Last week, OpenAI launched “Codex”, a new coding-focused AI agent within ChatGPT.1 According to OpenAI, Codex can autonomously handle software development tasks such as writing features, debugging code, and running tests. It operates on a version of OpenAI’s o3 model optimised for software engineering and is already being used by OpenAI’s internal teams and companies like Cisco and Kodiak Robotics.2
The rollout of Codex highlights the growing momentum behind agentic AI, a class of tools capable of performing multi-step tasks with minimal human oversight. Unlike traditional assistants that rely on prompts, agents like Codex can take initiative and manage more complex workflows.
The coding space has become a major battleground for this new type of AI. Startups such as Anysphere, the developer of Cursor, have seen rapid growth by offering AI assistants tailored to programmers.3 Cursor has become a daily tool for developers at companies like Instacart and Spotify, with strong adoption despite little marketing. In a sign of just how important OpenAI views this new and fast-moving segment of AI adoption, the company has recently finalised a three-billion-dollar acquisition of Windsurf, formerly Codeium, another agentic AI developer.4
Coding assistants are now one of the clearest examples of AI’s commercial traction. They help developers move faster, focus on higher-value work, and reduce the time spent on repetitive tasks. These tools also illustrate how AI is moving beyond content generation and into structured, technical workflows.
Agentic AI represents more than just an upgrade in functionality. It is redefining how users interact with software and how work gets done. As companies continue to refine these agents and scale them across industries, the ability for chatbots to “do”, and not just “talk” is becoming a major growth engine for the next phase of the AI cycle.
Access Artficial Intelligence with GXAI.
Looking for more ETF Express? Check out this week’s Market Moves and Commodity Calls.
