えあ

フィジカルえーあぃ　・ω・

JDLAジェネラリスト2019#2

[吉川真人🇯🇵深セン @mako_63]

Xiaomiのヒューマノイドロボットが自動車工場で「空コンテナの折りたたみ・回収」作業を開始し、現在の作業成功率は90%。
注目すべきは箱を畳めることではなく、小さな留め具を指先で解除し、両腕で折りたたみ、他のロボットと協調して物流を回す一連の工程をこなしている点です。
ロボットを自社の工場の実務でデータを集めていける会社は強いですね。ユースケースや売り先がなくて死ぬことはないから。Xiaomi以外ならTesla、Xpeng、ヒュンダイ（ボストンダイナミクス）、いずれトヨタ

[mimic @mimicrobotics]

Introducing FLUX-mimic, a next-generation Video-Action Model for general purpose dexterity, developed in partnership with @bfl_ai.

Late last year we published mimic-video and introduced Video-Action Models (VAM): a new family of robotics foundation models built on top of video generation models. We showed that robot control reduces to visual prediction, and that robot capability is downstream of improvements in video modeling accuracy. The obvious implication was that advances in the video modeling frontier would directly translate to increased capabilities in end-to-end robot learning.

FLUX-mimic is that thesis at frontier scale: We've applied our VAM architecture to the strongest video backbone available today, FLUX 3 from Black Forest Labs, and trained it on data from our own robots and wearables. General-purpose dexterity, running on a single GPU on premises.

Because the model already understands world dynamics, it needs far fewer demonstrations to learn a new task. This is game-changing for our mission to deploy robots to factory floors, where industrial robot data is scarce and expensive to collect.

We're now testing and deploying FLUX-mimic with manufacturing leaders like @Audi, on complex, multi-step manipulation long considered impossible for conventional automation.

[Stefan Weirich @stefan_weirich]

World models will eat robotics.
@bfl_ai x @mimicrobotics

[Yinghao Xu @YinghaoXu1]

🚀 LingBot-VA 2.0 is here!

After half a year of teamwork, we're thrilled to release LingBot-VA 2.0 — a native video-action foundation model for generalizable robot control.
Unlike prior world-action models that retrofit generic video generators for robot control, LingBot-VA 2.0 is natively pretrained from scratch as a video-action foundation model.

Three key insights:
🌍 Native Video-Action Pretraining for learning world knowledge that enables strong generalization.
🧩 Semantic Visual-Action Tokenizer for more accurate action prediction with robust prompt following
⚡ Foresight Reasoning enables the robot to think ahead while acting, delivering continuous, responsive control without interrupting execution.

It runs in real time on consumer-grade GPUs, supports up to 150 Hz control, and generalizes to unseen tasks.

👇 Demos below

[How To Prompt @HowToPrompt__]

NVIDIA just open-sourced a model that takes broken, blurry 3D scans and rebuilds them clean from any angle..

It's called ArtiFixer. It uses video diffusion to generate the camera angles you never captured, then reconstructs the scene from the generated frames.

→ 70x faster than anything
→ Finishes in 1 to 4 steps
→ Beats SOTA by 3dB
→ Works from just text prompts

100% Open Source.

[Charly Wargnier @DataChaz]

🚨 Forget LIDAR.

The Robbyant team just dropped a streaming 3D model that reconstructs scenes live, at ~20 FPS, over long sequences.

One single camera. Runs in real time. Open-source.

Entirely end-to-end.

NO iterative optimization tricks and no post-processing cleanup steps!

It outperforms both existing streaming approaches and several offline methods.

100% Free and open-source.

Repo, paper and model weights in 🧵↓

[Movez @0xMovez]

A senior Google engineer just dropped a 19-page PDF on "Loop Engineering" for LLM and agentic systems.

Act → Observe → Learn → Repeat

• Act: the LLM proposes a code transformation (tile this loop, parallelize that one).

• Observe: a compiler runs it and reports back - is it valid? faster? slower? by how much?

• Learn: the LLM reads that feedback and adjusts its next move.

• Repeat until it stops finding improvements.

The agent gets smarter purely from grounded feedback inside its own context window.

This 19-page PDF totally changed the way I’m building agentic systems today.

Read it now, then explore the article below.

[Anatoli Kopadze @AnatoliKopadze]

https://t.co/eq0oL4fIaR

[Sadao Tokuyama @tokufxug]

NVIDIAの動作生成AI「MotionBricks」。

複雑な制御や手作業を廃し、ブロックのように動作を組み合わせてリアルタイム生成。

35万以上のスキルを誇り、ゲームとロボティクスの双方で、即座に自律歩行や自在なスタイル変更を実現します。（詳細はリプ欄へ）

[Yoonseok Pyo @passionvirus]

Teaching our open humanoid @ROBOTIS AI Sapiens to dance! 🤖💃

We extracted 3D motion from raw 2D video (no MoCap device!) and trained it in NVIDIA Isaac Sim. Thanks to high-precision DYNAMIXEL-Q, we achieved an ultra-low Sim2Real gap for fluid movement!

- Video: https://t.co/XAGbiB9IZZ

[Sadao Tokuyama @tokufxug]

Meta AIらが開発した3Dメッシュ生成技術「MeshFlow」。

従来の1面ずつ順番に作成する遅い方式を覆し、全データを同時に並列処理。

画質の劣化（量子化誤差）を完全に無くしたことで、わずか1秒でプロ仕様の滑らかな3Dデータを生成します。（詳細はリプ欄）

[Jake Fitzgerald @earthtojake]

asked claude fable 5 to design a humanoid robot

2 hours and 1.4m tokens later, i got this

absolutely insane

[よこしゃん｜ヒューマノイド最前線🤖 @yokoshaan]

ヒューマノイドに歩行を覚えさせる学習を、ノートPCクラスのマシンで3分で終わらせる基盤が、無償公開されました。中国の清華大学発です。

清華大学智能産業研究院（AIR）が、ヒューマノイド／四足ロボット向けの強化学習基盤『UniLab』をオープンソースで公開しました。CPU 側で物理シミュレーションを高効率に回し GPU で学習する独自設計で、従来比3〜10倍速。ヒューマノイドの歩行学習が3分、四足は12秒で完了する事例を示しました。NVIDIA の CUDA に加え、Apple Silicon・AMD・Intel もそのままサポートし、Mac でも動作します。

これまでヒューマノイドの学習研究は高価な GPU クラスタを持つラボに限られていましたが、Mac で動く敷居の低さは大学や中小ラボの参入を一気に増やします。日本国内のロボティクス研究者にとっても、無視できないツール選択肢として浮上しています。

[NVIDIA AI @NVIDIAAI]

Introducing Cosmos 3: Our latest frontier model for Physical AI

Cosmos 3 is the world’s first fully open omnimodel with native vision reasoning, world and action generation.

Today we’re releasing Super (32B) and Nano (8B) variants.

[mizumaru@作曲依頼等募集中 @mizumaruP]

もうAIは買い切りの時代へ

UnityやBlender開発に使い放題なAIエージェント登場

コーディングだけじゃなく、調べ物や日常にも使える
Synaptic Code

セットアップは画面に従うだけ。複雑な設定なし。¥8500

[CyberRobo @CyberRobooo]

No wonder they say teleoperation is dead

Peking University’s DAGroup released HumanNet, a massive 1 million hour dataset of human-centric videos that turns everyday internet footage into gold for embodied AI.

It’s got first-person and third-person views, super detailed actions, object interactions, tool use, and long sequences of real behavior. Everything’s cleaned up and annotated with 3D poses, SLAM tracks, and robot-ready labels.

The crazy part is,Training on just 1,000 hours of their first-person videos performs as well as (or even slightly better than) 100 hours of actual robot teleop data on downstream tasks.

Basically, human videos can now stand in for expensive robot data. Scaling laws for robotics just got a whole lot more realistic.

Human priors + smart curation = the future of world models.

[Santi Torres @SantiTorAI]

🚨 ULTIMA HORA: Alguien acaba de conectar Claude directamente a Blender.

Ahora puedes construir modelos 3D complejos y geometrías avanzadas escribiendo solo un prompt de texto.

Sin código. Sin experiencia en 3D.

[そう｜Claude CodeのX運用で月800万インプ @so_ainsight]

これ、ガチで時代変わる。
Microsoftが「写真1枚から3Dモデルを作るAI」を無料公開しました。名前は TRELLIS.2。

写真を1枚渡すだけで、
・金属やガラスの質感までそっくり再現
・コップの中の空洞、葉っぱの薄さ、穴あきの服もそのまま再現
・速いモードなら3秒、高精細でも1分

しかも個人でも会社でも、自由に使ってOK。
ゲーム素材も商品モックも、写真さえあれば作れる時代です。

深掘り続きます👇

[Sadao Tokuyama @tokufxug]

自然言語から実ロボットを設計するオープンソース基盤が登場。

プロンプトからAIエージェントがCADコードを直接執筆し、7軸アーム等の3Dモデルを自動生成。

バイナリではなくソースコード制御のため、全工程がレビューと再現可能に。

設計から Kinematics 構築までを数分に短縮する新機軸。

[Ilir Aliu @IlirAliu_]

Your robot doesn’t need a policy anymore.
It can just write its own.
Coding Agents for Robotics:

Instead of training fixed models, robots become agents that:

• call perception and control APIs
• write code to solve tasks
• execute, observe, and improve in loops

This is a shift from “learning policies”
to programming behavior in real time.

And it works across robot arms, humanoids, and mobile systems.

What they built is deeper than it looks:

• full agentic toolkit (vision, depth, IK, grasping, navigation)
• CaP-Gym with 187 real manipulation tasks
• CaP-Bench benchmarking 12 frontier models
• CaP-Agent0 solving tasks without task-specific tuning
• CaP-RL boosting a 7B model from 20% → 72% in 50 iterations

Policies like VLAs become just another API call.

Which means:

You don’t train for every task anymore.

You let the agent figure it out on the fly.

From static policies to agentic systems that adapt in real environments.

Thanks for sharing, @DrJimFan !

Explore:
📍Project: https://t.co/6pHRqU2arI
Code: https://t.co/KQs840XU8A
Paper: https://t.co/2K0q6pnAOo

——

Weekly robotics and AI insights.
Subscribe free: https://t.co/9Nm01QUcw3