<?xml version="1.0" encoding="UTF-8" ?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Embodied AI Weekly | 1N Robotics</title>
    <link>https://1nrobotics.com/blog.html</link>
    <description>Weekly briefings and technical analysis on embodied intelligence, data, robot learning, and deployment.</description>
    <language>en</language>
    <lastBuildDate>Mon, 16 Nov 2026 21:04:27 +0800</lastBuildDate>
    <atom:link href="https://1nrobotics.com/feed.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>UMI after the gripper: Specialist and the engineering of useful robot data</title>
      <link>https://1nrobotics.com/posts/umi-specialist-data-collection-device-comparison.html</link>
      <guid isPermaLink="true">https://1nrobotics.com/posts/umi-specialist-data-collection-device-comparison.html</guid>
      <pubDate>Mon, 16 Nov 2026 21:04:27 +0800</pubDate>
      <category>Technical analysis</category>
      <description>Thirteen UMI Arena devices, a closer look at Specialist, and the tradeoffs between portable capture, contact sensing, data scale, and camera robustness</description>
    </item>
    <item>
      <title>From DUSt3R to Pi3X and MapAnything: geometry, ViewCrafter, and usable 3D worlds</title>
      <link>https://1nrobotics.com/posts/dust3r-pi3x-mapanything-viewcrafter-3d-robotics.html</link>
      <guid isPermaLink="true">https://1nrobotics.com/posts/dust3r-pi3x-mapanything-viewcrafter-3d-robotics.html</guid>
      <pubDate>Sun, 15 Nov 2026 09:21:00 +0800</pubDate>
      <category>Technical analysis</category>
      <description>How learned reconstruction meets generative rendering—and why content creators and robots need different definitions of a good world</description>
    </item>
    <item>
      <title>From BEVFormer to UniAD: a one-year route from Pi3X geometry to persistent navigation</title>
      <link>https://1nrobotics.com/posts/bevformer-uniad-pi3x-compass-one-year-roadmap.html</link>
      <guid isPermaLink="true">https://1nrobotics.com/posts/bevformer-uniad-pi3x-compass-one-year-roadmap.html</guid>
      <pubDate>Sun, 15 Nov 2026 02:25:00 +0800</pubDate>
      <category>Technical analysis</category>
      <description>What to transfer into Compass 1-style systems, which experiments must come first, and when a robotics team should continue, simplify, or stop</description>
    </item>
    <item>
      <title>Compass 1: what a single camera, geometric labels, and usable memory solve in robot navigation</title>
      <link>https://1nrobotics.com/posts/compass1-camera-geometry-memory-navigation.html</link>
      <guid isPermaLink="true">https://1nrobotics.com/posts/compass1-camera-geometry-memory-navigation.html</guid>
      <pubDate>Sat, 14 Nov 2026 16:53:00 +0800</pubDate>
      <category>Technical analysis</category>
      <description>Inside Archon’s navigation model: DJI Action 4 capture, monocular reconstruction, structured memory, video-action generation, and the gap to autonomous deployment</description>
    </item>
    <item>
      <title>From BEVFormer and UniAD to WorldEngine and SimScale: learning from the states that demonstrations miss</title>
      <link>https://1nrobotics.com/posts/bevformer-uniad-worldengine-simscale-robotics.html</link>
      <guid isPermaLink="true">https://1nrobotics.com/posts/bevformer-uniad-worldengine-simscale-robotics.html</guid>
      <pubDate>Sat, 14 Nov 2026 16:53:00 +0800</pubDate>
      <category>Technical analysis</category>
      <description>How spatial memory and planning-oriented learning led to interactive data engines, and which lessons transfer from autonomous driving to robotics</description>
    </item>
    <item>
      <title>Anchor1, SONIC, and SuperMimic: precision, scale, and the cost of control</title>
      <link>https://1nrobotics.com/posts/anchor1-sonic-supermimic-control-analysis.html</link>
      <guid isPermaLink="true">https://1nrobotics.com/posts/anchor1-sonic-supermimic-control-analysis.html</guid>
      <pubDate>Sat, 14 Nov 2026 16:53:00 +0800</pubDate>
      <category>Technical analysis</category>
      <description>Why a humanoid that can track a backflip can still miss a stationary reach, and what three control systems change about that problem</description>
    </item>
    <item>
      <title>Ropedia's next step: from human video to reusable robot worlds</title>
      <link>https://1nrobotics.com/posts/ropedia-xperience2-data-model-industry.html</link>
      <guid isPermaLink="true">https://1nrobotics.com/posts/ropedia-xperience2-data-model-industry.html</guid>
      <pubDate>Sat, 14 Nov 2026 16:15:00 +0800</pubDate>
      <category>Technical analysis</category>
      <description>Xperience 2, reconstruction, world models, and control: what the data can support, where the interfaces remain incomplete, and how industrial teams should measure value</description>
    </item>
    <item>
      <title>Robot learning after pretraining | Embodied AI Weekly</title>
      <link>https://1nrobotics.com/posts/embodied-ai-weekly-2026-11-13.html</link>
      <guid isPermaLink="true">https://1nrobotics.com/posts/embodied-ai-weekly-2026-11-13.html</guid>
      <pubDate>Fri, 13 Nov 2026 18:05:00 +0800</pubDate>
      <category>Weekly briefings</category>
      <description>AMP1, RoboScript, MR0 revisions, and NVIDIA's CoRL stack reveal where corrections and rehearsal improve robots—and where deployment costs remain unproven</description>
    </item>
    <item>
      <title>Ropedia W0: the data factory behind cross-embodiment world prediction</title>
      <link>https://1nrobotics.com/posts/ropedia-w0-data-model-analysis.html</link>
      <guid isPermaLink="true">https://1nrobotics.com/posts/ropedia-w0-data-model-analysis.html</guid>
      <pubDate>Fri, 13 Nov 2026 07:45:00 +0800</pubDate>
      <category>Technical analysis</category>
      <description>How human video, robot logs, and corrected kinematics become Xp-Data—and what the prediction and policy results actually establish</description>
    </item>
    <item>
      <title>Why agentic robotics is accelerating: NVIDIA's loops and TinyAstral's RSI</title>
      <link>https://1nrobotics.com/posts/agentic-robotics-nvidia-rsi-tinyastral.html</link>
      <guid isPermaLink="true">https://1nrobotics.com/posts/agentic-robotics-nvidia-rsi-tinyastral.html</guid>
      <pubDate>Thu, 12 Nov 2026 22:05:00 +0800</pubDate>
      <category>Technical analysis</category>
      <description>NVIDIA's navigation, manipulation, policy-research, and agentic WAM loops, compared with TinyAstral's harness RSI and compact-model distillation.</description>
    </item>
    <item>
      <title>From HumanWAM to GR00T 2: world-action models, data, and deployment</title>
      <link>https://1nrobotics.com/posts/humanwam-groot2-wam-comparison.html</link>
      <guid isPermaLink="true">https://1nrobotics.com/posts/humanwam-groot2-wam-comparison.html</guid>
      <pubDate>Thu, 12 Nov 2026 22:00:00 +0800</pubDate>
      <category>Technical analysis</category>
      <description>Architectures, aligned human and UMI data, benchmark protocols, reinforcement learning, and deployment costs across HumanWAM and GR00T 2.</description>
    </item>
    <item>
      <title>CoRL 2026 dexterous hands: what Sharpa helps us see</title>
      <link>https://1nrobotics.com/posts/corl-2026-dexterous-hands-sharpa.html</link>
      <guid isPermaLink="true">https://1nrobotics.com/posts/corl-2026-dexterous-hands-sharpa.html</guid>
      <pubDate>Thu, 12 Nov 2026 12:00:00 +0800</pubDate>
      <category>Technical analysis</category>
      <description>A CoRL 2026 dexterity analysis through Sharpa Wave, comparing learning routes, tactile feedback, real-robot protocols, generalization, and industrial value.</description>
    </item>
    <item>
      <title>Apple HALO: how a humanoid learns to move through a home</title>
      <link>https://1nrobotics.com/posts/apple-halo-humanoid-control-analysis.html</link>
      <guid isPermaLink="true">https://1nrobotics.com/posts/apple-halo-humanoid-control-analysis.html</guid>
      <pubDate>Thu, 12 Nov 2026 11:00:00 +0800</pubDate>
      <category>Technical analysis</category>
      <description>HALO's modular synthetic data, rendering, and training pipeline, with an analysis of generative rendering as a substitute for observation rendering and physical simulation.</description>
    </item>
    <item>
      <title>MuJoCo 4: The Industrial Value of Better Simulation</title>
      <link>https://1nrobotics.com/posts/mujoco-4-industrial-value.html</link>
      <guid isPermaLink="true">https://1nrobotics.com/posts/mujoco-4-industrial-value.html</guid>
      <pubDate>Thu, 12 Nov 2026 10:00:00 +0800</pubDate>
      <category>Technical analysis</category>
      <description>The industrial value of better contact simulation, collaboration, calibrated training, and measurable robot development costs.</description>
    </item>
    <item>
      <title>NVIDIA GR00T 2: World-Action Models Meet Deployment Reality</title>
      <link>https://1nrobotics.com/posts/nvidia-gr00t-2-wam-analysis.html</link>
      <guid isPermaLink="true">https://1nrobotics.com/posts/nvidia-gr00t-2-wam-analysis.html</guid>
      <pubDate>Thu, 12 Nov 2026 09:00:00 +0800</pubDate>
      <category>Technical analysis</category>
      <description>A deployment-focused analysis of GR00T 2's world-action architecture, data, generalization, dexterity, agentic recovery, and onboard latency.</description>
    </item>
    <item>
      <title>Did GPT-6 Astra Change the Robotics Roadmap?</title>
      <link>https://1nrobotics.com/posts/gpt6-astra-robotics-new-stack.html</link>
      <guid isPermaLink="true">https://1nrobotics.com/posts/gpt6-astra-robotics-new-stack.html</guid>
      <pubDate>Sat, 07 Nov 2026 21:00:00 +0800</pubDate>
      <category>Technical analysis</category>
      <description>A post-release evaluation of how GPT-6 Astra changes the division of labor among VLA, WAM, agentic systems, embodied in-context learning, and tactile control.</description>
    </item>
    <item>
      <title>From Predicting the Next Frame to Orchestrating the Next Step: Agentic WAM, MoT, and Dexterous Control</title>
      <link>https://1nrobotics.com/posts/nvidia-agentic-wam-mot-dexterity-analysis.html</link>
      <guid isPermaLink="true">https://1nrobotics.com/posts/nvidia-agentic-wam-mot-dexterity-analysis.html</guid>
      <pubDate>Sat, 07 Nov 2026 15:00:00 +0800</pubDate>
      <category>Technical analysis</category>
      <description>An analysis of the three-expert MoT in NVIDIA-linked AW-0, compared with EgoScale, T-Rex, DexTacWAM, and Dexplore.</description>
    </item>
    <item>
      <title>From Human Video to Robot Action: DA Group's HumanScale and HumanWAM Roadmap</title>
      <link>https://1nrobotics.com/posts/humanscale-humanwam-analysis.html</link>
      <guid isPermaLink="true">https://1nrobotics.com/posts/humanscale-humanwam-analysis.html</guid>
      <pubDate>Sat, 07 Nov 2026 09:00:00 +0800</pubDate>
      <category>Technical analysis</category>
      <description>How Open-HumanScale-50K, HumanWAM-Vanilla, and HumanWAM-Goal transfer egocentric human experience into robot action.</description>
    </item>
    <item>
      <title>Tactile reaction, field reliability, and collaborative robot learning</title>
      <link>https://1nrobotics.com/posts/embodied-ai-weekly-2026-11-06.html</link>
      <guid isPermaLink="true">https://1nrobotics.com/posts/embodied-ai-weekly-2026-11-06.html</guid>
      <pubDate>Fri, 06 Nov 2026 09:00:00 +0800</pubDate>
      <category>Weekly briefings</category>
      <description>Four developments that moved embodied AI closer to useful physical systems this week.</description>
    </item>
  </channel>
</rss>
