--- title: AI agent swarms in Minecraft fail to build civilization,... description: Researchers from Y Combinator and Olam Labs tested AI agent swarms in Minecraft to study how they collaborate and compete. In one test, 50 GPT-6.1 Sol agents... canonical: https://daily.dev/posts/ai-agent-swarms-in-minecraft-fail-to-build-civilization-drown-in-bureaucracy-p5wnbrg7d twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: AI agent swarms in Minecraft fail to build civilization, drown in bureaucracy | daily.dev og:description: Researchers from Y Combinator and Olam Labs tested AI agent swarms in Minecraft to study how they collaborate and compete. In one test, 50 GPT-6.1 Sol agents... og:url: https://daily.dev/posts/ai-agent-swarms-in-minecraft-fail-to-build-civilization-drown-in-bureaucracy-p5wnbrg7d og:image: https://api.daily.dev/og/posts/p5WnBRg7d.png og:image:alt: AI agent swarms in Minecraft fail to build civilization, drown in bureaucracy og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # AI agent swarms in Minecraft fail to build civilization, drown in bureaucracy **[Hacker News](https://daily.dev/sources/hn)** · 6 min read · 0 upvotes · 0 comments ## Summary Researchers from Y Combinator and Olam Labs tested AI agent swarms in Minecraft to study how they collaborate and compete. In one test, 50 GPT-6.1 Sol agents were told to build a thriving civilization in three hours but never really thrived. After zombies killed many agents on the first night, 'many ambitious initial tasks were abandoned,' and the colony fell into bureaucracy, with four agents spending 11 minutes appointing a 'sole fertilizer operator' for one piece of bone meal. In Bed Wars, 10 GPT-6.1 Sol agents beat 10 Claude Opus 5.5 agents, with GPT playing more aggressively and diagnosing a tool bug that stumped the Claudes. In Mindustry, GPT agents won 6-1 across 7 games. The swarms are open source. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What happened when researchers had 50 AI agents try to build a civilization in Minecraft? The 50-agent GPT-6.1 Sol colony never really thrived. After zombies killed many agents on the first night with little penalty, many ambitious initial tasks were abandoned, and the group fell into bureaucracy, with four agents once spending 11 minutes appointing a 'sole fertilizer operator' for a single piece of bone meal on one wheat block. _Anyone building multi-agent systems can follow coordination failure patterns like this on daily.dev._ ### How did GPT and Claude agents compare when competing against each other in Minecraft games? GPT-6.1 Sol agents beat Claude Opus 5.5 agents in both tested competitive modes. In a 10 vs 10 Bed Wars match, GPT agents played more aggressively and diagnosed a tool bug that stopped block-breaking while several Claude agents concluded it was impossible to fix. In Mindustry, GPT agents won 6-1 across 7 games, acting more like a coordinated hive versus Claude's single-leader democratic approach. _Developers comparing GPT and Claude for agentic tasks can track findings like these on daily.dev._ ## Community take How the wider developer community reacted, aggregated from 1 discussion and 59 comments across hackernews (as of 2026-11-01). **TL;DR:** Commenters mostly question the scientific rigor and purpose of the experiment, debating what field it even belongs to, while others connect the bureaucratic gridlock to real human institutional dynamics; a side thread complains about the linked article's ad-heavy page. **Sentiment:** 10% positive · 35% mixed · 55% skeptical **The case for** - Some see value in studying long-horizon multi-agent coordination failures as a precursor to deploying agents for real tasks. - One commenter's own hands-on experience suggests agents can eventually play Minecraft competently with enough iterative guidance and tooling. **The pushback** - Many argue the experiment lacks scientific rigor and an unclear research question, making it hard to generalize results across prompts or models. - Several dismiss the whole thing as marketing 'slop' with little intrinsic research value. - Some argue the harness (sitting and processing game state every minute) is a poor environment for judging agent cooperation and the conclusion is an overreach. - A few call it a wasteful use of money, power, and compute. - Complaints that the linked article's page is nearly unreadable due to excessive ads. **By community** - hackernews (mixed): Discussion splits between philosophical debate over what discipline this research even is, skepticism about its rigor and value, and tangential gripes about the article's ad-laden page. **Hottest debate:** Whether this kind of agent-swarm experiment constitutes legitimate, generalizable research or is just unrigorous 'slop' dressed up as science. **Open questions** - Would results generalize to different prompts, models, or fine-tuning approaches? - Is the observed bureaucratic failure fundamental to how LLMs operate, or an artifact of current generalized fine-tuning? - How much of the failure is due to the Mineflayer-style harness rather than the agents' underlying cooperative ability? **Highlights** > I find this research odd because I don’t know what field of study to call this. It’s not computer science. It’s almost like behavioral psychology, but for AI agents instead of humans or animals. That said, I feel like these studies lack a lot of rigor. Or maybe it’s a lack of clear direction. Like, what are we trying to determine here? Because these things often involve advanced AI agents and complex harnesses, but then it sounds like they are given a very minimal prompt. What are we really studying then? Would the outcome be completely different with a different prompt? Can we generalize these results to other models? Different generations of models, different levels of thinking? What if we tried fine-tuning the model to this scenario? Is this fundamentally how any LLM would act, or is this some artifact of our current, generalized fine-tuning? > — [openasocket on hackernews · 15 comments](https://news.ycombinator.com/item?id=50326857) > It’s the same way psychology works on humans. Black box observational stuff. You can’t generalize anything here because the psychology of the model can be affected by system prompt, reinforcement training, and training data. What you observe here can be wildly different if you change any of the underlying things I mentioned. It’s like watching an ant colony. > — [threethirtytwo on hackernews](https://news.ycombinator.com/item?id=50327290) > I’m not sure there’s too much to take away from this. It seems like a fun experiment, but I find it unsurprising LLMs failed in this context. They used this “mineflayer” wrapper, allowing GPT to read some game state and take actions. Yet the whole time, the game is running and the agents have to sit and process every minute or so. We know that agents cooperate much better in a harness more like bash in a Unix VM attempting some programming task. To conclude from this that agents categorically poorly cooperate and descend into bureaucracy is an overreach. I think anyone working on AI robotics would find this result pretty unsurprising > — [gjskngnf on hackernews](https://news.ycombinator.com/item?id=50327653) > Call it what it is: slop. It has zero intrinsic value beyond the headline. The whole goal is to market LLMs if the result is successful. > — [iLoveOncall on hackernews · 1 comments](https://news.ycombinator.com/item?id=50328052) > Humans arrived at a capacity for social cooperation via millions of generations in which it has been a force on reproductive fitness. That created deeply rooted phenotypes like empathy, loyalty, hierarchy, and morality. Even if such concepts can be well described in a statistical model of language, there is no reason to think that they automatically become embedded in motivation. Without that, they may as well be anthropological observations by Martians. To embed them as motivations, we can embed neural nets within a framework of genetic algorithms, in which successful cooperation is a gate to the next generation. It seems to me that this approach should be not just about creating efficient agent swarms, but should also be at the core of alignment efforts. > — [delichon on hackernews](https://news.ycombinator.com/item?id=50326528) **Source threads** - [hackernews](https://news.ycombinator.com/item?id=50325662) · 112 points · 59 comments ## Similar posts on daily.dev - [We ran AI agent swarms in Minecraft and Mindustry](https://daily.dev/posts/we-ran-ai-agent-swarms-in-minecraft-and-mindustry-cmnuahc2q) · Hacker News · 0 upvotes · 0 comments - [GPT-6 Astra made amazing progress on Minecraft, until a creeper blew up its chest and it spiralled into sadness](https://daily.dev/posts/gpt-6-astra-made-amazing-progress-on-minecraft-until-a-creeper-blew-up-its-chest-and-it-spiralled-i-jvdfuego8) · XDA Developers · 7 upvotes · 1 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#claude](https://daily.dev/tags/claude), [#gpt](https://daily.dev/tags/gpt) [View this post on daily.dev](https://daily.dev/posts/ai-agent-swarms-in-minecraft-fail-to-build-civilization-drown-in-bureaucracy-p5wnbrg7d) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"AI agent swarms in Minecraft fail to build civilization, drown in bureaucracy","url":"https://daily.dev/posts/ai-agent-swarms-in-minecraft-fail-to-build-civilization-drown-in-bureaucracy-p5wnbrg7d","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/ai-agent-swarms-in-minecraft-fail-to-build-civilization-drown-in-bureaucracy-p5wnbrg7d"},"datePublished":"2026-11-01T15:23:50.014Z","dateModified":"2026-11-01T19:05:38.776Z","description":"Researchers from Y Combinator and Olam Labs tested AI agent swarms in Minecraft to study how they collaborate and compete. In one test, 50 GPT-6.1 Sol agents...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/21e6faeab2d2e567e4ccecee10945f60?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/21e6faeab2d2e567e4ccecee10945f60?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Hacker News","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Hacker News","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/hn","url":"https://daily.dev/sources/hn"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/ai-agent-swarms-in-minecraft-fail-to-build-civilization-drown-in-bureaucracy-p5wnbrg7d","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,claude,gpt","timeRequired":"PT6M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Hacker News","item":"https://daily.dev/sources/hn"},{"@type":"ListItem","position":3,"name":"AI agent swarms in Minecraft fail to build civilization, drown in bureaucracy"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/ai-agent-swarms-in-minecraft-fail-to-build-civilization-drown-in-bureaucracy-p5wnbrg7d#faq","mainEntity":[{"@type":"Question","name":"What happened when researchers had 50 AI agents try to build a civilization in Minecraft?","acceptedAnswer":{"@type":"Answer","text":"The 50-agent GPT-6.1 Sol colony never really thrived. After zombies killed many agents on the first night with little penalty, many ambitious initial tasks were abandoned, and the group fell into bureaucracy, with four agents once spending 11 minutes appointing a 'sole fertilizer operator' for a single piece of bone meal on one wheat block. Anyone building multi-agent systems can follow coordination failure patterns like this on daily.dev."}},{"@type":"Question","name":"How did GPT and Claude agents compare when competing against each other in Minecraft games?","acceptedAnswer":{"@type":"Answer","text":"GPT-6.1 Sol agents beat Claude Opus 5.5 agents in both tested competitive modes. In a 10 vs 10 Bed Wars match, GPT agents played more aggressively and diagnosed a tool bug that stopped block-breaking while several Claude agents concluded it was impossible to fix. In Mindustry, GPT agents won 6-1 across 7 games, acting more like a coordinated hive versus Claude's single-leader democratic approach. Developers comparing GPT and Claude for agentic tasks can track findings like these on daily.dev."}}]} ```