Four top AI models face off in Civilization VI: Claude nukes France and still loses, exposing perception and execution gaps
Original title:四大顶级 AI 对决《文明 VI》!Claude「核平」法国,结果还是输了
AI summary
Liam Wilkinson, a former data scientist at the UK Prime Minister's Office, built 76 MCP tools and put four models, including Claude Opus 4.6, GPT-5.4 and Gemini 3.1 Pro, into Civilization VI for 23 matches. Playing Portugal, Claude spent 50 turns developing nuclear weapons and nuked Toulouse as France closed in on a culture victory, but France ultimately won a diplomatic victory. Wilkinson found that the AI proactively checked global state only 1–2% of the time (a perception blind spot) and executed only 48–66% of plans within the next 10 turns (the knowing-doing gap). The conclusion: intelligence is not the bottleneck; perception and execution are.
Key Takeaways
- Liam Wilkinson,a former data scientist at the UK Prime Minister's Office,built 76 MCP tools and…
- 4 and Gemini 3.
- 1 Pro,into Civilization VI for 23 matches.
- Playing Portugal,Claude spent 50 turns developing nuclear weapons and nuked Toulouse as France…
- Wilkinson found that the AI proactively checked global state only 1–2% of the time (a perception…
Quick Facts
前首相府数据科学家让 Claude、GPT 等打《文明 VI》,揪出了 AI 的「感知盲区」和「知行差距」——更聪明的大脑解决不了睁不开眼、伸不出手的问题,做智能体的必须直面这两个工程瓶颈。
Related entities
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: Four top AI models face off in Civilization VI: Claude nukes France and still loses, exposing perception and execution gaps. June 28, 2026. https://new.nat6.net/en/news/cmqx7behn03g3slp0pzrofqz9/