David🌊

Community is Everything
Web3
Reply Guy

[David🌊 @davidbasser69]

Big progress this week. The cleaner long vs short horizon split should make the training data much more useful.

[Axis Robotics @axisrobotics]

Axis Weekly

Last week, we closed the remaining replay and runtime gaps between the browser, policy server, and physics stack — then scaled the coverage of our task generation engine (TaskGen).

An articulated asset library is now in the generation path, the full RoboCasa scene grid is online, and a cleaner long- versus short-horizon task split is ready for training and distillation.

Key updates:

• Replay & runtime: Policy and human-trajectory replay now match the frontend path, with web runtime reaching full replay fidelity. Scene loads faster, and simulation pauses automatically when idle.

• Articulated assets & RoboCasa: TaskGen now includes a library of 27 articulated object families (4 variants each). All RoboCasa scenes are online as a 50×50 layout and style grid.

• Horizon splits & next DAgger step: A cleaner long- vs. short-horizon split is ready, LIBERO Pro shows consistently high success rates, and the next round focuses on filtering correction segments that change closed-loop behavior.

Details below 👇

[David🌊 @davidbasser69]

@nGRND_io The shift from extraction to creating value is what makes this interesting.

[David🌊 @davidbasser69]

@nGRND_io @CCNCitizens @Futurist_conf @KurtRobsonn nGRND on Moon

[David🌊 @davidbasser69]

@nGRND_io nGRND one more step ahead

[David🌊 @davidbasser69]

The interesting part of the @AxisRobotics × @KaitoAI campaign isn’t just the 0.2% $AXIS reward pool.

It’s the mechanism behind it.

Kaito measures the attention you create around Axis through Mindshare.
Your invited users then create a second layer of value by generating verified, on-chain trajectories.

So the loop becomes:
Attention → contributors → data → Physical AI

That’s a much more interesting creator model than simply rewarding impressions.

[nGRND_io @nGRND_io]

We've spent centuries improving how we extract natural resources.

nGRND starts one step earlier.

The new question is how do you create value without extraction?

[nGRND_io @nGRND_io]

CCN - Crypto Citizens Network reporter Kurt Robson interview with nGRND Chair at the Toronto Futurist Conference

https://t.co/lgsj9lMLSr

@CCNCitizens @Futurist_conf @KurtRobsonn
#preservedgold @nGRND_io

[nGRND_io @nGRND_io]

For thousands of years, we've assumed gold has one purpose:

Extract it.
Sell it.
Move on.

nGRND starts from a different question: what if verified gold could create value without extraction?

That's not a new asset, it's a new economic model.

[David🌊 @davidbasser69]

@nGRND_io Gold🚀

[nGRND_io @nGRND_io]

nGRND is built around resources measured in geological time.

Building around geology changes the questions you ask.

Not only:
"What's happening this quarter?"

But also:
"What still matters decades from now?”

[David🌊 @davidbasser69]

Kaito is measuring attention.

Axis is measuring what that attention can actually turn into.

That’s what makes the @AxisRobotics × @KaitoAI campaign interesting to me.

It’s not just a mindshare leaderboard. The referral side connects content to real users, real trajectories, and verified onchain activity.

Attention → contributors → data → better Physical AI.

That loop makes a lot more sense than farming impressions for the sake of impressions.

[David🌊 @davidbasser69]

@mdabdurrahim98 @axisrobotics Exactly. Better data could be the real advantage here.

[David🌊 @davidbasser69]

What if the biggest upgrade for a robot isn’t a better brain…

but a better memory?

A robot can’t learn from situations it has never seen.

That’s why the idea behind @AxisRobotics stands out to me. More environments, more tasks, more variations, more edge cases → a richer pool of experience for Physical AI to learn from.

Maybe the race isn’t just about who builds the smartest robot.

Maybe it’s about who can give that robot the most useful experiences.

[Axis Robotics @axisrobotics]

What Happens After Post-Training Data Collection?

Since launching Axis V2, human-gated DAgger correction has been a core part of our data engine. In this thread, we share a series of experiments exploring how to best process and use post-training data.

The core finding: not every human intervention helps. But if you verify which specific actions actually change the outcome — and train only on those — the improvements are real and they scale.

Results:

- 660 corrections collected → 161 (24.4%) entered training
- Each training snippet is ~0.8 seconds
- Naively imitating full human trajectories dropped success from 40.0% to 36.7%, while 161 verified snippets lifted the three-seed average to 48.3% on the same evaluation set — and reached 48.8–52.5% across three seeds in a separate paired evaluation.

Read the full blog: https://t.co/ASYarP7CgU

Details below ⬇️

[David🌊 @davidbasser69]

Everyone talks about building better robotics models.

But a model can only learn from what the data teaches it.

That’s why I think the real physical AI race is becoming a data race.

@AxisRobotics is interesting to me because they’re focused on scaling the part that could matter most: collecting, improving, and diversifying the data robots learn from.

Better models are important.

But without better data, how far can they really go?

[David🌊 @davidbasser69]

This one was definitely a different level on @AxisRobotics 👀

Workbench Tool Arrangement – Long Horizon
7 steps, multiple objects, precise positioning, a 90° rotation, and finally stacking the tape measure on the tray.

The task has a 4/5 difficulty rating with only a 58.4% pass rate, so it definitely takes more than just moving things around randomly.
Took me 3:13 to get it done successfully and upload the data. 🦾
Longer tasks like this are where the accuracy really matters.

Back to the grind.

[David🌊 @davidbasser69]

Just checked my @KaitoAI Aura and somehow I’m sitting at 20 with rank 5,577.Not crazy numbers but it feels good to finally have something showing up after the Yaps days.If you’re still waiting to claim yours, you can use my link:
https://t.co/eO2A6hjDdO to see how this whole Aura thing develops.

[David🌊 @davidbasser69]

The 7-day @AxisRobotics leaderboard is getting competitive 👀

Currently at #375 with 294 trajectories and a 66.7 avg score.

The top spots are already pushing 1K+ trajectories, so there’s plenty of room to climb, back to the grind.

Want to start contributing too?
https://t.co/bOLXFiI5p5

[David🌊 @davidbasser69]

@TheBullWeb3 @KaitoAI @axisrobotics Physical AI + InfoFi is a pretty interesting combo. 30 days to build mindshare, and 100% unlocked at TGE makes it even more interesting.

[Kaito AI 🌊 @KaitoAI]

Today we are welcoming @axisrobotics as the first token-based campaign on Kaito Katalyst.

This is also the first campaign to run a referral multiplier.