of harnesses
13:56 - Self-improving harnesses
18:35 - @sethkarten: Prime Agent, a self-improving RLM harness
21:50 - Context as an L1, L2, L3 cache
24:51 - From Turing machine to von Neumann computer
28:33 - Messaging between agents
30:04 - ARC-AGI results
33:09 - Emulator Bench and GPU kernels
37:30 - @JonSaadFalcon: OpenJarvis, personal AI on personal devices
38:26 - How far behind are local models
39:21 - The five primitives of a personal AI stack
42:47 - Letting cloud models optimize your local stack
43:53 - 800x cheaper than the cloud
45:58 - @josh__france and @jbellregan: QM, YC's agent harness for work
47:29 - A history of YC's internal agents
49:24 - OpenClaw and a fleet of 50 agents
51:04 - Pulling the brain out of the sandbox
54:43 - Letting the agent choose its own sandbox and model
57:16 - The grind tool: budgets on goals
58:50 - Agents don't understand social context
50
Astra knows what the artist is trying to make
I build AI tools inside Blender. This gap surprised me.
Left: Claude Fable 5.1
Right: GPT-6 Astra
Same brief: a forest path in Blender, 12 seconds each.
Fable’s output is very subpar compared to Astra. My guess is that Anthropic just hasn’t put much Blender data into its training corpus yet.
It’s not that the model is dumb; Blender simply isn’t part of its distribution yet.
Similar to what Karpathy has said about earlier models being very bad at chess. Then chess data went into the next training run, and they improved immensely.
The same thing will happen here.
1
81
