Epic tells developers not to write code by hand
Epic Games has delayed an unannounced shooter set in the Marvel and Star Wars universes and demanded that the team rebuild production around AI tools — from programming to art creation. One of the project's directors, according to Bloomberg, put the task briefly: "No more handwritten code." Instead of writing code, developers are expected to correct and direct the work of AI agents.
The scale of the decision goes beyond piecemeal automation: the proposal is to run the code and graphics of an active commercial project through an AI-centric process.
The project was behind schedule
The shooter with Marvel and Star Wars characters had previously been planned for release in March 2027. No new date was given to the team. Sources familiar with the development say the project was moving more slowly than Epic Games had counted on: the company underestimated both the timelines and the number of people it would require.
The team was instructed to work through AI tools, including Claude from Anthropic. Epic's communications director Jake Jones told Bloomberg that the creative drivers of the project remain the developers themselves, and that AI is needed to speed up production.
"We are moving to AI-accelerated development led by Epic's developers — they are the ones who remain the authors and the creative driving force of everything we do."
Jake Jones, communications director at Epic Games
Shortly after publication, Epic head Tim Sweeney confirmed that the restructuring concerns the entire company. In his words, a "huge pivot" is under way at Epic: people are being reallocated in order to realign around maximum acceleration with AI. The goal, Sweeney insists, is not to do the same volume of work with fewer people, but to let the current teams ship more, and then to pass those same capabilities on to developers working on Unreal Engine.
The AI pivot against the backdrop of the March layoffs
In March, Epic laid off more than 1,000 people — about 20% of its staff. The reason was a decline in engagement with Fortnite: Sweeney admitted the company was spending substantially more than it earned. Epic additionally found more than $500 million in savings on contractors, marketing and unfilled vacancies. At the time, the company explicitly stressed that the layoffs were not connected to AI, and presented the restructuring as a way to fund the transition to Unreal Engine 6 and the next generation of the Fortnite ecosystem.
Sweeney insists this is about increasing output, not replacing people. But the proximity of the two events will inevitably sharpen the argument over whether generative tools support teams — or compensate for the resources removed during the restructuring.
The engine was already being prepared for this kind of work
Technologically, Epic has been heading toward this for more than one release. Unreal Engine 5.8 brought experimental support for MCP — large language models gained more direct access to the editor's workflows. At State of Unreal, the company presented AI-assisted development as one of the pillars of Unreal Engine 6 — alongside Verse, compatible assets and a tighter link between Unreal projects and Fortnite.
The benefit for studios is clear: less routine, faster experiments. Along with it come questions:
- the maintainability of code that nobody on the team wrote;
- authorship of the result and the data the models were trained on;
- the consistency of generated results;
- developers' ability to understand and control systems created by agents.
The use of AI in studios has long since stopped being a rarity in itself. A CESA survey showed that 85.8% of the Japanese developers who took part use generative models at least occasionally. A study commissioned by Google Cloud produced a figure of 87% of surveyed developers using AI agents — most often for testing, localization, programming assistance and processing large volumes of production data.
Sony publicly holds a more cautious line: the company says AI is capable of "unlocking the creative potential" of PlayStation studios, but in the examples disclosed the matter concerns QA, engineering support, animation processing and repetitive technical tasks, while people and performers remain at the center of the process. There is a fundamental difference between automating a specific operation and moving the code and graphics of a commercial game into an AI-centric process.
What this means for the industry
Epic's pivot looks like the continuation of a trend in which economic pressure and the availability of agentic tools have coincided in time. Companies that have been through layoffs are looking for a way to preserve their portfolio ambitions with smaller teams, and an AI pipeline looks to management like a quick answer. That is probably why statements about accelerating production sound louder than answers to questions about the quality and maintenance of the result.
If the trend continues, the nearest changes will affect the requirements placed on specialists: value will shift toward those who can set tasks for agents, check and fix generated systems, and keep architecture and visual style within bounds. For studios on Unreal Engine, Epic's promise to extend its internal capabilities to the engine also matters — then the discussion about "AI-first" will stop being the internal affair of a single company. The delayed shooter itself will serve as the test: if the move to the new pipeline does not bring the project back on schedule, the argument about accelerated production will weaken.
What lies behind such a pivot by Epic: a bet on cost savings, the failure of the shooter itself, the industry's general fatigue with expensive AAA?
Поддержки нескольких видеокарт нет до сих пор, активная камера так до сих пор выбирается через ж..у, реедер в режиме spartial невообразимо забивает всю память но зато не писать код руками. Ну может хоть ии допишет. )))
Даже на иллюстрации к цитируемой статье (Fortnite. Изображение: Epic Games) все, включая характеры/образы, позы, эстетику, формы и детали - вторичное, 100500 раз виденное и переработанное, бездарное, шаблонное, эпигонское нейросетевое ...
На что они рассчитывают? )