"Programming is hard. So, instead of doing that, we vibe coded a new Epic Games Launcher."
That is not a line from a forum post or a developer's social media joke. It is how Epic Games opened the official announcement for a rebuilt Epic Games Store client, a piece of software that millions of people use to buy games, manage accounts, and process payments.
The pitch that follows is a loud, specific competitive claim: a launcher that cold-boots 46% faster than Steam, built by AI agents, shipping with performance numbers generated by a vendor Epic itself commissioned, and released as an unfinished preview that Epic describes as "light on features." It also arrives with a first-ever Linux build attached. That combination is what makes this launch worth examining closely, not as another launcher update, but as the first real test of whether AI-assisted development can ship consumer software that holds up under scrutiny.
What Epic Actually Shipped
The new launcher is available now at launcher.epicgames.com as a preview build for Windows, macOS, and Linux. According to Epic's official announcement, the client was rebuilt entirely from scratch on a new tech stack and architecture, covering both the shell (the download manager and library) and the store itself.
The headline figures are laid out plainly. Epic claims a cold boot time of 5.47 seconds, compared with 10.07 seconds for Steam, a 46% improvement. System tray restore comes in at 0.36 seconds against Steam's 0.42 seconds, a 14% gain. The library reportedly loads 30 times faster than the previous launcher, startup is twice as fast, and the whole application occupies just 300MB.
Worth noting: Epic previously floated a projection of 5x faster boot times during earlier development discussions. The delivered numbers fall well short of that, landing at roughly 2x. That gap between promise and product is the first hint that this launch deserves a careful read rather than a victory lap.
On the feature side, the preview includes a redesigned library that can detect and import games from Steam and the Xbox PC app, a persistent cross-game voice chat system called Epic Parties, a customizable navigation bar, and an in-app feedback tool for developers. Migration support lets users carry over installed games, settings, and cloud saves.
The Benchmark Problem
The performance comparison was conducted by Keywords Studios, an independent vendor. It was also commissioned by Epic. Those two facts are not contradictory, but they matter enormously for how much weight the numbers should carry.
No independent outlet has verified the figures. No third-party reviewer has published its own cold-boot tests on matched hardware under controlled conditions. Until that happens, every speed claim in Epic's blog post should be treated as vendor-commissioned data, not established fact.
There is also a numbers issue worth clearing up. Some coverage has framed the result as the launcher being "1.8x faster than Steam." That is the same claim as the 46% figure, just expressed as a ratio: 10.07 divided by 5.47 is approximately 1.84. The underlying data is identical; only the framing changes. Epic's own precise numbers, 5.47 seconds versus 10.07 seconds, are the ones to hold onto.
Why does the source of a benchmark matter so much? Because methodology is everything in performance testing. What hardware was used? Was Steam running with the same background processes, the same library size, the same cache state? Was the comparison run on a cold system, or with warm caches that favor one client? A commissioned benchmark is not inherently dishonest, but it is inherently shaped by the questions the commissioner wants answered. Readers should wait for third-party testing before drawing conclusions about which launcher is actually faster in everyday use.
What 'Vibe Coding' Actually Means Here
The term "vibe coding" was coined by Andrej Karpathy in February 2025 and was later named Collins Dictionary's Word of the Year for 2025. It refers to the practice of directing AI agents to write software, often with minimal human intervention in the actual code generation.
Epic Games Store GM Martin Keely offered a more supervised description of how it worked here. According to Keely, AI agents wrote and reviewed all of the code, but the team supervised the entirety of the process. AI was used for code efficiency and architecture, not for artwork or design.
There is an important distinction to draw, and Keely drew it himself. The launcher team did not adopt the "no more hand-written code" model that Bloomberg reported was mandated elsewhere at Epic. Those are separate initiatives, and conflating them misrepresents what happened on this project. That distinction matters again when placed against the company-wide AI pivot described below, but the launcher team's approach should be evaluated on its own terms.
Still, the concerns critics raise about vibe coding apply directly to a storefront. A launcher handles payment information, account credentials, and entitlement checks. Code that was generated and reviewed by AI agents raises legitimate questions about accountability when something breaks, long-term maintainability as the codebase grows, and whether AI review catches the same classes of security vulnerabilities a human reviewer would. Epic says the process was supervised. Whether that supervision was sufficient is a question only time and security research will answer.
The Bigger AI Pivot, and the Layoff Context
The launcher launch does not exist in a vacuum. CEO Tim Sweeney has announced a "huge pivot underway" to retool Epic for what he calls "maximum AI acceleration," framing the shift as a way to boost output rather than reduce headcount. Bloomberg reported an "AI first" mandate on an unannounced Star Wars and Marvel shooter, and that a director on an unannounced Disney title told staff "No more hand-written code." As noted above, these are separate initiatives from the launcher team's supervised approach, but they form the backdrop against which the launcher's "vibe coded" branding will be read.
In March, Epic laid off more than 1,000 staffers. Sweeney said those layoffs were not related to AI. That statement and the company-wide AI pivot can both be true, and there is no evidence the layoffs and the launcher's development are connected. But their proximity in time shapes how the announcement is received: a company telling the industry it is retooling around AI, weeks after cutting a thousand jobs, while shipping an AI-built product with unverified performance claims, invites a level of skepticism that a normal launcher update would not.
Linux, Fortnite, and the Long Game
The Linux build is the most concrete piece of good news here. It is the first time Epic has shipped a launcher for the platform, though engineering director Rick Deist has said it will not work on every distribution. Epic is targeting a set of security features that cover many, but not all, Linux distros.
The bigger prize is Fortnite. Epic is rebuilding Easy Anti-Cheat as "EAC Linux v2," detailed in an official Epic Online Services announcement, and says Fortnite will come to Linux once the system works with a critical mass of distribution maintainers. That is a conditional promise, not a date.
In practice, this release is a soft beta in everything but name. Keely described it as shipping "in lieu of a beta," and Epic is not pushing it to existing users. Adoption depends on word of mouth while the team collects feedback. Migration support for installed games, settings, and cloud saves lowers the switching cost for anyone willing to try it.
The competitive backdrop is hard to ignore. The Epic Games Store launched in December 2018 as a direct Steam challenger, built on a 12% revenue cut and weekly free-game giveaways. Its launcher has been criticized for years as slow and feature-poor compared with Steam, which has dominated the market since 2003. Rebuilding around speed is a pointed shot at the one weakness Epic's critics have cited most consistently.
The Verdict That Hasn't Arrived Yet
Epic is making a loud, specific, competitive claim about beating Steam at the thing it has been most criticized for. That deserves attention. But the evidence is vendor-commissioned, the client is an unfinished preview that Epic itself calls light on features, and the "vibe coded" framing raises real questions about software that handles payments and accounts.
Rather than declare a winner, here is what to watch. Independent benchmarks on matched hardware, run by outlets with no relationship to Epic. Security research into how AI-written code holds up when someone tries to break it. Whether the feature-light preview fills out into something that can genuinely compete with Steam's maturity. And whether Fortnite on Linux actually materializes, or remains a conditional promise tied to distro maintainers Epic does not control.
The meta-point is larger than any launcher. This release is a public experiment in whether AI-assisted development, supervised by humans but executed by agents, can ship consumer software that survives contact with real users at scale. For the first time in a major consumer software launch, the central question is not simply whether the product works. It is whether the process that built it can be trusted at all, and Epic has offered no independent evidence either way.






Comments
Join the Conversation
Share your thoughts, ask questions, and connect with other community members.
No comments yet
Be the first to share your thoughts!