The New 3DS build displays a native‑60 fps race with stereoscopic 3D and a bottom‑screen HUD.
F-ZERO X PORT TO NINTENDO 3DS REACHES 60 FPS IN 24-DAY SPRINT
*A 24‑day sprint turned a black screen into a native‑60 fps stereoscopic racer on the New 3DS. A second 43‑day phase added Old 3DS support and a 64DD Expansion Kit, all from a single source tree.*
The post‑mortem of G‑Diffuser, the F‑Zero X decompilation, reveals a rapid port to Nintendo 3DS that delivered a native 60 fps stereoscopic racer on the New 3DS in just 24 calendar days. The effort split into two distinct phases: an initial sprint from 2026‑08‑11 to 2026‑09‑03 that established core performance, followed by a 43‑day extension from 2026‑09‑07 to 2026‑10‑19 that added Old 3DS support, a native rendering engine, and the 64DD Expansion Kit. Together they produced three console builds from a single codebase, each bootable from the HOME menu or Homebrew launcher and fully functional without a shipped ROM.
How was the 24‑day sprint organized?
The first phase ran from 2026‑08‑11 to 2026‑09‑03, a 24‑calendar‑day sprint that moved from research to a playable race. One person directed the effort while a long‑lived Claude session orchestrated 147 subagents. Fifteen active days were logged across the sprint, with three pauses when weekly plan allowance ran out. The team committed 250 changes, expanding the port/3ds codebase to 15,039 lines and adding 48 pure‑delta patches. Parallel foundation streams tackled feasibility, boot bugs, HUD, and GPU vertex‑transform work. By the end of the sprint the build ran at native 60 Hz on New 3DS, ready for release.
What performance gains were achieved from emulator to hardware?
Initial emulator runs capped at 12–20 fps. The first hardware run on New 3DS recorded 48–50 fps, and a product‑grade build on 2026‑08‑28 reached roughly 51 fps in minimal crowd conditions. Subsequent optimizations raised median frame rate to 59.6 fps, with the tenth‑percentile floor at 51.4 fps in 30‑machine crowds. Fifty‑three percent of frames hit the 60 fps cap, up from 14 % a week earlier. The second phase introduced a native renderer that skips per‑frame display‑list walks, allowing the Old 3DS to keep pace with comparable results.
What token usage and monetary cost did the first phase incur?
Across all sessions and agents in the first phase, the workflow consumed 3.17 billion tokens. At the published API list prices—$10 per MTok for the Fable tier and $50 per MTok for cache writes—the token bill totaled $4,974. The cost page clarifies that this figure covers only the first phase; the second phase’s expenses are not included. Token consumption reflects the intensive agentic workflow, with each subagent contributing to the overall compute budget.
What features does the v1.0.0 release provide across consoles?
Version 1.0.0 delivers three console builds from a single source tree: Old 3DS/2DS, New 3DS/New 2DS XL, and New 3DS with the 64DD Expansion Kit. All builds boot from the HOME menu (.cia) or Homebrew launcher (.3dsx) and read pre‑baked asset archives from the SD card—no ROM or disk image ships with the game. The port includes Grand Prix, Time Trial, Practice, Death Race, and the machine editor, with full music, DSP‑driven effects, stereoscopic 3D, and a bottom‑screen dashboard offering three HUD layouts. The native renderer is the default on both consoles, preserving the 60 fps pacing achieved in the first phase.
Frequently asked questions
How long did the initial 3DS port sprint take?
The first phase ran from 2026‑08‑11 to 2026‑09‑03, a 24‑day calendar sprint.
What frame rate did the New 3DS build achieve after optimization?
Median frame rate reached 59.6 fps, with the tenth‑percentile floor at 51.4 fps and 53 % of frames hitting the 60 fps cap.
How many tokens and how much money were spent in the first phase?
The first phase consumed 3.17 billion tokens, costing $4,974 at the published API list prices.
The F‑Zero X 3DS port demonstrates how a tightly coordinated agentic workflow can compress years of development into weeks while hitting console‑grade performance. By unifying builds across legacy and current hardware and adding the 64DD Expansion Kit, the project sets a new benchmark for homebrew decompilation efforts. Future ports may replicate this model, leveraging native renderers and token‑based compute budgeting to accelerate delivery without sacrificing quality.
Sources: [https://hypnotize.works/gdx-3ds/postmortem/index.html]