OpenAI DevDay 2026: Early Tests of Dots, Spaces and Fast Models

Claire Vo shares early impressions from OpenAI DevDay 2026, including a personal agent, collaborative Spaces, connected Sites and developer tools. These are her initial tests and opinions, and product availability and pricing can change.

Vo describes testing a Dot that can work through several interfaces and delegate tasks. She likes some of the results and proactive suggestions but says the relationship between Dot, chat and Codex interfaces remains confusing, so she withholds a full review.

She sees Space as a shared place for people and agents to work on documents and other artifacts. Her example is a collaborative scratchpad for tasks, notes and decisions; whether organizations adopt it remains to be seen.

Vo describes Sites as a way to share a deployed app and use connected plugins as data sources. She emphasizes that a viewer’s access depends on their own connection and permissions, and sees this as relevant to internal tools.

She discusses GPT-6.1 Sol in terms of her preferred balance of capability, speed and cost. Her model preference and the event’s quoted prices are specific to that time and workload.

Vo demonstrates early access to a Decisions API with vision, using it to select video frames for a thumbnail and classify a hot dog image. She says the API was not yet generally released when she recorded and plans a more formal comparison.

She experiments with a faster, more expensive model setting in a collaborative drawing app and a small 3D world that changes in response to prompts. The examples show possibilities and some latency, not a general performance benchmark.

Vo reports spending about $97 on roughly half an hour of the interactive game experiment. The example illustrates how fast inference can create new experiences while making cost measurement important; it is not a universal price estimate.

She closes by recommending that teams explore the collaboration tools and that developers evaluate the new model and API options for their own tasks. Her review remains preliminary.