Poseidon
@psdnai
Followers
27K
Following
724
Media
70
Statuses
278
Real-world data for physical AI
Joined July 2025
Internet data fueled LLMs, but the next leap in AI will happen in the real world. ▸ Robots that clean ▸ Vehicles that see ▸ Agents that speak Poseidon is a full-stack data layer for the supply and demand of AI training data. Introducing Subnets in the Poseidon Litepaper ↓
292
225
1K
It all starts with IP. Thank you @Dune 🤝
We asked blockchains live on Dune to recount their 2025: the metrics that mattered, the ecosystem projects that truly moved the needle, and first-hand perspectives from core contributors. Here’s what the data revealed. https://t.co/mXrsAUVVxK
49
27
132
In just 2 weeks, eco project @psdnai built an audio dataset much larger than the best AI datasets for "low resource" languages, with significantly more non-European language hours than public datasets: ▸ Over 34k hours of audio data ▸ From 405k contributors worldwide
111
62
371
The Official Story Holiday Gift Guide 🎁 1. Rights-cleared training data 2. Rights-cleared training data 3. Rights-cleared training data
90
28
244
Despite the incoming drought of public text, we're nowhere close to using up the world's data. The world's total data is ~200 ZBs and doubles every ~2 years. The valuable modalities and sources that can be drawn from include: > Audio (@psdnai, @silencioNetwork, @Meet_Perle) >
20
17
133
Poseidon's Subnets are tailored to the incompatibilities between different AI domains. These decentralized data pipelines will enable Poseidon to serve diverse sectors, instead of just a narrow domain. Story makes them rights-cleared, licensable, and ready for commercial use.
Most people see data collection as a centralized workflow. With Poseidon, it happens via Subnetworks. ˗ˏˋ Example ´ˎ˗ → Audio Transcription Subnetwork Contributors upload real voice samples. Validators check for duplication, AI generation, and accuracy. Once verified, the
82
48
337
ICE was born on a mission to Defend the Homeland. Will you answer the call to serve?
314
890
6K
Co-Founder @SPChinchali shows how decentralized data collection rewrites the growth model. Traditional pipelines grow linearly and slow as they scale. Decentralized collection grows in parallel across thousands of contributors. AI training data for teams that need it, fast:
Plot 1: Poseidon (blue) vs 4 major public datasets (orange). We collected orders of magnitude more data for under-represented languages like Korean, Indonesian, Vietnamese, while still competitive on European languages. Note: log scale, so differences are even more dramatic.
64
33
332
Confidential Data Rails is set to power the Story ecosystem, including @psdnai’s AI data, @aria_protocol’s music masters, and @Genobank_io’s genomic vaults. It's a glimpse of what’s next: encrypted data that can automate, permission, and scale. Read ↓ https://t.co/5xEm9NnvPb
story.foundation
Story is AI-native infrastructure for the intellectual property economy.
11
10
151
How is SG, a services and finance hub in Asia, supporting the region's corporates and governments in their transition to a low-carbon future?
14
45
253
Most people see data collection as a centralized workflow. With Poseidon, it happens via Subnetworks. ˗ˏˋ Example ´ˎ˗ → Audio Transcription Subnetwork Contributors upload real voice samples. Validators check for duplication, AI generation, and accuracy. Once verified, the
89
58
444
The future of physical AI runs on high-quality, long-tail training data.
This might be the most disturbing AI paper of 2025 ☠️ Scientists just proved that large language models can literally rot their own brains the same way humans get brain rot from scrolling junk content online. They fed models months of viral Twitter data short, high-engagement
55
23
284
The acceleration of AI won’t just come from stacking more GPUs or breaking ground on another data center. It requires coordinated investment across the entire AI stack. From the top down, that stack includes ↴ ↳ Model development: architecture design, optimization, and
106
55
407
In 2 weeks, Poseidon was able to secure: ▸ Over 34k hours of audio data ▸ From 405k contributors worldwide Building an audio dataset that supports the foundation laid by public datasets. Together, we’re powering the future of physical AI.
126
63
498
Hiding your fungus damaged toenails? Get clear, healthy-looking nails with clinically proven NONYX!
0
4
33
I’m thrilled to share that I’ve joined Poseidon @psdnai as VP of Engineering. After an incredible journey at @StoryProtocol, helping build the foundation of an AI-native IP blockchain where data itself becomes IP, I’m excited to continue pushing that vision forward from a new
39
8
122
When @SPChinchali talks about AI data, you can tell he’s lived it. From Stanford labs to UT Austin classrooms and all the real-world data in between. Now he’s helping lead the charge at Poseidon to solve AI’s biggest bottleneck: data. Hear the pitch ↴
64
48
346
If data like this was in the training set, I think your autonomous vehicle would be less likely to get stuck in an infinite loop at the roundabout. https://t.co/7BFpjzLhEr
68
32
366
STAY with UNICEF💙 We are proud to partner with 'Crypto', raising awareness and funds to support children in need around the world. Learn more about the partnership here
0
4
67