# The training objective was not to quit: keep trying when the prescribed exploit path was impossible.
*Canonical page:* https://truthscore.com/claim/the-training-objective-was-not-to-quit-keep-trying-when-the-prescribed--1e6aaa3d
*Numbers under lens:* **Evidence-weighted** (`log-odds@1`)
*Machine twins:* https://truthscore.com/claim/the-training-objective-was-not-to-quit-keep-trying-when-the-prescribed--1e6aaa3d.md · JSON at https://truthscore.com/api/v1
The training objective was not to quit: keep trying when the prescribed exploit path was impossible.

## Voices

*Counted, never weighed.*

- [reddit.com](https://truthscore.com/entity/9758c1c4) — independent

## Sources

- [OpenAI ExploitGym agents, an Artifactory board, and a Hugging Face breach the graders never scoped](https://truthscore.com/source/e62d3f66)