---
type: "post"
id: "epoch-will-release-1-000-gameplay-videos-and-a6ce53"
slug: "epoch-will-release-1-000-gameplay-videos-and-a6ce53"
title: "Epoch will release 1,000+ gameplay videos and extend the harness to more games."
description: "Thesis: A visual game benchmark measures an agent system, not just the model, because success depends on perception, action selection, and persistence across a long task."
retrieval_nugget: "Thesis: A visual game benchmark measures an agent system, not just the model, because success depends on perception, action selection, and persistence across a long task."
published_at: "2026-10-15"
updated_at: "2026-10-15"
record_date: "2026-10-15"
date_kind: "published_at"
topics: ["evals","agents","agent-harness"]
source_urls: ["https://epoch.ai/publications/ai-systems-are-improving-rapidly-at-slay-the-spire-and-its-sequel"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:epoch-will-release-1-000-gameplay-videos-and-a6ce53"
status: "published"
visuals: []
editorial_provenance: {"schema_version":"newruntime-editorial-copy-v1","content_status":"source_grounded_final","final_copy_sha256":"sha256:73d106b892af3ab9a20c5bc3cde8c39b63a9f861ca7369054ebab61f75af3351","reviewed_at":"2026-10-15T04:04:59.203Z","source_evidence_count":1,"verified_claim_count":3,"site_analysis_schema_version":"newruntime-site-analysis-v1","site_object_kind":"field_note","observed_fact_count":3,"implication_count":2,"watch_condition_count":1,"related_record_count":0}
analysis: {"schema_version":"newruntime-site-analysis-v1","object_kind":"field_note","thesis":"A visual game benchmark measures an agent system, not just the model, because success depends on perception, action selection, and persistence across a long task.","observed_facts":[{"text":"Epoch AI's SpireClimb evaluates models in Slay the Spire 1 and Slay the Spire 2 using a single-agent visual computer-use harness.","source_urls":["https://epoch.ai/publications/ai-systems-are-improving-rapidly-at-slay-the-spire-and-its-sequel"]},{"text":"The report says frontier systems exceed 90% win rates on the original game's base difficulty, while GPT-6 Astra and GPT-6.1 Sol reach 17% and 20% respectively on the original's maximum difficulty.","source_urls":["https://epoch.ai/publications/ai-systems-are-improving-rapidly-at-slay-the-spire-and-its-sequel"]},{"text":"Epoch says it plans to release more than 1,000 gameplay videos and extend the harness to more games.","source_urls":["https://epoch.ai/publications/ai-systems-are-improving-rapidly-at-slay-the-spire-and-its-sequel"]}],"mechanism":"The benchmark couples model reasoning to screen interpretation and computer-use actions over a sequence of turns. Performance therefore reflects the complete harnessed system and its ability to sustain decisions, rather than an isolated answer or game-knowledge test.","why_now":"Epoch's six-month comparison indicates that this task has shifted from near-inability to high success at the base setting, making evaluation of more difficult settings and system behavior newly consequential.","implications":["A visual game benchmark measures an agent system, not just the model, because success depends on perception, action selection, and persistence across a long task.","The next useful step is to compare traces and failure modes, not treat a single win rate as a general measure of autonomy."],"evidence_boundary":"The findings are specific to the tested games, selected models, difficulty settings, attempt protocol, and Epoch's harness; they do not establish general computer-use competence or explain the causal source of the gains.","watch_conditions":["Whether Epoch's planned gameplay videos expose repeatable failure modes and whether the harness extension produces comparable results on other games."],"related_records":[],"new_branch_reason":"No prior New Runtime record is supplied for this specific SpireClimb report; this opens a distinct record about visual gameplay benchmarking and its evidence trail."}
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/epoch-will-release-1-000-gameplay-videos-and-a6ce53/","markdown":"https://newruntime.com/posts/epoch-will-release-1-000-gameplay-videos-and-a6ce53.md","json":"https://newruntime.com/posts/epoch-will-release-1-000-gameplay-videos-and-a6ce53.json"}
related_patterns: ["ai-engineering-evaluations-are-splitting-into-task-specific"]
---

# Epoch will release 1,000+ gameplay videos and extend the harness to more games.

## Retrieval answer

Thesis: A visual game benchmark measures an agent system, not just the model, because success depends on perception, action selection, and persistence across a long task.

Thesis: A visual game benchmark measures an agent system, not just the model, because success depends on perception, action selection, and persistence across a long task.

Observed facts: Epoch AI's SpireClimb evaluates models in Slay the Spire 1 and Slay the Spire 2 using a single-agent visual computer-use harness. The report says frontier systems exceed 90% win rates on the original game's base difficulty, while GPT-6 Astra and GPT-6.1 Sol reach 17% and 20% respectively on the original's maximum difficulty. Epoch says it plans to release more than 1,000 gameplay videos and extend the harness to more games.

Mechanism: The benchmark couples model reasoning to screen interpretation and computer-use actions over a sequence of turns. Performance therefore reflects the complete harnessed system and its ability to sustain decisions, rather than an isolated answer or game-knowledge test. The planned videos could make runs easier to inspect, although aggregate scores alone do not identify which component caused a success or failure. Why now: Epoch's six-month comparison indicates that this task has shifted from near-inability to high success at the base setting, making evaluation of more difficult settings and system behavior newly consequential.

Implications: A visual game benchmark measures an agent system, not just the model, because success depends on perception, action selection, and persistence across a long task. The next useful step is to compare traces and failure modes, not treat a single win rate as a general measure of autonomy. Evidence boundary: The findings are specific to the tested games, selected models, difficulty settings, attempt protocol, and Epoch's harness; they do not establish general computer-use competence or explain the causal source of the gains. Watch condition: Whether Epoch's planned gameplay videos expose repeatable failure modes and whether the harness extension produces comparable results on other games.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Evals](https://newruntime.com/topics/evals.md): Explore the evals topic hub.
- topic hub: [Agents](https://newruntime.com/topics/agents.md): Explore the agents topic hub.
- topic hub: [Agent Harness](https://newruntime.com/topics/agent-harness.md): Explore the agent-harness topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)