Eigen RadarAI
Archive
Analysis

Cursor brings code hosting inside its own editor

Cursor's Origin layer brings repositories into the editor; Google previews Gemini Enterprise editions for legal and finance teams, while Mozilla.ai opens shared agent skills to Octonous users.

Artificial Intelligence··Evening
A modular metal apparatus on a bright workbench joins golden branches into one glowing blue arrow-shaped rail.

Google opens separate editions for legal and finance work

Google Cloud has put two specialised Gemini Enterprise editions into preview. The legal version targets law firms and corporate legal teams with prebuilt skills for contract redlining, playbook creation, regulatory scanning and data-subject access requests. It connects to systems including iManage, NetDocuments, Docusign, Everlaw, RelativityOne and Thomson Reuters HighQ, inheriting their permissions. The financial-services edition, previewed the same day, is aimed at capital-markets and corporate-banking work. Its managed Financial Research Agent offers more than 50 foundational skills, confidence scores, data snapshots for auditing and source citations, and connects to services including FactSet, Moody's, MSCI, PitchBook and SEC Edgar while preserving existing entitlements.[1], [2]

Origin moves the repository into the editor

Cursor released Origin, a git-based code-hosting layer built into its editor, as an early beta for Pro, Teams and Enterprise subscribers. It places repositories, pull requests, browser code navigation, stacked pull requests and agent-aware merge queues inside Cursor. For projects mirrored from GitHub, GitHub remains the primary repository and pushes still land there. InfoQ reports that the launch coincided with a multi-hour GitHub outage affecting Actions, APIs, git operations and Copilot. Early community reaction focused less on the feature list than on who holds the code and how the data is handled.[3]

Mozilla.ai makes skills shareable across a workspace

Mozilla.ai opened Agent Skills to all Octonous users. Teams can store reusable instruction sets written in plain Markdown in the open Agent Skills format and invoke them with slash commands inside a conversation. A skill can contain product information, a research method, an operating procedure or review criteria; users can write one manually, draft it with a built-in tool, or import it from a GitHub URL or an upload. Skills remain private by default but can be shared across a workspace. Mozilla.ai gave no pricing or licence terms in the announcement.[4]

References

  1. News sourceGoogle CloudGoogle opens a legal edition of Gemini Enterprise in preview↩
  2. News sourceGoogle CloudGemini Enterprise gets a banking edition built around a Google research agent↩
  3. News sourceInfoQCursor starts hosting code inside its own editor↩
  4. News sourceMozilla.aiMozilla.ai adds a shared skills library to its agent platform↩
AI safety and regulationThe frontier model race

More analysis from this desk

View all Brief analyses from the Artificial Intelligence desk →