Epic Games is using AI to write Unreal Engine code and develop features, according to executive vice president of development Marcus Wassmer. He says tools including Anthropic’s Claude and OpenAI’s Codex have helped resolve longstanding bugs and support automatic monitoring of Fortnite’s PC stability.
Game Developer’s October 28, 2026 report relays Wassmer’s comments to This Week in Video Games. His account describes a substantial change in how Epic builds its software: engineers guide AI-generated code toward an intended architecture, while most current development is concentrated in streams attached to Fortnite, Unreal Editor for Fortnite, and Unreal Engine 6.
The important distinction is between AI writing code and AI controlling development. Epic is openly embracing the former. Wassmer says engineers still direct the work, although his statement leaves important questions about verification and measurable results unanswered.
What Epic Says Claude and Codex Are Doing
Wassmer points to engine maintenance as one concrete application. According to Game Developer, tools such as Claude and Codex helped eliminate “non-deterministic cooking bugs” that Epic had been trying to resolve for a decade.
Cooking prepares game content for a target platform. A nondeterministic problem can cause runs that should behave consistently to produce different outcomes. That makes investigation difficult: an apparently successful run may simply mean the problem did not occur that time.
Epic’s example is consequential because it concerns reliability in an established development pipeline, beyond generating a new feature. If an intermittent cooking problem disappears, developers could spend less time investigating inconsistent builds and more time working on their projects. That is the practical value of the claim, assuming the fix holds across relevant workloads.
Wassmer also says AI tools help automatically monitor Fortnite’s PC stability, flagging combinations of hardware and drivers that cause problems. His account frames this as a way to find patterns that would take more work to identify manually.
These remain Epic’s reported results. The published accounts provide no bug reproductions, before-and-after stability measurements, controlled productivity comparisons, or independent validation of the claimed gains.
The two examples also describe different jobs. Fixing a cooking bug changes software; identifying a troublesome driver and hardware combination helps direct an investigation. An alert can be useful without resolving the underlying fault. I’d keep those outcomes separate when judging Epic’s progress, because detection and repair need different measures of success.
Human-Led Architecture, AI-Written Code
Wassmer’s description goes well beyond occasional code completion. In comments reproduced by Polygon on October 28, he says:
“our engineers are no longer writing the bulk of the code”
That is a significant statement about who, or what, produces the implementation. It should not be softened into a claim that Epic merely uses AI for small suggestions.
Wassmer also says engineers guide the tools toward the architectural result Epic wants. He explicitly rejects generating the engine in a single pass:
“We’re not ‘one-shotting’ the engine”
His account describes an iterative process in which developers direct AI-generated work while keeping the engine within memory, performance, and scalability requirements. Those constraints matter for a game engine: a feature can function correctly in isolation yet consume too much memory or perform poorly as a project grows.

The distinction is useful, but human direction is a description of the workflow, not a quality measurement. The cited reporting does not detail Epic’s code-review rules, testing requirements, or approval process for AI-generated changes.
My view is that architectural control is the more useful question than the label “vibe coding.” That label can cover very different practices. Epic’s explanation gives engineers responsibility for the intended result and technical boundaries. Assessing the approach requires knowing how those boundaries are checked as the generated implementation changes.
Polygon also reports that Tim Sweeney denied AI use for Unreal Engine updates about a year earlier. That provides context for the newer commitment. Without the earlier statement’s full scope in the opened evidence, it would be too strong to reconstruct exactly when Epic’s internal practices changed.
Unreal Engine 6 Is the Main Development Focus
Wassmer says almost all current development is happening in streams attached to Fortnite, Unreal Editor for Fortnite, known as UEFN, and Unreal Engine 6. His comments place AI-assisted work inside those ongoing development efforts, according to Game Developer.
He also leaves open a possible Unreal Engine 5.9 release, with a much narrower description.
| Development area | What Wassmer says |
|---|---|
| Fortnite, UEFN, and UE6 | Almost all current development is happening in associated development streams. |
| Possible Unreal Engine 5.9 | If Epic makes it, the update would be relatively modest, target specific systems, and include AI-developed changes. |
The condition attached to 5.9 matters. It is a possibility, not a confirmed release commitment. Nor does Wassmer’s statement say AI’s contribution to that update would itself be small; he describes the proposed update’s overall scope as modest.
For studios planning engine upgrades, I would separate two decisions. Epic’s choice of development tools concerns how future changes are produced. A studio’s upgrade decision still depends on the particular release: its fixes, compatibility, performance, and consequences for that studio’s project.
The UE6 connection also should not be inflated into a claim that Epic is replacing the entire engine with an autonomous rewrite. Wassmer describes ongoing development streams and iterative work under engineering constraints. His comments do not provide a UE6 release date or a complete account of its architecture.
How to Evaluate Epic’s Stability Claims
Epic is presenting AI as a way to maintain and modernize a large, aging codebase while developing features faster. adds that Wassmer cited faster feature development and more opportunity for engineers to explore ideas they previously lacked time to tackle.
Frequently Asked Questions
3 questions
1Which Claude and Codex versions is Epic using?
The cited reports name Claude and Codex but do not identify model versions or provide a reproducible tool configuration.





