Open Fleet Learning Infrastructure for Physical AGI
San Francisco
- Tweets 23
- Following 12
- Followers 91
- Likes 40
Avala AI retweeted
🇺🇸🇺🇸 Happy 250th Birthday, America! 🇺🇸🇺🇸
250 years ago, a handful of founders made a radical bet: that ordinary citizens, given liberty and a stake, could govern themselves. That wager built the most dynamic civilization in history.
Now, at the dawn of AI, Robotics, and The Age of Abundance, the bet renews. When the machines do the work, will ordinary people own a piece of them, or only consume what they produce?
The answer is the same one America has always reached for: make more founders. More owners. More people placing the bet instead of living inside someone else’s.
Full piece, 250 years forward to the 500th: nitter.net/Emal/status/2073135817…
🧵
2
2
8
9,201
Vision-Language-Action models solved the robot control problem. Physical Intelligence proved it with π0.
But scaling is blocked by data infrastructure. Teams waste months stitching 5+ incompatible tools together.
The architecture race is over. The data race just began!
avala.ai/news/vla-robot-brai…
1
1
2
795
Avala AI retweeted
shipping this open source because the ecosystem needs it! rust, clean architecture, MIT licensed. PRs welcome :) github.com/avala-ai/rust-cod…
2
3
190
Avala AI retweeted
Great map @emily_yu @aurorafeng_01 @deeptt!
Would love to see @AvalaAI on here — we're building full-stack 4D world-data infrastructure for AV and robotics: annotation, sensor fusion, dataset management, ingestion, and indexing.
We'd show up in a few of these categories 😄 Happy to share what we're seeing on the ground!
avala.ai
1
3
293
Real-world AI makes all the difference
Ashok Elluswamy, VP of AI at @Tesla on self-driving:
"It's so obvious you can solve this with cameras. Why wouldn't you solve with cameras? It's 2026. The self-driving problem is not a sensor problem, it's an AI problem. The cameras have enough information already. It's a problem of extracting the information, which is an AI problem."
(via @aelluswamy's presentation at the 2026 ScaledML Conference on January 29th)
84
307
3,472
145,559
Rivian just announced custom silicon that rivals Tesla's autonomy hardware. But it's their LiDAR decision that really caught our attention 🤯
RAP1 delivers 1,600 TOPS on 5nm—4x their current NVIDIA system. Combined with an end-to-end Large Driving Model, they're betting on vertical integration.
Is camera-only perception a trillion-dollar bet or a physics problem waiting to happen 👀?
Our technical deep dive: about.avala.ai/news/rivian-c…
5
7
72,553
VLAs are converging as the standard for general-purpose robots. But here's what teams discover the hard way:
The data that trained GPT-4 won't train your robot.
Why robot data is fundamentally different—and what that means for building systems that actually work: about.avala.ai/news/vla-data…
1
134
Our Favorite Papers from NeurIPS 2025 about.avala.ai/news/neurips-…
1
3
168,832
Avala AI retweeted
We are excited to have @AvalaAI as a platinum sponsor of #Actuate2025!
As the leading developer conference for physical AI, Actuate brings together top companies and engineers to share technical challenges, present breakthroughs, and help accelerate the path to commercialization for the entire industry. We’re grateful for the support of Avala and all of our sponsors to bring this event together.
We are 4 weeks away and tickets are moving fast! Get your ticket before general admission pricing ends: hubs.li/Q03Fx5Ty0
2
5
547
We need to build more ethical AI, aligned with human values and objectives. Find out more at avala.ai
ALT We need to build more ethical AI, aligned with human values and objectives. Read on to find out why...
ALT AI systems need to work for everyone. Building aligned models means eliminating the algorithmic biases that typically come with tools that don’t work well across BIPOC communities, producing more performant outputs for all people.
ALT Open access is more than API access Accountability requires transparency. Transparency means anyone can observe and examine model behavior and the provenance of the datasets on which those models are trained. A truly open platform will provide these audit tools.
ALT Share transformative opportunities AI could usher humanity into a new age of abundance, ending social and geographical inequalities. But only if it puts people first, and is aligned with our values and objectives.
1
1
621
Building better aligned AI, with human feedback from a global team of domain experts and data labelers #avalaai #peoplefirstai #aialignment avala.ai
1
2
454