Coding Tools Gain Autonomy While Humanoid Robots Fine-Tune Physical Skills

The briefing · 25 Oct 2026 · 66 tracked changes

Coding Tools Gain Autonomy While Humanoid Robots Fine-Tune Physical Skills

A quiet wave of updates shows AI coding assistants taking on larger system tasks, while humanoid robotics platforms focus on physical coordination and precision.

AI Programming Agents Expand Independence

Anthropic decreased its monthly entry price for Claude Code to $17 and introduced scheduled tasks. Lovable evolved into a full-stack builder with frontend and backend generation, while Cognition added Slack and Teams integrations alongside a local desktop interface for Devin. Google also shifted Gemini Code Assist to its 3.5 Flash model and added a command-line interface.

Why it matters Software creation is becoming more hands-off as AI coding agents take on multi-step, full-stack tasks on their own.

claude-code lovable devin gemini-code-assist

Built from 6 tracked changes

Hardware Physical Control Gets Smarter

Unitree's G1 robot received hybrid force-position controls and support for the UnifoLM model, alongside continuous over-the-air updates. Nvidia added a dedicated humanoid foundation model to its Isaac GR00T platform, while Sanctuary AI's Phoenix gained hydraulic hand controls. For autonomous driving, Waymo integrated pedestrian action recognition.

Why it matters Robots are acquiring the precise sensory and motor skills needed to navigate and manipulate the real world safely.

unitree-g1 nvidia-isaac-groot sanctuary-phoenix waymo-driver

Built from 4 tracked changes

Context Windows Expand Across Asian LLMs

Zhipu AI updated its AutoGLM and Z.ai GLM platforms to version 5.3, with both supporting a one-million-token context window. Moonshot AI introduced its K3.1 Pro model for the Kimi writing assistant, bringing a responses API and expanding its context window to over one million tokens. Alibaba updated Qwen Chat to version Qwen3-2507, maintaining a 256K context limit.

Why it matters Larger context windows allow AI models to analyze much longer documents and retain more conversational history at once.

autoglm zhipu-glm kimi qwen-chat

Built from 2 tracked changes

Highly Regulated Industries Adopt Targeted Tools

Rogo added automation workflows alongside compliance measures designed for the EU AI Act. Hebbia introduced a collaborative workspace, data extraction, and document synthesis to streamline due diligence. Meanwhile, Abridge deployed dedicated revenue cycle documentation features to help healthcare systems capture clinical information.

Why it matters Finance, law, and medicine require specialized AI features built specifically to navigate strict compliance and complex document formats.

rogo hebbia abridge

Built from 3 tracked changes

See the raw change log