Tech • IA • Crypto
ChatGPT Work: finally, a useful app!
7/10TL;DR
OpenAI’s new ChatGPT Work app introduces a local workspace with plugins, skills, and automation features, but early use reveals structural inconsistencies and unfinished functionality.
KEY POINTS
New App Replaces “Classic” Version
A new ChatGPT desktop application is rolling out, distinct from the older version now labeled “ChatGPT Classic.” Users must ensure they install the updated app, which introduces a redesigned interface and new “Work” capabilities aimed at productivity and structured workflows.
Workspace Concept Mirrors Competitors
The app introduces ChatGPT Work, a workspace-oriented environment comparable to systems developed by competitors like Anthropic’s Claude Cowork. It allows users to operate within a local directory, theoretically enabling persistent knowledge, structured files, and reusable workflows across sessions.
Confusion Between “Projects” and Workspaces
A key usability issue lies in the distinction between projects and the actual workspace. Projects function as simple conversation folders, while the workspace is meant to act as a full working environment. This overlap creates confusion and risks limiting the tool’s intended capabilities if misused.
Local File System Integration
ChatGPT Work can connect to a local folder on the user’s computer, enabling file-based organization. It generates directories and documents, often in Markdown format, intended to serve as a growing knowledge base. However, inconsistencies were observed in where files are actually created versus where the system claims they are stored.
Markdown Knowledge Architecture
The system relies on Markdown files to structure information and memory. This approach allows compatibility with external tools like Obsidian and supports modular knowledge building. However, missing metadata headers in generated files raise concerns about indexing efficiency and long-term usability.
Skills and Automation Capabilities
Users can create skills, reusable instructions that automate recurring tasks such as accounting workflows or reporting. Unlike some competing systems, skills must currently be created through conversation with the AI, with no direct import or duplication feature available.
Plugins Combine Tools and Skills
The platform includes plugins that bundle together external integrations and skills. For example, a Google Calendar plugin enables both system access (via APIs) and predefined automation behaviors. This hybrid model aligns with emerging standards for AI tool ecosystems.
MCP Integrations Enable External Access
ChatGPT Work supports MCP-style connections to services like email, banking, or cloud storage. These integrations allow the AI to retrieve and process real-world data, forming the backbone of advanced automation scenarios.
Integrated Browser for Live Interaction
A built-in browser allows users to open websites directly inside the workspace and interact with them alongside the AI. While still unstable, this feature hints at real-time collaboration between user, AI, and web content within a single interface.
“Computer Use” नियंत्रण Feature
A notable addition is Computer Use, which enables the AI to control the user’s device—mouse, keyboard, and applications. This could allow full task automation beyond the browser, though current functionality appears limited and may not yet work reliably on Windows systems.
Performance and Stability Issues
Early testing shows multiple inconsistencies: mismatched directories, incomplete file structuring, and partially functional features. These issues suggest the product is still in an early deployment phase, with expected refinements ahead.
CONCLUSION
ChatGPT Work signals a major shift toward fully integrated AI work environments, but its current implementation remains uneven, requiring further development before it can reliably match or surpass competing solutions.