Back to list
Industry NewsOpenAILibreOfficeOpen Source

OpenAI's ChatGPT Desktop App Discovered to Bundle Full LibreOffice Suite and Development Runtimes

A technical investigation into the OpenAI Codex desktop app, recently rebranded as ChatGPT, has revealed a massive 1.7GB local runtime environment stored within the user's cache directory. Discovered by developer Simon Willison, the 'codex-primary-runtime' folder contains a comprehensive suite of tools, including full installations of Python and Node.js, alongside native binaries for Git, Poppler, and the LibreOffice open-source office suite. These components are integrated into the application via a 'skills' system, which provides the AI with the necessary instructions to utilize these local binaries for document processing and code execution. This discovery highlights the significant local infrastructure OpenAI is deploying to support its desktop application's advanced capabilities.

Hacker News

Key Takeaways

  • Massive Local Footprint: The ChatGPT (formerly Codex) desktop app installs a 1.7GB runtime folder named codex-primary-runtime in the user's ~/.cache/ directory.
  • Bundled Open-Source Software: The runtime includes a full copy of the LibreOffice suite, which forked from OpenOffice.org in 2010, to handle document-related tasks.
  • Development Environments: Full installations of Python and Node.js are bundled within the app to support local code execution and processing.
  • Native Binary Integration: The app utilizes native binaries for Git and Poppler, managed through a specific "skills" architecture located in the application's plugin directory.
  • Rebranding Transition: The discovery confirms the transition of the OpenAI Codex desktop application into the current ChatGPT desktop ecosystem.

In-Depth Analysis

The Discovery of the Codex Runtime

While performing a routine system cleanup using the utility OmniDiskSweeper, developer Simon Willison identified an unusually large allocation of space within the ~/.cache/ folder. The investigation pinpointed a directory titled codex-runtimes, specifically the codex-primary-runtime subfolder, which occupies approximately 1.7GB of disk space. This folder is associated with the OpenAI Codex desktop application, a tool that has since been rebranded under the ChatGPT name.

The presence of such a substantial runtime environment in a cache folder suggests that the application relies heavily on local execution capabilities rather than purely cloud-based processing for certain tasks. The specific path identified—~/.cache/codex-runtimes/codex-primary-runtime/plugins/openai-primary-runtime/plugins/documents—indicates a modular structure where document-handling capabilities are treated as specialized plugins or "skills."

A Comprehensive Suite of Local Tools

The contents of the codex-primary-runtime folder reveal a sophisticated local environment designed to handle a wide array of file formats and programming tasks. The most notable inclusion is a full copy of LibreOffice, the open-source office suite. By bundling LibreOffice, the ChatGPT app gains the native ability to parse, convert, and manipulate complex document formats (such as .docx, .xlsx, and .pptx) without requiring the user to have the software installed independently or relying on cloud-based conversion APIs.

In addition to office software, the runtime includes:

  • Python and Node.js: Full installations of these runtimes allow the AI to execute scripts and manage package-based workflows locally.
  • Poppler: A PDF rendering library that enables the application to process and extract data from PDF documents.
  • Git: The inclusion of the Git version control system suggests the app has native capabilities for repository management and code tracking.

These tools are not merely stored; they are actively integrated into the AI's workflow. The directory structure includes a "skills" folder, which contains the logic and instructions that tell the Codex/ChatGPT engine how to locate and invoke these specific native binaries to perform complex user requests.

Industry Impact

The Shift Toward Local AI Infrastructure

The discovery that OpenAI is bundling 1.7GB of third-party open-source software within its desktop client signals a shift in how AI companies are approaching desktop integration. By including full runtimes like Python, Node.js, and LibreOffice, OpenAI is reducing latency and potentially improving privacy by performing document processing and code execution on the user's local machine. This "thick client" approach contrasts with the "thin client" model where all processing is offloaded to the cloud.

Utilization of Open-Source Ecosystems

This development underscores the critical role of the open-source ecosystem in powering proprietary AI applications. By leveraging established tools like LibreOffice (which has a lineage dating back to OpenOffice.org) and Poppler, OpenAI can provide robust document support without reinventing the underlying technology. However, the inclusion of such large binaries also raises questions regarding disk space management and the transparency of what is being installed on a user's system under the guise of a simple AI application.

Frequently Asked Questions

Question: Why does the ChatGPT desktop app need a full copy of LibreOffice?

Based on the discovery of the "documents" plugin folder and associated "skills," the app uses LibreOffice binaries to process, read, and potentially convert various office document formats locally on the user's computer.

Question: Where is this 1.7GB runtime folder located on the system?

According to the report, the files are located in the user's cache directory at the following path: ~/.cache/codex-runtimes/codex-primary-runtime/.

Question: What other programming tools are bundled with the application?

The application includes full installations of Python and Node.js, as well as the Git version control system, allowing it to handle a variety of coding and development tasks locally.

Related News

How Google Duplex Evolved to Call Businesses on Your Behalf in Gemini
Industry News

How Google Duplex Evolved to Call Businesses on Your Behalf in Gemini

A Google Research Blog entry titled “How Google Duplex evolved to also call businesses for you in Gemini” describes its subject as the evolution of Google Duplex to support calls to businesses on a user’s behalf in Gemini. The entry is dated November 20, 2026, and the supplied article text consists only of the label “Generative AI.” The headline identifies the capability and connects it to Duplex, but the available text does not explain how the system evolved, how users request a call, or where the feature is available. The headline establishes the topic, but the supplied material does not provide a technical account or rollout details.

How AI and Lean Are Helping Automate Formal Verification to Fight Software Slop
Industry News

How AI and Lean Are Helping Automate Formal Verification to Fight Software Slop

In Latent Space, Richard MacManus examines formal verification as an emerging way to fight what the headline calls “slop.” The central claim is that proving software correct with mathematics has traditionally required specialist work, and that AI and Lean are now helping automate that process. The distinction is important: the reported change concerns how the work of proving correctness is done, not what counts as a proof. The supplied excerpt presents automation as assistance, without saying that it replaces specialists or completes proofs independently. It also does not specify which parts of the process AI and Lean help automate, how much work they save, or which software has been verified this way.

AI and Human Claim Solutions to the 167-Year-Old Riemann Hypothesis
Industry News

AI and Human Claim Solutions to the 167-Year-Old Riemann Hypothesis

TechCrunch AI reports that AI claimed a solution to the Riemann hypothesis, a mathematical problem the headline describes as 167 years old. A human also put forward a competing solution. The report, written by Tim Fernholz and published on November 20, 2026, describes mathematicians as astonished by the AI-driven result. The supplied excerpt does not explain either proposed solution, identify the AI system or the human claimant, or say whether either claim has been verified. The distinction between a claimed solution and a confirmed one is central to reading the report: the excerpt establishes that two claims are being reported, but it does not establish that either has resolved the hypothesis.