OpenAI just made your entire tech stack obsolete...

OpenAI’s latest Dev Day push reframes ChatGPT from a chatbot into an app platform—an approach that could make traditional websites and mobile apps feel obsolete for many AI-powered workflows. The pitch is simple: instead of opening a separate app (say, Spotify) and manually tapping controls, users would open ChatGPT and ask it to perform actions across services. That shift matters because it turns ChatGPT into a distribution channel, with OpenAI banking on its roughly 800 million weekly active users to attract developers early. Early adopters could profit from building “ChatGPT apps” that sit at the center of everyday tasks, effectively acting like an operating system for connected actions.

The rollout isn’t starting with everything at once. The transcript notes that “apps and ChatGPT” are currently being trialed by major, recognizable apps, and it argues the model could be especially useful for high-friction services—real estate is offered as an example, where an AI could help evaluate options faster than a human-only process. Still, there’s skepticism: in many cases, a clean app interface may be easier than routing routine actions through a chatbot, which can add complexity rather than reduce it.

Beyond the platform idea, the announcement list gets more concrete for developers. OpenAI’s API updates include access to Sora 2 and GPT5 Pro, both described as available through the API. The transcript also flags pricing and cost risk: Sora 2 is listed at 50 cents per 1 second of high-quality video, translating to roughly $1,800 for an hour-long output—an expense that could quickly balloon during experimentation. For teams adding voice or image generation, the transcript mentions two smaller, cheaper models designed to reduce costs while still delivering useful capabilities.

A separate productivity play targets developer workflow. OpenAI released a GitHub action for Codex that automatically reviews code quality on every new pull request. The same automation is positioned as a multi-purpose tool—checking for security flaws, generating documentation, and handling other review-related tasks—aimed at cutting down time spent on manual code review.

The most ambitious tooling update is “agent kit,” described as a unified set of tools for building, deploying, and optimizing agents. It uses a flowchart-style canvas to let developers implement logic without stitching together as much custom code. That could broaden access for non-engineers, but the transcript warns that no-code or low-code systems often hit a complexity ceiling that eventually forces a return to traditional engineering.

Finally, the transcript adds a sponsor note: JetBrains launched an AI coding agent called Juni, integrated with the JetBrains IDE and designed to understand the context of an entire codebase. It’s portrayed as slower than some code generators but strong on context and accuracy, with asynchronous job handling via integrations like GitHub—an approach aimed at reducing context switching for developers on larger projects.