Latest
← Signal: Tech & AI

Epic Games Launcher rebuilt with AI coding, claims faster performance than Steam

Confirmed1 source · Oct 14, 2026

Epic released a new Preview build of its Games Launcher using AI-assisted development, claiming startup and restoration speeds that exceed Steam's.

Epic Games Launcher rebuilt with AI coding, claims faster performance than Steam
Image via Ars Technica

What happened

Epic Games announced a publicly available Preview build of a newly rebuilt Epic Games Launcher on October 14, 2026, developed using AI as a tool to write and review code from scratch rather than modifying existing code. Epic Games Store General Manager Martin Keely stated the team used AI to replicate existing functionality in a different tech stack, allowing a small team to rewrite the codebase faster and support Windows, Mac, and Linux. Epic claims the new Launcher achieves faster cold startup to the store page and slightly faster restoration from system tray to Library compared to Steam. Ars Technica's testing confirmed the new Launcher reaches the store page faster on cold start, though the claimed 0.06-second difference in restoration speed appeared negligible in their testing.

Context

Epic has operated its Game Store and Launcher as a competitor to Valve's Steam for nearly eight years. The new build represents a significant engineering approach—using AI to rebuild core functionality rather than iteratively updating it—potentially signaling how game platform developers may leverage AI for competitive development velocity. If performance claims hold in wider testing, this could address a longstanding complaint about the Epic Games Launcher's responsiveness relative to Steam.

What's disputed

The practical significance of performance differences is unclear; Epic claims the new Launcher is 'way faster and more responsive,' but Ars Technica's testing found the restoration speed advantage (0.06 seconds) too small to notice in practice. Broader performance comparisons across different hardware configurations and use cases remain untested in the source material.