This article is part of a series showing internal examples of how OpenAI applies its own technologies and APIs. The tools described are used internally at OpenAI and are presented here to illustrate how frontier AI supports use cases across teams.
Role and challenge
For Chad Nelson, Creative Specialist at OpenAI, Codex has moved beyond an engineering tool to become a collaborative creative partner. Chad’s role is to explore what OpenAI models enable for creative teams and translate those possibilities into demos, workflows, and ideas that brand partners can use.
Accustomed to the limitations of standard creative software, Chad values that Codex can consider more project context: it can know which project he’s on, which brand is involved, and what the strategic goals are. As Chad puts it: “The tools I’ve been using my whole career have no idea what project I’m on, what brand I’m working for, what product I’m trying to pitch. Codex and ChatGPT have the context.”
What Codex helps with in creative work
- Context handling: Codex can ingest and interpret brand books, style guides, fonts, and composition rules, allowing it to better align with a project’s creative needs.
- Campaign direction generation: A campaign’s direction is influenced by the brand, the product, and the emotional tone. Chad found Codex can combine those inputs in ways traditional tools cannot.
- Building custom tools and workflows: In natural language, Chad can describe a workflow and ask Codex to help implement it—connecting API keys, creating prototype interfaces, or designing custom UI elements such as sliders and controls for visual exploration.
Chad describes the capabilities: “With Codex, I can actually create tools, user interfaces, sliders, and controls. I can adjust camera composition or lighting, and create real-time shadows with depth and dimension.”
Faster ideation and stronger pitches
Codex speeds up creative exploration. Recently Chad had one day to develop a broad set of campaign ideas for a new client. Under normal circumstances he might have developed five directions, or perhaps ten with more time, many of them small variations on the same concept.
Using Codex, Chad provided project context—the brand and its products, some guidelines, the client’s goals, and the intended tone. In a single day, Chad and Codex generated 50 campaign directions; Chad then reviewed and distilled those into the 10 strongest ideas. “What I found is Codex is allowing me to riff creatively and to explore all these different options with a greater sense of freedom,” he says. With more concepts to choose from, he brought a stronger set of options into the pitch despite the tight deadline.
Reducing the gap between idea and execution
When Chad has a prototype in mind, he can describe it and begin building immediately with Codex. If workflows require new modalities or API integrations, he can ask Codex to help connect them. If he encounters an obstacle, he explains it in natural language and works through it without breaking the creative flow.
Chad’s experience illustrates broader ways creative teams are using AI: to build systems that create, evaluate, and refine ideas as work progresses. As Chad summarizes: “One thing that I think I’ve learned through using Codex is that if you have an idea for a creative workflow or a solution or an interface, just ask it. Tell it what you’re thinking of, and suddenly in 10 minutes you’ll be looking at a prototype for it.”
Why this matters
These internal examples show how generative models can shrink the distance between idea and implementation and provide creative professionals with new tools for faster iteration, wider experimentation, and more efficient technical solutions.
