Copilot leak lists integrations; Minimax M3+ and Nvidia N1X

Microsoft's Copilot app leaked on June 1, 2026, revealing a consolidated "super app" interface and internal plans to connect multiple third-party services. The dataset of files and screenshots shared across public repositories and social channels includes UI mockups, API stubs, and a list of integration partners referencing Slack, Gmail, Zoom, and several CRM tools.

The leak surfaced alongside two separate developments in the AI stack. Minimax published artifacts that point to a new M3+ variant, and Nvidia left traces of an N1X accelerator in driver and roadmap materials that have circulated among developers.

Copilot super app: what the leak shows

The leaked Copilot materials depict a single-window workspace that can host applets from different services. Mockups show a persistent left rail for navigation, a central canvas for multiapp task threads, and a gallery-style marketplace for extensions and connectors. Several integration examples include message threading with Slack, email drafting tied to Gmail labels, and calendar-aware scheduling linked to enterprise calendars.

Code snippets and API stubs in the leak reference authentication flows and telemetry endpoints. One set of files lists scopes and permission prompts for read and write access to third-party accounts. Another contains orchestration logic for passing context between applets, implying a messaging or intent layer that would let one plugin feed inputs to another inside the Copilot UI.

The materials also include notes on extensibility: SDK stubs labeled for JavaScript and a native wrapper for Windows and macOS. There are references to subscription tiers and notes about developer billing, but no firm launch timeline or pricing details appear in the leaked set.

Security and privacy comments are present in the drafts. They call for consent flows and per-connector data handling policies, but those sections are under revision in the files, indicating internal debate about how much data the Copilot host will surface to plugin developers.

Minimax M3+ and Nvidia N1X updates

Minimax's repositories and release branches show a new model artifact tagged M3+. Release notes attached to the commit mention improved instruction-following and latency optimizations compared with the M3 baseline. There are also example evaluation notebooks and small-scale benchmark results emphasizing multi-turn coherence, though the artifacts stop short of full public evaluation sets or third-party benchmark comparisons.

Separately, Nvidia references an N1X accelerator in recent driver commits and internal boards. The references include a device code name and a note about firmware that adjusts tensor core scheduling for low-latency inference. Public-facing marketing content has not appeared, but the developer traces suggest Nvidia is preparing hardware aimed at dense inference workloads and longer context models.

Taken together, the model, hardware, and platform traces show activity across layers of the AI stack: an application-level hub for end users, a model variant promising incremental improvements, and a hardware family tuned for inference.

Why it matters

If Microsoft moves Copilot toward a single-window app hosting third-party services, it changes where and how users run AI-assisted workflows and raises new privacy and data-movement questions for enterprises and regulators. The M3+ and N1X developments show vendors optimizing models and silicon in parallel, which will affect deployment choices, costs, and performance trade-offs for cloud and on-premise AI workloads.