OpenAI DevDay 2026
My takeaways from DevDay in San Francisco: useful AI applications, discovery, marketing, competing ecosystems, and the value of human interaction.
I attended OpenAI DevDay in San Francisco. The announcements included ChatGPT Space, Dots, Ultrafast, and updates to Codex.
Explore my full field notes, with the launches, sources, and photos from the day.
Here’s what I’m taking away:
AI’s application era is becoming more concrete
Dots and Space put ongoing tasks and shared work at the center of the experience. As models become more capable, the challenge is turning those capabilities into products that fit how people actually work.
You don’t know what you don’t know
Models can already do far more than many people realize, but it’s hard to ask for something you don’t know is possible. That makes discovery part of the product experience. Good examples and clear starting points help people see what AI could do for their own work and craft.
As building gets easier, marketing matters more
Making a product doesn’t mean people will notice it or want to use it. Finding your users means understanding their needs and culture, then finding a creative way to reach them at the right moment.
I see two approaches developing alongside each other
Open-weight models and tools that support multiple providers give companies more choice in what they build with. Model providers are also building integrated products around their own models, which means they can compete with the products built on those same models.
Human interaction may become something we pay more for
As companies automate more customer interactions, access to a person could become a premium service. I can see that mattering most when our situation doesn’t fit a standard process and we need someone to understand the context and exercise judgment.