Spud and the Agent Question: What OpenAI's Next Model Means for Delivery Teams

OpenAI's Spud isn't officially announced yet. The model, codenamed Spud and often referred to as GPT-5.5, completed pretraining around 24 March 2026. But the rumours and leaks tell us something important.

OpenAI is betting heavily on agentic AI, and that bet has direct implications for how project delivery teams will work over the next 12 to 24 months.

Details are sparse by design. OpenAI hasn't officially confirmed the model's existence or capabilities. But internal signals and industry analysis suggest Spud represents a generational leap in agentic reliability.

Unlike models that excel at conversation or analysis, Spud is being optimised for autonomous task execution. It focuses on planning multi-step workflows, using tools reliably, and executing complex tasks with minimal human intervention.

This is a different kind of capability than raw reasoning power or coding ability. A model can be brilliant at explaining complex concepts but terrible at using tools reliably. A model can excel at writing code but fail at planning a sequence of actions to accomplish a goal.

Agentic AI requires a different set of capabilities. It needs the ability to break down complex problems, maintain context across multiple steps, recover from errors, and know when to ask for human input.

This focus on reliability is crucial because an agent that fails halfway through a task is often less useful than no agent at all.