OpenAI launches "genius" o4 model with a programming CLI tool

The video discusses OpenAI’s launch of its new “genius” reasoning models, 03 and 04 mini, alongside the introduction of the Codeex CLI tool for developers, while reflecting on the competitive landscape of AI coding tools. Despite some skepticism about the effectiveness of these tools, the host encourages developers to embrace the advancements in AI technology while acknowledging the limitations still present in current offerings.

In a recent video, the host discusses OpenAI’s launch of two new reasoning models, 03 and 04 mini, which are being touted as “genius” level AI. This release follows closely on the heels of other updates, including GPT-4.1 and GPT-4.5. The host humorously reflects on the skepticism surrounding claims of AI genius, referencing a local saying in San Francisco about being fooled multiple times. Despite the skepticism, there’s a sense of excitement about the rapid advancements in AI technology, particularly in the realm of programming tools.

The video highlights the introduction of an open-source command-line interface (CLI) tool called Codeex, which is designed to assist developers in writing, executing, and analyzing code directly from their terminal or integrated development environment (IDE). The host expresses frustration over the current state of coding tools, despite investing heavily in various AI coding assistants. The competitive landscape in Silicon Valley is described as chaotic, with companies racing to capture the attention of software engineers, especially those looking for easier coding solutions.

The host mentions that OpenAI is rumored to be in talks to acquire Windsurf, a VS Code fork that has integrated AI features, for $3 billion. This reflects the growing trend of AI-enhanced development tools, with competitors like Microsoft releasing significant upgrades to their own coding assistants, such as Copilot. The host notes that while many tools are emerging, Google’s Gemini 2.5 is currently regarded as the best programming model available, further complicating the landscape for developers.

In a hands-on demonstration, the host tests the capabilities of OpenAI’s Codeex by attempting to create a simple YouTube clone. However, the model struggles with unclear requirements and takes a long time to process commands, ultimately resulting in empty directories. The host compares this experience with Claude Code and Firebase Studio, noting that while Firebase Studio is faster, it also fails to meet specific coding requests. The overall sentiment is that while these AI tools have potential, they still have significant limitations.

The video concludes with a balanced perspective on the current state of AI coding tools. The host advises viewers not to fall for the hype surrounding OpenAI’s genius claims, nor to succumb to the pessimism of those who deem these tools worthless. Instead, the host encourages developers to embrace the opportunities presented by these technologies, emphasizing that it is an exciting time to be in the field. The video also promotes MX, a video infrastructure API, as a valuable resource for developers looking to integrate video features into their applications.