OpenAI Dev Day: The Agents are Coming
Sam sounded a bit "Attack of the Clones" but we're all trying to have fun here
OpenAI’s Dev Day gave a clear glimpse into the future of AI: agentic workflows are coming, my bet is H1 2025. And you can see the pieces coming together with the Dev Day release on October 1: cheaper APIs, real-time voice interactions, higher token limits, and a demo of the full o1 model. OpenAI is laying out the puzzle pieces for autonomous agents.
Imagine it’s early 2025 and you go to bed after assigning an agent the task of writing a full research report. It could not only draft the content but also format the document, generate references, but it could also submit it to the appropriate platform and get you paid—all without further guidance. This is going to lead to some comical errors and excellent blog posts.
It also means new modes of work. Right now, AI helps us write and generate ideas, but soon agents will take over entire workflows. Instead of manually guiding each step, we’ll need to learn new skills: delegating complex writing and submission tasks, freeing ourselves to focus on other priorities. While there’s lots of fear around whether there will be other human tasks around, I think the smarter question is to ask ourselves how AI-as-colleague is going to open up new ways to imagine our work patterns. For instance, what if our ticket hygiene was always perfect because an agent kept Jira up to date? That would be nice.
So here’s today’s sharp thought: We don’t have agents yet, but we should view the rest of 2024 as a chance to think about how we’ll live and work differently with agents. What does it look like to have agents as users in our software systems? As buyers? What skills do we need to learn to delegate more effectively with agents? The future is arriving either way, but we can be more ready if we think about it now.