Five releases stand out in the July 9 Radar: Meta’s Muse media models, OpenAI’s GPT-Live voice models, Codex inside JetBrains IDEs, a broad set of GitHub Copilot updates for Visual Studio Code, and a mobile workflow for handing merge conflicts to Copilot. Together, they show AI products moving beyond chat boxes and into media creation, live conversation, coding environments, and everyday software maintenance.
The reporting date is July 9, 2026. The official announcements reviewed below were published on July 7 and July 8; this Radar does not imply that every item launched on July 9.
At a glance
- Muse Image and Muse Video: Meta launched Muse Image and previewed Muse Video, with different availability for each model.
- GPT-Live: OpenAI began a global rollout of two full-duplex voice models in ChatGPT Voice.
- Codex in JetBrains: GitHub added Codex as a public-preview agent provider and expanded agent customization controls.
- Copilot in VS Code: GitHub’s June release roundup covered browser tools, parallel sessions, cost visibility, model discovery, and Autopilot changes.
- GitHub Mobile: Maintainers can start a Copilot cloud-agent merge-conflict workflow from a pull request on mobile.
Meta launches Muse Image and previews Muse Video
Meta announced Muse Image and Muse Video on July 7. Muse Image launched across the Meta AI app and meta.ai, Instagram Stories in the United States, and WhatsApp in a limited set of countries. Meta said Facebook availability would follow. Muse Video was presented as an early preview and was described as coming later to creators and Meta AI.
The distinction matters: Muse Image was an available product release, while Muse Video was not yet generally available. Meta described Muse Image as supporting instruction-following, image editing, multi-reference composition, tool use, and self-refinement. Those are provider-reported capabilities, not results independently reproduced by Kingy AI.
Kingy AI assessment: Muse points toward media models that can plan, use tools, and revise outputs instead of treating generation as a single prompt-response step. Teams evaluating it should still review source rights, identity consistency, embedded text, factual details, provenance signals, and the availability rules for each Meta surface.
GPT-Live brings continuous voice interaction to ChatGPT
OpenAI’s July 8 GPT-Live announcement introduced GPT-Live-1 and GPT-Live-1 mini. OpenAI describes the models as full duplex: they can continuously process audio while producing a response, allowing the system to decide when to speak, listen, pause, interrupt, or invoke a tool.
The rollout began globally across ChatGPT on iOS, Android, and the web. OpenAI said GPT-Live-1 would become the default for Go, Plus, and Pro users, while GPT-Live-1 mini would become the default for Free users. The July 8 announcement also said API access was planned; that dated statement should not be read as a current API-availability guarantee.
Kingy AI assessment: The important shift is architectural rather than cosmetic. Continuous interaction and delegation allow a voice interface to keep a conversation moving while another model handles search or deeper reasoning. Real deployments still require review of language coverage, interruptions, background noise, latency, safety behavior, and the limits of any connected tools.
Codex becomes a public-preview agent provider in JetBrains
GitHub’s July 7 JetBrains update added Codex as an agent provider in public preview. Developers install the Codex CLI, configure its path in GitHub Copilot settings, and then select Codex from the agent picker without leaving the IDE.
The same release expanded hooks and MCP-server management, added approval settings for Copilot CLI sessions, and introduced support for custom models configured by Copilot Business and Enterprise administrators. GitHub notes that Business and Enterprise administrators must enable the editor-preview-features policy before members can use the Codex provider.
Kingy AI assessment: The release gives JetBrains teams more choice over agents, models, tools, and approval behavior inside an established development environment. Those controls help teams shape a workflow; they do not by themselves guarantee secure execution, correct code, policy compliance, or successful unattended operation.
GitHub’s VS Code roundup expands the agent workspace
GitHub published its June 2026 Copilot roundup for Visual Studio Code on July 8. It covers versions 1.123 through 1.127, shipped through June and early July, rather than one single feature released on the publication date.
The roundup includes generally available agentic browser tools, parallel agent sessions, multiple chats within a session, clearer session and subagent cost visibility, model-provider discovery through the Marketplace, and changes to Autopilot behavior. It also describes managed settings, workspace-trust improvements, and secret storage for MCP OAuth credentials.
Kingy AI assessment: VS Code is becoming an orchestration surface for longer-running agent work. That makes visibility and control more important: teams need to know which model and tools are active, what permissions were granted, what a session consumed, and how generated changes will be reviewed before merging.
GitHub Mobile can hand merge conflicts to Copilot
GitHub’s July 8 mobile release lets a maintainer start a Copilot cloud-agent workflow from the merge box of a pull request with conflicts. Tapping “Fix with Copilot” prepares a comment asking Copilot to resolve the conflicts; submitting it launches the agent.
GitHub said the workflow was available in the latest production build of GitHub Mobile for iOS and Android. The launch announcement does not establish a specific price or AI-credit charge for this action, so this Radar makes no pricing claim.
Kingy AI assessment: Mobile handoff can keep a pull request moving when a maintainer is away from a workstation, but convenience does not replace review. Conflict resolution can change program behavior, tests, dependencies, or generated files, and the resulting commit should receive the same scrutiny as any other automated code change.
What connects these releases
Across media, voice, and software development, the common pattern is delegation. Muse can use tools and refine an image; GPT-Live can delegate deeper work while maintaining a conversation; coding agents can operate inside IDEs, browsers, and pull-request workflows. The interface is becoming less important than the chain of decisions and tools behind it.
For buyers and builders, that shifts evaluation toward operational questions: What can the agent access? Which actions require approval? How is usage made visible? Can a person inspect and reverse the result? What availability limits apply today? A polished demonstration is useful, but those controls determine whether an agent fits a real workflow.
Method and availability note
This Radar uses the five official provider announcements linked above. Product facts and availability statements are attributed to those dated public sources. The Kingy AI assessments are editorial interpretation, not hands-on testing, performance measurement, sponsorship, endorsement, or a buying recommendation. No vendor benchmark, customer result, funding claim, private evidence, or pricing assertion is used.
Availability, policies, model behavior, and plan terms can change after publication. Readers should confirm current product documentation before deployment. Continue with the AI Launch Tracker or browse the latest AI news.
Sponsor path
Sponsor Kingy AI launch coverage
Launching an AI product that needs clear demos, creator education, and buyer trust? Sponsor a Kingy AI video or launch feature.
The Kingy Brief
Follow The Kingy Brief.
One consequential launch, one pricing, limit, or shutdown change, one hands-on test, one exact prompt or Test Pack, and one try / watch / skip verdict.
Free · Choose your subjects · Double opt-in · Unsubscribe anytime