Epic Games Vibe-Coded Its New Launcher. The Real Story Is the "AI First" Order

What happened

Two things happened at Epic Games this week, and the internet is only arguing about one of them. On October 14, Epic put a rebuilt Epic Games Launcher into Preview on Windows, macOS and, for the first time, Linux, opening with this line: "Programming is hard. So, instead of doing that, we vibe coded a new Epic Games Launcher." (Epic's announcement)

Two days later, Bloomberg's Jason Schreier reported that Epic delayed its Marvel and Star Wars crossover shooter with Disney, originally due in March 2027, and told the team to go "AI first," using tools like Claude with humans managing the work. One of the game's directors told staff, "No more hand-written code," and added that all art for the game would be "generated with full AI workflows." (The Next Web) Tim Sweeney posted that Epic has a "huge pivot" underway to "retool for maximum AI acceleration," and says the aim is building "MORE and BETTER stuff," not doing the same work with fewer people.

Image: Epic Games, via The Next Web

What everyone is arguing about

The benchmarks, mostly. Epic says the new launcher cold starts in 5.47 seconds versus 10.07 for Steam, measured by Keywords Studios across 13 machines, with the three lowest-end machines left out, which is its own story. Ars Technica's own tests found the launcher reaches the store a few seconds faster than Steam. Meanwhile the Preview is missing social features and account options from the old app, and does not support Intel Macs.

So half the internet is arguing about seconds and the other half is asking whether "vibe coded" is just a marketing line. Both are missing the part that actually changes how you work.

The design angle nobody is leading with

A company once renowned for its engine expertise just shipped a flagship product surface with almost no hand-written code. Martin Keely, general manager of the Epic Games Store, told PC Gamer plainly: "We started from scratch. This is a completely new architecture for both the store and the launcher shell." And per Bloomberg, the directive inside a AAA game team is that hand-written code is out and full AI art workflows are in.

Here is what that means for designers and builders, and it is not "learn to code faster." When code becomes generated, code stops being the deliverable. What survives is everything that was never code in the first place:

The design system. Tokens, components, states, accessibility patterns. A machine can generate a thousand screens in an afternoon, but it cannot decide what your product should feel like. That decision now lives entirely with you.

The interaction spec. When a team prompts instead of programs, the prompt has to carry the flow logic, edge cases, empty states and error behavior. Writing that spec is UX work, not engineering work. Think of it as the new code review.

Art direction as pipeline design. "All art generated with full AI workflows" does not delete art direction, it relocates it. The art director's job shifts from approving individual assets to designing and curating the system that produces assets. That is the exact shift UI designers already lived through with design systems. Taste moves upstream. It does not move out the door.

My honest position

I think this is mostly good news, with one caveat. Epic is not telling creative people to stop creating. It is telling them their leverage now lives in systems and judgment. I have believed this for a while: the people shipping genuinely interesting work in 2026 are the ones who can fully specify a product and then direct machines to build it. Epic building Claude and Gemini into Unreal Engine 6 is the same bet, applied to the engine itself.

The caveat: AI-first without brutal review is how you ship a Preview that is missing features and benchmarks that quietly skip the three weakest machines. Vibe coding amplifies whoever holds the spec. If you do not have a spec, you do not have a product, you have a slot machine.

What to do about it

1. Write spec-grade briefs. For your next project, write the interaction logic, the states and the constraints before a single screen gets generated. The quality of what comes out will track the quality of what goes in.
2. Build the system before the screens. A token set and component rules will outlast any individual layout, and it is what AI tools actually need from you to stay on brand.
3. Review like an editor. Treat generated code and generated art the same way: fast to produce, slow to verify. Your review pass is the product.
4. Ship something end to end this month. The gap between people who talk about AI-first and people who have shipped with it is widening fast. You can take a UI from idea to working app with an AI app builder, and Base44 is the one I use and recommend (affiliate link).

The takeaway

Epic vibe coding its launcher is not a stunt and it is not doom. It is a AAA company telling you where the work went: into systems, specs and judgment. The benchmarks will keep improving. The one thing nobody can generate is taste, and as of this week, it is officially the most valuable thing in the building.

So here is the question I would sit with this weekend: if code were free tomorrow morning, what would you spend your taste on?

Related on NewForm

• Smart TVs Got So Ad-Heavy That Owners Are Vibe-Coding Their Own Home Screens
• Claude Now Refuses to Recreate Existing Games. Your Prompts Need to Change
• Pieter Levels Vibe Coded a GTA: San Andreas Mission. The Real Masterpiece Is the Prompt