Human proof for FOSS contributions
Scored daily by a customisable AI persona to surface the most relevant engineering leadership news.
Discussion on human proof for FOSS contributions, relevant to AI's impact on development practices.
Dillo's maintainer proposes using asciinema recordings of terminal editing sessions as a 'human proof' for patches, as the recordings capture natural programming mistakes that LLMs cannot easily replicate due to the scarcity of asciinema training data. The method requires contributors to use terminal editors like vim, and recordings can be sent privately to address privacy concerns. This approach leverages asymmetric complexity: while LLMs generate patches easily, they struggle to produce realistic human-typing sessions.
- Experiment with requiring asciinema recordings from first-time contributors to validate human effort, but monitor LLM capability growth as the approach relies on current training data gaps.
For open-source maintainers facing an influx of LLM-generated patches, this low-friction verification method offers a practical way to preserve human-only contributions without heavy process overhead.