InfoWorld has taken Microsoft to task over Visual Studio Code’s increasingly AI-heavy release cadence, arguing that the editor is being repositioned as an agent front end at the expense of conventional IDE improvements.
In a July 20 opinion piece, InfoWorld senior writer Serdar Yegulalp points to VS Code 1.126 through 1.128 as evidence of the shift. Microsoft’s release notes support the broad observation: recent updates have concentrated on multi-chat agent sessions, model controls, agent-host infrastructure, browser tools, AI credit reporting, and Copilot workflow features.
VS Code 1.126, released June 24, introduced session-level AI cost information, multiple chats in an agent-host session, agentic code-review feedback, and the Agent Host Protocol. Its notable non-AI change was a safer default for opening unfamiliar folders: new folders now open in Restricted Mode rather than immediately prompting users to trust them.
Version 1.127, released July 1, pushed further into the agent workflow. Microsoft made integrated-browser tools generally available, allowing agents to open pages, inspect content and errors, take screenshots, and navigate web applications. The release also added agent session groups, pull-request feedback actions, generated PR descriptions, and agent credit visibility.
VS Code 1.128 followed on July 8 with more multi-chat agent work and generally available image support in Chat. It also added OS-level keyboard-shortcut registration, the conventional editor enhancement Yegulalp singled out as useful amid an otherwise AI-led update.
Microsoft’s documentation now describes VS Code as shipping with AI agents built in. Depending on permissions and configuration, an agent can edit project files, run commands, use browser tools, and iterate on errors. That is a meaningful expansion of the editor’s scope, particularly for organizations already using GitHub Copilot.
That distinction matters for managed desktops. The newer browser and agent features can reach code, terminals, and—where allowed—network resources, so admins should ensure their VS Code policies match existing Copilot and data-handling rules. Microsoft’s 1.127 release notes also introduced file-based managed Copilot settings on Windows through
Yegulalp’s complaint is ultimately about priorities, not the removal of core editor capabilities: users can still build language-specific environments through extensions, but Microsoft’s release narrative now puts AI agents first.
Developers can disable the AI features they do not use, while administrators should review the relevant VS Code and Copilot policies before deploying newer builds.
In a July 20 opinion piece, InfoWorld senior writer Serdar Yegulalp points to VS Code 1.126 through 1.128 as evidence of the shift. Microsoft’s release notes support the broad observation: recent updates have concentrated on multi-chat agent sessions, model controls, agent-host infrastructure, browser tools, AI credit reporting, and Copilot workflow features.
Agents dominate the changelog
VS Code 1.126, released June 24, introduced session-level AI cost information, multiple chats in an agent-host session, agentic code-review feedback, and the Agent Host Protocol. Its notable non-AI change was a safer default for opening unfamiliar folders: new folders now open in Restricted Mode rather than immediately prompting users to trust them.Version 1.127, released July 1, pushed further into the agent workflow. Microsoft made integrated-browser tools generally available, allowing agents to open pages, inspect content and errors, take screenshots, and navigate web applications. The release also added agent session groups, pull-request feedback actions, generated PR descriptions, and agent credit visibility.
VS Code 1.128 followed on July 8 with more multi-chat agent work and generally available image support in Chat. It also added OS-level keyboard-shortcut registration, the conventional editor enhancement Yegulalp singled out as useful amid an otherwise AI-led update.
Microsoft’s documentation now describes VS Code as shipping with AI agents built in. Depending on permissions and configuration, an agent can edit project files, run commands, use browser tools, and iterate on errors. That is a meaningful expansion of the editor’s scope, particularly for organizations already using GitHub Copilot.
The practical Windows angle
For Windows developers who do not want the AI layer, VS Code still works as the extension-driven editor it has long been. Microsoft says built-in chat, inline suggestions, and Copilot extensions can be disabled through thechat.disableAIFeatures setting. The company also provides enterprise policies to disable agent mode and restrict agent network access, third-party tools, and MCP servers.That distinction matters for managed desktops. The newer browser and agent features can reach code, terminals, and—where allowed—network resources, so admins should ensure their VS Code policies match existing Copilot and data-handling rules. Microsoft’s 1.127 release notes also introduced file-based managed Copilot settings on Windows through
%ProgramFiles%\GitHubCopilot\managed-settings.json, offering an option for machines not enrolled in MDM.Yegulalp’s complaint is ultimately about priorities, not the removal of core editor capabilities: users can still build language-specific environments through extensions, but Microsoft’s release narrative now puts AI agents first.
Developers can disable the AI features they do not use, while administrators should review the relevant VS Code and Copilot policies before deploying newer builds.
References
- Primary source: InfoWorld
Published: 2026-07-15T18:21:16+00:00
Loading…
www.infoworld.com