The news
Decomp.dev has removed its public leaderboard of game decompilation projects. The operators say AI-assisted forks have reduced visible progress metrics to a contest over raw completion numbers while flooding the scene with unfinished ports. The change ends open ranking for retro game reverse-engineering efforts that once let contributors compare code coverage and function matches across teams.
Context
Until the shutdown, decomp.dev functioned as a shared index where multiple groups tracked their work on classic titles. Projects displayed percentages for reversed code and matched functions, creating a public record of ongoing preservation work. Once AI tools entered the picture, new forks began copying established repositories, running automated generation on the remaining sections, and releasing PC ports ahead of the original teams. The site’s creator points to this pattern, rather than any limit in decompilation techniques, as the direct reason the public index no longer served its purpose.
Details
The closure applies only to the public leaderboard. Teams that request continued private tracking can still use the platform for internal progress logs, but the open rankings that drove external attention are gone. The stated trigger is a surge of forks that reach high completion percentages by applying AI to existing codebases and then publish rough ports. These forks emphasize visible speed and numeric milestones over code accuracy or long-term maintainability.
Original project members have noted that their repositories reach a certain stage only to see duplicate efforts appear shortly afterward with similar completion claims. The new forks often require little additional human review before release. Site operators describe the outcome as a loss of reliable community signal: raw counts no longer distinguish careful, sustained work from automated duplication.
No new moderation rules or technical filters were added before the decision. The operators instead chose to withdraw the public index entirely rather than attempt ongoing enforcement against the forks. The pattern emerged after several well-known decompilation projects reached partial completion, at which point multiple AI-driven copies surfaced with faster port releases.
Why it matters
Public leaderboards measure what can be counted. When AI reduces the cost of producing those counts, the numbers stop reflecting sustained human effort and start reflecting access to automation. Decomp.dev’s choice to drop the scoreboard rather than police every fork shows one community deciding the metric itself had become counterproductive.
For developers focused on preservation, the absence of public rankings lowers the reward for outsiders racing to finish first with lower-quality output. The work of verifying functions, cleaning up generated code, and maintaining compatibility remains slow and manual. Removing the visible contest reduces the pressure to match automated speed.
For groups that prioritize quick ports over code quality, the change removes one source of visibility but leaves the underlying AI tooling untouched. Forks can continue on private repositories or other hosting platforms. The episode demonstrates how quickly automated assistance can strip value from public progress tracking when the tracked activity depends on human judgment that is hard to automate reliably.
The decision also highlights a practical limit of open metrics in technical communities. Once the cost of inflating those metrics drops, the original signal becomes noise. Decomp.dev responded by removing the signal instead of trying to restore it through additional rules. That response leaves original teams with fewer external incentives to publish incremental updates while giving automated efforts less of an audience to chase.
---
Sources:
No comments yet