# Epoch AI newsletter reports AI job-automation tests and Slay the Spire benchmark results

> Original title: The Epoch Brief - October 23, 2026

- Source: Epoch AI · The Epoch Brief
- Published: 2026-10-23T19:29:17.000Z
- Bad Influence: https://badinfluence.news/items/qjcnw1zu5jlhz6h6rzrruymip
- Original: https://epochai.substack.com/p/the-epoch-brief-october-23-2026

## Summary

Epoch AI's newsletter reports that six leading models could not yet automate its own work in 11 tasks, with none meeting its quality standards. It also reports that frontier models now win over 90% of Slay the Spire runs at base difficulty, up from struggling to beat the game six months earlier. Top human players still do better, at an estimated 85% of runs at the original's hardest difficulty.

## Why it matters

The newsletter ties its benchmark results and benchmark audits together, which helps readers gauge how far AI capabilities and their measurements can be trusted.