Ece Özbilen reports from San Francisco:
What happened at OpenAI DevDay 2025?

The first time I visited San Francisco was eight years ago. The city welcomed me with heavy rain and fog so thick that, even with the Golden Gate Bridge right in front of me, I couldn’t see it. I went back to Istanbul sick and disappointed. This time, the city was completely different. It was a balmy week in October, the temperature close to 30 degrees, the sky clear, and the streets filled with driverless Waymo cars. I was back in San Francisco for OpenAI DevDay.

At this event, OpenAI shares its strategic decisions, vision, and latest technologies, models, and tools through live demos and sessions led by its own team. The goal is not merely to exchange information but to inspire developers, encourage them to create the next generation of AI solutions, and in doing so, expand the ecosystem and help the field of artificial intelligence mature. Held physically in San Francisco, DevDay also reaches millions of viewers around the world through live and recorded broadcasts.

For anyone wishing to attend OpenAI DevDay, the process is transparent but highly selective. The call for applications is usually announced during the last week of July. Submissions remain open for about a week, and anyone can apply individually.

To briefly explain the process: you start by filling out a short online form that asks for basic information.  What truly makes a difference, though, is the section where you describe yourself. The clearer and more authentic your response, the more likely you are to stand out.

Results are announced in early August. If accepted, you have just one week to complete your registration and pay the 650-dollar fee. After that, the OpenAI team keeps you informed with regular updates to help you prepare for the event. And before the journey even begins, the first emotion everyone seems to feel is the same: curiosity.

Attending OpenAI DevDay in person for the first time felt like stepping into a world I had only followed online until now. Everyone I met came from different countries, cities, organizations, startups, universities and individual projects. It was a clear reminder of how diverse, rich and inclusive the OpenAI ecosystem really is.

Throughout the event, one thing stood out for me: the limited representation of women in technology. Even at a global gathering as influential as DevDay, the number of women participants was noticeably low.

Meeting people with different perspectives, experiences and stories not only gave me an opportunity to broaden my technical knowledge, it also deeply inspired me.

Now, I’ll try to summarize the most remarkable moments of the event, focusing on the three main sessions that shaped the impact and direction of this year’s DevDay.

The event began with these words from Sam Altman. That sentence perfectly captured the journey OpenAI has taken over the past three years. Since the very first DevDay in 2023, the OpenAI ecosystem has expanded at remarkable speed, clearly showing that the transformation in artificial intelligence will only continue to accelerate. The data Altman shared in his keynote speech, comparing the past three years, also revealed that this transformation is set to move even closer to the center of our lives in the years ahead.

• Developers: The number of developers building with OpenAI has doubled from 2 million to 4 million.

• Users: ChatGPT’s weekly active users have jumped from 100 million to 800 million.

• API usage: The number of tokens processed through the API has rocketed from 300 million to 6 billion per minute.

This year’s DevDay opened with a new and rather special tradition.

Teams that built applications processing 10 billion, 100 billion, and even 1 trillion tokens were invited on stage to be recognized for their achievements.

To put 10 billion tokens into perspective, that’s roughly the equivalent of 40 million pages of text or reading the entire Harry Potter series 400,000 times.

OpenAI described the commemorative plaque given to developers as follows:

“We launched this project to recognize the real-world change our users on the OpenAI Platform are making, and the milestones you've reached along the way... illuminating this major moment and reflecting our bright future together.”As of today, only 141 companies have reached these remarkable milestones.

In the opening session, Sam Altman outlined Open AI’s platform strategy and the latest developments under four main themes.

• First, ChatGPT is evolving into an application platform.

It will soon be possible to access services such as Booking.com, Figma, Spotify, Coursera, Canva, Zillow and Expedia directly through the ChatGPT interface. In short, ChatGPT is becoming a kind of “super app” where different needs can be met in one place.

Watching the live demo and seeing how easy it is to interact with a learning video on Coursera, almost like having a conversation with the app, or create a dynamic and playful poster using Canva, helped me imagine how soon this innovation could generate an inspiring new level of interaction between applications within ChatGPT itself.

• Another notable innovation was the introduction of a toolkit called AgentKit.

AgentKit can be seen as a specialized set of tools designed for developers and organizations on the OpenAI platform to simplify the transition from prototype to production in AI agent solutions.

It’s a comprehensive suite built to securely create, deploy, and manage agents. The set includes Agent Builder for creating visual multi-step workflows, ChatKit for building conversational interfaces for end users, Connector Registry for securely connecting agents to internal systems, data, or third-party services, and finally Evals for evaluating agent performance.

It's exciting to see the prediction that “2025 will be the year of agents” already coming true. OpenAI has successfully captured this transformation and provided developers with a practical, accessible, and secure infrastructure.

That said, for those who have already begun integrating agents into workflow automation tools such as n8n, there’s still room to explore and advance within this new platform.

• In the AI-assisted software development process, artificial intelligence is evolving from a co-pilot into a true business partner.

Since it was a developer-focused event, nearly every session featured live demos that made participants an active part of the process. Perhaps the most impressive and entertaining moments for everyone were the spontaneous demos where OpenAI interacted directly with the conference hall’s sound and camera systems.

For engineering teams, the most notable announcement was the Codex integration with Slack, allowing software teams to receive direct coding support within chat environments. In addition, new monitoring tools, environment controls, and analytics dashboards were introduced for large-scale use.

Finally, the updates for GPT 5 Pro, Sora 2, and the Mini Realtime APIs signaled that OpenAI aims to give developers even greater flexibility in its next phase of growth.

Altman closed the session with these words: "We're watching something significant happen, I think. Software used to take months or years to build - you saw that it can take minutes now. To build with AI, you don't need a huge team, you need a good idea and you can just sort of bring it to reality faster than ever before."

Greg Brockman, who co-founded OpenAI alongside Sam Altman, Elon Musk, and Ilya Sutskever and continues to serve as the company’s president, led the session titled “Developer State of the Union.” The discussion began with OpenAI’s founding mission and its journey of growth to date.

It’s truly striking to realize that the models which have so profoundly transformed our perspective on artificial intelligence were born from such a simple vision. As Steve Jobs once said, “Simple can be harder than complex, because you have to work hard to get your thinking clean to make it simple.”

One of the most intriguing parts of this session was the information shared about OpenAI’s open models (GPT OSS). These models offer developers the flexibility to build applications that can run locally, offline, or on-premises within company systems. In short, organizations can now use AI not only in the cloud but also freely within their own infrastructures.

The numbers clearly show the scale of this impact:

·       In just two months, the models were downloaded 23 million times on the Hugging Face platform.

·       Developers from 147 countries participated in a hackathon organized around these models.

One example from the hackathon was particularly striking:

A decentralized “mesh agent” was developed for use during natural disasters when communication networks collapse. This agent can connect people offline, translate languages instantly, and direct users to the nearest emergency resources.

These examples highlight the potential of open-source and small language models to bring artificial intelligence “into people’s pockets.” They also suggest that we are on the verge of a new wave of transformation in this direction.

All these innovations were showcased through extensive live demos. Those interested can access the session recordings and materials on the event’s official website.

The day’s most anticipated session, “A Conversation with Sam and Jony” , brought Sam Altman and Jony Ive together on stage.

Sir Jony Ive is undeniably one of the legends of the technology and design world—the creative mind behind iconic products that have come to define an era, including the iMac, iPod, iPhone, iPad, and Apple Watch.

What perhaps most distinguishes him from other designers is not only his ability to shape form but his power to redefine and transform our relationship with technology itself.

For this reason, the announcement that OpenAI would collaborate with him on its first hardware product created great excitement in the tech world. The product was expected to be unveiled by the end of the year and to boost OpenAI’s valuation by nearly one trillion dollars. As the year draws to a close, however, the launch now appears to be postponed, with the earliest expected date being 2027.

While watching this session at the event, one question kept running through my mind: would there be any new information about the process? So far, the only thing we’ve learned is that the product will not be a phone. By the time the session ended, it was clear that this curiosity remained unresolved — not only for me, but for many others in the audience who left with the same question lingering in their minds.

The session began in a warm and conversational tone, focusing mainly on design philosophy and product vision. Rather than concrete technical details, the emphasis was on the values shaping the product and the perspective it offered for the future.

Altman noted, “We have entered an era where computers can see, think, and understand.” Yet despite this revolution, he added, user experience is still defined by outdated products and interfaces. In his view, good design is what will bridge this gap.

Ive said, “We are standing at the threshold of a new generation of technology,” and expressed his belief that these technologies have the potential not only to make us more productive but also to help us become better versions of ourselves. For him, the goal is not merely to create physical objects but to build a mental space of interaction around them.

Altman closed the conversation with a question that perfectly captured the spirit of the day:

·       "AI will clearly change the world in all sorts of ways, some really huge. What do you most hope that we get right? As we go through this technological revolution, what do you most hope we discard or preserve or put out?"

Ive’s response was both simple and thought-provoking:

·       "I think that it would be these tools - this is almost too naive sounding to say it - but that they make us happy and fulfilled and more peaceful and less anxious and less disconnected."  
We may have to wait a little longer for the launch, but this session showed that what is truly worth waiting for is not just the product — it’s the vision that will make it possible.

It's been almost three weeks since OpenAI DevDay. What stayed with me after the event was not just the new model updates or live demos, but rather the strong impression of a team evolving into a more systematic and professional structure. It’s clear that OpenAI is undergoing a transformation aimed at organizational maturity.

Since the event, this impression has been reinforced by concrete steps. Last week, OpenAI completed its long-planned restructuring process. The company is now organized under two entities: a nonprofit foundation (OpenAI Foundation) and a for-profit business arm (OpenAI Group PBC).

All these developments signal a powerful shift not only in the world of artificial intelligence but also in how organizations of the future will operate. So how is this transformation affecting us — and how will it continue to do so?

Perhaps the answer lies not only in technology itself but also in how we think about it.

The increasing affordability, accessibility, and individual-level adoption of AI means that employees are now adapting faster than their organizations. This clearly shows the direction of change: artificial intelligence is spreading not “top-down” from management but “bottom-up” from employees.

For organizations, this represents both a warning and an opportunity. As the World Economic Forum’s recent Chief People Officers Outlook report emphasizes, the key to navigating this change is human-centered transformation — rethinking organizational structures, roles, processes, and talent strategies.

Of course, embracing technology alone is not enough; governance models for data and AI are equally critical. Sustainable transformation requires clear frameworks and transparent communication on how data is used, who oversees AI decisions, and how accountability is ensured.

In short, the period ahead is not about “experimenting with AI,” but about “rethinking with AI.” The real difference will not come from the size of a company’s technology investments, but from who uses artificial intelligence, for what purpose, and how.