- category
- detector
- maturity
- emerging
- approach
- Self-hosted, CPU-runnable ensemble of 23 detection engines (8 neural classifiers, 7 statistical methods such as log-rank, GLTR, perplexity and Binoculars, 8 linguistic heuristics) whose scores are combined by a logistic model fitted on the project's own public SlopBench benchmark into a single calibrated 0-100 score.
- access
- open-source
- owns
- An MIT-licensed, self-hostable AI-text detector that exposes every engine's score alongside one calibrated verdict, described by its README as 'VirusTotal for AI-generated text'.
- distinctFrom
- Closest are the hosted single-vendor text classifiers (gptzero, pangram, copyleaks, originality-ai). SlopTotal is an open-source, self-hosted multi-engine ensemble that publishes per-engine scores and its benchmark, rather than a proprietary single model behind an API.
- notes
- MIT license and 59 stars confirmed via the GitHub API; the repo was created 2026-03-21 and last pushed 2026-10-10, so it predates the 2026-09-19 window by creation date and the in-window event is the Show HN launch (posts reported on 2026-09-29, 2026-10-05 and 2026-10-09). The README self-reports AUC 0.942 on its own SlopBench benchmark (1,626 human and 1,626 AI texts), 1.7% of human texts flagged Likely AI, and 15% of non-native English essays flagged; all author-reported on the project's own benchmark and not independently verified. Included as a gap-fill by creation date, not a new release.
