Developer Tools

Codex for almost everything

The model behind GitHub Copilot now translates English to SQL, Excel formulas, and 12+ languages.

Deep Dive

OpenAI has significantly broadened the application scope of Codex, the AI model that powers GitHub Copilot, by making its API generally available. Initially focused on code completion and generation, Codex now demonstrates proficiency across a surprising range of tasks. Early adopters have deployed it in over 70 different applications, showcasing its ability to translate natural language into SQL queries, generate complex Excel formulas, and even convert commands between more than a dozen programming languages. This move signals a shift from Codex being a specialized coding tool to a general-purpose engine for interpreting and executing intent.

The expansion turns Codex into a foundational layer for building natural language interfaces across software. Developers can now integrate the API to create features where users instruct applications using plain English. For example, a user could ask a data dashboard to "show me last quarter's sales by region" and Codex would generate the necessary database query. This reduces the need for users to learn specific query syntax or formula structures, lowering the technical barrier to using powerful software. The model's performance, trained on both natural language and billions of lines of public code, allows it to understand context and nuance in user requests.

Key Points
  • Codex API is now generally available, powering over 70 third-party applications.
  • It translates natural language to SQL, Excel, and 12+ programming languages beyond just code completion.
  • Enables developers to build software that users can command with plain English instructions.

Why It Matters

Democratizes complex software use by letting people command tools with English, not code syntax.