Physicist, Programmer, Designer
Joined December 2024
- Tweets 2,292
- Following 90
- Followers 10,275
- Likes 23
1,923 Photos and videos
The 2025 forecast just became reality.
The US government pulling the plug on Fable 5 for foreign isn't a glitch. We've officially transitioned from AI tools to "state-controlled weapons."
- Mid 2025: simple agents (ok)
- Late 2025: The $100B compute arms race begins (ok)
- 2026: AI starts upgrading itself. (ok)
> May 2027: Mandatory security clearances < WE ARE HERE
- Late 2027: Full-on nationalization of frontier models (Next up)
the crazy part is that it’s not even 2027 yet.
ai-2027.com/
Jun 13
The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.
The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.
Access to all other Claude models is not affected.
We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.
Read our full statement: anthropic.com/news/fable-myt…
1
11
1,049
Jun 12
VideoMDM- builds high-quality 3D human animations using regular 2D videos instead of expensive motion capture suits.
- Text-to-Motion
- Cheap scaling
- Captures complex motions
videomdm.github.io/
9
46
2,826
Jun 12
LTX-2.3 Sci-Fi Cinema LoRA
Nice cinematic science-fiction style LoRA for cities, alien worlds, spacecraft interiors etc.
Makes good cyberpunk environments
huggingface.co/SOLRICKS/ltx-…
1
4
32
2,060
Jun 12
Surflo. turns a few standard photos into clean 3D models for game dev/VR.
- creates usable meshes
- geometric consistency
- predicts millions of oriented surface points
- beats VGGT-1B and DepthAnything-3
anttwo.github.io/surflo/
4
45
320
14,280
Jun 12
MoVerse: Single-image to interactive 3D world modeling.
It doesn't just guess what's behind you; it builds a permanent scaffold so the room doesn't shift when you turn around.
- expands NFOV photos into 360 navigable scenes
- Wan2.1 Pano DiT
- TAEHV (fast VAE decoder), 8 FPS on an RTX 4090
- stable geometry
- persistent spatial memory.
perfect for real estate & design
orange-3dv-team.github.io/Mo…
22
116
5,402
Jun 11
Chatterbox Multilingual V3. General-purpose 0.5B TTS model
- improved speaker similarity
- low hallucinations for cross-lingual voice cloning
- more natural, conversational speech
huggingface.co/ResembleAI/ch…
1
16
120
5,447
Jun 11
Orion4D FXMax, Professional color grading and sharpening suite for ComfyUI
- split-screen previews
- instant slider adjustments
- cinema-grade Tone Mapping
- Catmull-Rom Curves
- skin tone protection
- film grain
- edge-aware sharpening
github.com/orion4d/Orion4D_F…
1
8
62
3,524
Jun 10
Getting around to building these libraries at last:
> causal_conv1d 1.6.2.post1
> NATTEN 0.21.6
for torch 2.12/2.13, cuda 13.2/13.0, python 3.14/3.13
github.com/wildminder/AI-win…
1
1
23
2,022
Jun 10
DiP: new pixel-space model, no VAE needed.
- Crisp textures without compression artifacts.
- 10x faster than prior pixel models
- 1.79 FID on ImageNet
huggingface.co/zhen-nan/DiP
1
11
102
6,114
Jun 10
MilliVid - long AI videos that stay consistent from start to finish.
- maintains object permanence across hundreds of frames.
- ViT-B AE DiT-B architectures.
davidcharatan.com/millivid/
2
6
43
2,452
Jun 10
Lip Forcing. Real-time autoregressive lip synchronization.
- causal streaming with 2-step denoising
- 40x speedup vs OmniAvatar
- fits in 9GB VRAM
- Wan 2.1
- low latency
Face stays natural, mouth matches the audio perfectly
cvlab-kaist.github.io/LipFor…
6
13
159
6,583
Jun 10
Interesting one from Alibaba.
ABot-Earth - converts flat satellite images into interactive 3D worlds.
- brings high-detail 3D views to regions that currently only have flat 2D maps.
- generates at 10 mins/sq km
- OGC 3D Tiles compliant
- makes city-scale digital twins UAV simulation.
abot-earth.amap.com/?tab=pla…
5
75
4,524
Jun 9
UniSHARP - universal sharp monocular view synthesis.
- supports perspective, fisheye, 360°
- 3.1s inference
- handles hidden areas with dual-layer geometry
- uses UniK3D 3DGEER
- SOTA fidelity.
insta360-research-team.githu…
4
52
3,088
Jun 9
5-Day AI Agents course! it's free and Google-backed!
It’s perfect if you’re tired of just chatting with AI and want to start building actual autonomous workers.
- the "Vibe to Live" focus
- day 3 is the gold mine. Long-term memory
- only an 1-2h a day for a week.
kaggle.com/competitions/5-da…
2
10
1,092
Jun 9
well Claude Fable 5 is basically a "leave it alone and let it work" model.
> Elite Autonomy. It solves real-world GitHub issues with far less human "babysitting" or steering.
> Massive Coding Leap
> Self-Correction! Intelligent self-verification, so it catches its own bugs before handing the code back to you.
> Cybersecurity Powerhouse. It’s a beast at finding and patching vulnerabilities
> Spatial Reasoning. Much better at understanding complex system architectures and layouts.
> Efficiency. Uses tokens more intelligently
Say hello to your personal senior engineer
Introducing Claude Fable 5: a Mythos-class model that we’ve made safe for general use.
Its capabilities exceed those of any model we’ve ever made generally available.
1
5
15
1,534
Jun 9
SCAIL-2 just landed in ComfyUI.
huggingface.co/Comfy-Org/SCA…
Jun 9
HOT! SCAIL-2 just dropped!
End-to-end character animation via in-context conditioning
- no skeleton middleman, it copies pixels directly
- no glitches, no messy hands
- 512p/704p
- unified architecture for character replacement and multi-character tasks.
- zero-shot generalization to animal-driven and mesh-based control.
teal024.github.io/SCAIL-2/
2
41
276
21,428
Jun 9
HOT! SCAIL-2 just dropped!
End-to-end character animation via in-context conditioning
- no skeleton middleman, it copies pixels directly
- no glitches, no messy hands
- 512p/704p
- unified architecture for character replacement and multi-character tasks.
- zero-shot generalization to animal-driven and mesh-based control.
teal024.github.io/SCAIL-2/
10
68
461
44,238
Jun 9
AnchorWorld -a first-person world simulator, turns body movements into interactive video.
- Wan2.2 SMPL-X motion and head-mounted trajectories
- uses image, 3D pose, text
- high trajectory accuracy
- 480p, 77 frames
yuli0103.github.io/AnchorWor…
16
109
6,950
Jun 9
MAMMA!
You can do the Matrix Bullet Time effect with only 4 cameras.
It's a markerless motion capture system that turns standard multi-view video into accurate 3D digital models.
No expensive hardware and physical markers. Tracks complex human interactions like dancing/hugging.
mamma.is.tue.mpg.de/
1
18
91
4,775