Sunsetting decomp.dev
B · Useful, limited First-hand account from the maintainer of the matching-decompilation community hub on why AI broke its community model; credible practitioner signal, anecdotal and without numbers.
Summary based on the full text·Original: about 5 minutes to read
In short
The maintainer of decomp.dev, a 2024 hub that tracked progress on matching decompilation of video games, is closing its main index and new project additions (October 2026). Recent models make decompilation much faster with human guidance, but hostile AI forks that "finish" others' projects, racing ports and a leaderboard turned into a numbers game have damaged the community. The site keeps existing project pages and history.
Key numbers
- 2024year decomp.dev launched to centralise matching-decompilation progress
- early 2026when, with Opus 4.6 and GPT-5.4, models appeared trained on the community's own decompilation tools
As reported by the source.
English·Web page·opens decomp.dev
What it reports
- From early 2026 (Opus 4.6, GPT-5.4), models appeared trained on matching decompilation and knew the community's tools, but needed heavy cleanup, cheated with code that matched machine code but was not plausible source, and cost too much per task.
- Newer models such as GPT-6 Astra follow complex directions, cross-reference symbols and debug data, write natural code and act as strong "permuters", testing many source variants across files to find a match, faster and cheaper.
- Harms named: forks that finish projects with AI without contributing, opportunists racing to ship half-working PC ports from others' work, and "decomp" being used to mean "PC port".
- The leaderboard that once motivated volunteers now rewards burning tokens to push percentages.
- The maintainer declines to ban AI-assisted projects: "slop" is hard to define and moderate for a sole hobbyist. The site code stays unlicensed and may not be reused.
How they got it
The maintainer's own observation of the projects and community over two years.
Limits
- No counts of hostile forks, ports or AI-assisted projects.
- One maintainer's view; other community members may differ.
- Saved copy is the homepage announcement only.
Why it matters for companies
Open communities run on effort that is visible and credited. When AI makes the last stretch of a project cheap, outsiders can capture the result without carrying the earlier work, and leaderboards start rewarding volume. Companies that rely on open-source projects, internal knowledge bases or contribution metrics should expect the same pressure: measure quality and provenance, not percentage complete.
Summary based on the full text. Read the original for the full argument and data.Read the original on decomp.dev