What happened
This Wednesday Epic Games put out a Preview of a rebuilt Epic Games Launcher, the desktop app people use to buy, install and start games from the Epic Games Store. The first line of the announcement is the reason everyone is talking about it:
"Programming is hard. So, instead of doing that, we vibe coded a new Epic Games Launcher that's way faster and more responsive than the current one, and has a new look."
The same post calls the release "light on features" and asks you to think of it "as a speedboat and not a yacht". It runs on Windows, macOS and, for the first time, Linux. Your account, purchases, installed games, cloud saves and progress carry over, and bugs go through a "Send Feedback" button inside the app.
Epic also posted on X: "We vibe coded the Epic Games Launcher to be faster than the other guy." The other guy is Steam, and the launch was sold almost entirely on speed.
What Epic actually shipped
Strip away the joke and the Preview is a fairly ordinary list of client features, the kind any frontend team would recognise:
- Faster cold starts, quicker switching between library and store, and less load while the app sits in the background.
- A page per game with its status, activity, details and play controls in one place.
- Collections, a grid view and a list view, and a top bar you can customise.
- Search that filters as you type.
- A downloads view showing download, read and write speeds, plus active, queued, scheduled and failed downloads.
Fortnite on Linux is not there yet. Epic says it is rebuilding Easy Anti-Cheat and working with distro maintainers first.
The speed numbers, and what they show
The Preview site puts the comparison front and centre. Epic's figures:
- Cold start: 5.47 s for the new launcher against 10.07 s for Steam, which Epic rounds to "1.8x faster startup than Steam".
- Opening the store: "3x faster than Steam".
- From the background to the library: 0.36 s against Steam's 0.42 s, or "1.2x faster".
- A headline that reads "Library, 30x faster", without saying faster than what.
Each claim carries an asterisk, and the page does not say what hardware, network or cache state sits behind it. Ars Technica tested it and found the cold start to the store page "a few seconds faster than Steam", which supports the biggest claim. On the tray-to-library number, Ars says the 0.06 seconds gap is "so small as to be unnoticeable", and it was in their testing too.
For context, Ars also notes that at Unreal Fest in June Epic had promised a "5x faster cold start" for Launcher V2. That target was against Epic's own old launcher, not Steam, so the two numbers do not contradict each other, but they are not the same promise either.
Then Steam answered
A day later, RespawnFirst reported that Valve had pushed a Steam client update and that Steam's cold boot was now 5.23 seconds, just ahead of Epic's figure. The article does not say how it measured that, and nothing on the Steam side confirms it was a reply to Epic. Still, the story took off: the Hacker News thread reached 329 points and 343 comments.
The interesting part of that thread is that almost nobody argues about AI. They argue about whether startup speed should ever have been this bad. One commenter points out that asking a model to "rank the top causes of slow startup" and then fix them is "one of the most trivial things you can do with AI". Another asks the obvious question: test both launchers on the same machine before declaring a winner. Neither company's numbers do that in public yet.
What "vibe coded" means here
"Vibe coding" usually means describing what you want to a model and accepting what comes back without reading it closely. Epic's own words are looser than that. Martin Keely, who runs the Epic Games Store, told IGN (as quoted by Ars) that AI "made us more efficient and allowed a small team to rewrite the code base and be able to launch it faster across platforms". Earlier, he told PC Gamer (again via Ars) that the team "started from scratch" and used AI as a tool to write and review new features.
So the honest reading is: a small team did a full rewrite with heavy AI help, shipped it to three desktop platforms, and chose the most provocative possible phrase for the marketing. That fits the wider company picture. Days earlier, Epic CEO Tim Sweeney wrote that Epic has "a huge pivot underway" to "retool for maximum AI acceleration", per the same Ars piece.
Why this matters if you build web apps with AI
Three things are worth taking from it.
Speed became the feature again. A large company launched a client with fewer features than the old one and led with startup time. When a rewrite gets cheap, the old excuse — "we cannot touch that, it would take a year" — gets weaker, and performance becomes something you can compete on. That applies to a dashboard or a Next.js app as much as a game launcher.
A rewrite is now a realistic option, which makes it a risky one too. If a small team can regenerate a code base, the hard part moves to what the old code knew: edge cases in payments, odd download states, the bugs that were fixed years ago and never written down. Epic is honest that the Preview "will have some bugs". The interesting test is the next six months, not launch day.
Publish numbers people can check. Epic's page has numbers, which is better than most launches. But numbers with an asterisk and no method invite exactly the reply it got: a competitor or a reporter shows a different number the next day. If you claim your app is faster, say on what machine and how you measured.
What is not known yet
- Which AI tools or models Epic used. Neither the announcement nor the Preview site says.
- What the launcher is built on. Epic has not said whether it is a web view, a framework like Electron or Tauri, or native code.
- How the Steam comparison was measured. Hardware, cache state and network are not given.
- Whether Steam's update was a response, and how the post-patch figure was measured.
- How much of the code a person reviewed, and what testing guards purchases. "Some bugs" is all Epic has said.
My take (Amar)
I like that Epic said it out loud. Plenty of teams lean on AI heavily now and describe it as "engineering velocity" in a slide nobody reads. Putting "we vibe coded it" in the first sentence of a launch is a bet that users care about the result, not the process, and I think that bet is mostly right. Nobody opens a game launcher to admire the code.
What I would not copy is the measurement. A speed claim is the one thing a frontend developer can verify in an afternoon, and Epic left the method out, so the conversation became a fight about one number on unknown machines. If I shipped a rewrite like this, the speed page would say what device, cold or warm cache, and link a script anyone can run. The AI part of this story is less interesting than the boring part: a fast client is a choice someone has to keep making, by hand or by prompt, every release.
