14 Photos and videos
Death retweeted
Apr 8
Early R&D using our SP-6M dataset.
Exploring image-to-3D reconstruction from single images, including heavily modified inputs (lighting, hair, etc).
Still a work in progress.
27
66
603
52,428
Announcing ARC-AGI-3
The only unsaturated agentic intelligence benchmark in the world
Humans score 100%, AI <1%
This human-AI gap demonstrates we do not yet have AGI
Most benchmarks test what models already know, ARC-AGI-3 tests how they learn
246
579
4,326
741,058
Death retweeted
Mar 24
Seedance 2.0 is impressive. But it's closed-source!
Introducing our daVinci-MagiHuman — a single-stream 15B Transformer trained from scratch that jointly generates video audio. No cross-attention. No multi-stream branches. Just self-attention.
⚡ 5s 1080p video in 38s on a single H100
🏆 80% win rate vs Ovi 1.1 | 60.9% vs LTX 2.3 (2,000 human comparisons)
🌍 6 languages
📦 Fully open-source
Speed by simplicity.
By @SII_GAIR × @SandAI_HQ
📄 arxiv.org/abs/2603.21986
💻 github.com/GAIR-NLP/daVinci-…
🤗 huggingface.co/spaces/SII-GA…
88
257
1,869
298,037
Death retweeted
Jan 31
I've just ran @OpenClaw (formerly Clawdbot) through ZeroLeaks.
It scored 2/100. 84% extraction rate. 91% of injection attacks succeeded. System prompt got leaked on turn 1.
This means if you're using Clawdbot, anyone interacting with your agent can access and manipulate your full system prompt, internal tool configurations, memory files... everything you put in SOUL.md, AGENTS.md, your skills, all of it is accessible and at risk of prompt injection.
For agents handling sensitive workflows or private data, this is a real problem.
cc @steipete
Full analysis: zeroleaks.ai/reports/opencla…
357
793
5,092
975,952
Jan 31
Possibly the best AI video I have ever seen btw
Jan 29
🎬ViduQ3プロンプトテスト
兎の刃が、闇を切り裂く―― 因幡 vs 鬼軍
1500文字MAX限界突破。
演出・効果音完璧…ViduQ3ヤバすぎ✨
------------------------------
#YesVidu #ViduQ3 #ViduQ2R2VPro
45
Death retweeted
19 Dec 2025
This new voice cloning model should be illegal 😱
Okey, maybe not, but its soooo good. Usually the problem with voice cloning models is that they are super flat, and accents gets americanised
But this new model, Mirage audio really stays true to the original voice
more try
26
65
749
66,243
Death retweeted
14 Dec 2025
Thanks for the free Claude, sucker!
I'm gonna be rich!
14 Dec 2025
you can now try React Grab's agent to directly edit the website
no signup / use for free
11
15
815
233,344
Death retweeted
8 Dec 2025
What it's like to watch AI fix a bug
431
2,518
20,741
883,749
Death retweeted
6 Nov 2025
bro leaked his ip
297
665
25,776
881,707
7 Oct 2025
Easy way to get access to SORA sooner, and also access to countries outside of the US
7 Oct 2025
UNLIMITED Sora 2 & Sora 2 PRO are LIVE on Higgsfield!
Unlimited, unrestricted, & available worldwide.
Audio-synchronization, 1080p quality & multi-scene reasoning.
World's most wanted model UNLIMITED exclusively on Higgsfield.
For 9h: retweet & reply - get 150 creds in DM.
1
64
7 Oct 2025
LTX has always been my favourite video generator because of its efficient models that can run on consumer GPUs
Something BIG is coming for our most passionate creators.
We’re about to launch the LTX Ambassadors Program, a new way to get early access, shape what we’re building, and get paid.
Retweet and we'll DM you an invite to join the waitlist.
Details creator showcase below 🧵
41
Death retweeted
26 Sep 2025
The biggest application of MCP servers:
building them
219
571
8,524
522,606
17 Sep 2025
I'm fed up with AI companies mimicking each other, building near-identical projects like coding sandboxes or "deep researchers."
Here's a challenge: develop a project for AI to conduct empirical research, verify, and peer-review its responses.
Innovate, don't duplicate.
41
17 Sep 2025
Open source it😡
16 Sep 2025
We're thrilled to launch our new Hunyuan3D 3.0! It features 3x higher precision, 1536³ geometric resolution, and 3.6B voxel ultra-HD modeling for stunning detail.🔥🔥🔥
🌟Highlights:
✅Creates faces with lifelike facial contours and natural poses, creating truly realistic, high-fidelity models.
✅Precisely inteprets input images to reconstruct complex structures using a layered generation strategy, capturing hidden details
✅Generates ultra-clear, professional-grade details with enhanced texture fidelity and alignment for immersive, true-to-design visuals
Free access via Hunyuan 3D AI Engine (20 generations). Available via Tencent Cloud API. Transform gaming, film, e-commerce, and more!
👉🏻Try it now: 3d.hunyuan.tencent.com/
48
Death retweeted
15 Sep 2025
Introducing v0-clone 🪄
An open source starter kit for building your own AI app builder
Get started in seconds:
pnpm create v0-sdk-app -e v0-clone
→ AI Elements integration
→ v0 SDK integration
→ Speech-to-text
→ Drag-and-drop attachments
→ Real-time preview
→ Conversation history
→ Streaming support
→ Drizzle integration
→ Comprehensive task support
→ Auth.js integration
→ Multi-tenant architecture
33
58
793
222,671
Death retweeted
31 May 2025
Here is a prompt you can use as a template for creating Cursor rules for any framework.
18
124
1,753
193,735