--- title: Can Skills Learned in Games Transfer to Real-World Work? description: Good Start Labs, spun out of Every with $3.6 million in funding, trains AI models on games like Diplomacy and the railroad strategy game 1830 to teach... canonical: https://daily.dev/posts/can-skills-learned-in-games-transfer-to-real-world-work--hoz6ln9so twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Can Skills Learned in Games Transfer to Real-World Work? | daily.dev og:description: Good Start Labs, spun out of Every with $3.6 million in funding, trains AI models on games like Diplomacy and the railroad strategy game 1830 to teach... og:url: https://daily.dev/posts/can-skills-learned-in-games-transfer-to-real-world-work--hoz6ln9so og:image: https://api.daily.dev/og/posts/HOz6LN9so.png og:image:alt: Can Skills Learned in Games Transfer to Real-World Work? 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. # Can Skills Learned in Games Transfer to Real-World Work? **[Latent Space](https://daily.dev/sources/latentspace)** · 7 min read · 2 upvotes · 0 comments ## Summary Good Start Labs, spun out of Every with $3.6 million in funding, trains AI models on games like Diplomacy and the railroad strategy game 1830 to teach transferable skills such as strategic thinking and goal-directed execution. Co-founder Alex Duffy explains that training design matters enormously: fine-tuning a 30B model on 1830 using a multi-turn terminal-agent harness improved performance on a Finance-Agent benchmark, while a single-turn question-answering design did not, despite both improving in-game play. The company sells reinforcement learning data and learning environments to frontier AI labs, and is now working toward a unified general game intelligence model. Duffy argues evidence so far supports that goal-directed execution and reasoning transfer across domains, citing the 1830 finance result and earlier Diplomacy-to-customer-support transfer, though broader real-world transfer remains unproven. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### Does training an AI model on a board game actually improve its performance on unrelated real-world tasks? Yes, but only with the right training design. Good Start Labs fine-tuned a 30B model on the strategy game 1830, which has a stock-market mechanic, and tested it on financial research tasks. A single-turn question-answering training approach did not improve performance on the Finance-Agent benchmark, but a multi-turn terminal-agent design that used tools to explore, plan, and adapt did improve it. _Anyone evaluating training methods for agentic AI can follow ongoing coverage of these experiments on daily.dev._ ### Why did Claude Opus 4 lose consistently to OpenAI's o3 model when both played the game Diplomacy? OpenAI's o3 won all the games by planning future betrayals, a core strategic element of Diplomacy, while Claude Opus 4 refused to lie and consequently got destroyed in play. This divergence, observed in a 2025 Twitch stream of frontier models playing Diplomacy, revealed that different frontier models exhibit distinct personality traits like betrayal, collaboration, and theory of mind when facing game scenarios. _Developers comparing model behavior across vendors can track these personality-trait differences via daily.dev._ ### What does Good Start Labs actually sell to AI companies? Good Start Labs sells reinforcement learning data and learning environments, primarily to frontier AI labs. The data comes in two forms: trajectories of agents playing games (recording observations, decisions, actions, and outcomes) and custom data generated by agents playing live inside specific publishers' games. The learning environments are full games models can play end to end, anonymized and stripped of personally identifiable information before sale. _Teams evaluating RL data vendors for model training can keep up with this space through daily.dev._ ## Similar posts on daily.dev - [General Intuition’s $2.3B bet that video games can train AI agents for the real world](https://daily.dev/posts/general-intuition-s-2-3b-bet-that-video-games-can-train-ai-agents-for-the-real-world-lby2hikz0) · TechCrunch · 0 upvotes · 0 comments - [General Intuition’s $2.3B bet that video games can train AI agents for the real world](https://daily.dev/posts/general-intuition-s-2-3b-bet-that-video-games-can-train-ai-agents-for-the-real-world-swqgbgq2j) · TechCrunch · 0 upvotes · 0 comments --- Tags: [#machine-learning](https://daily.dev/tags/machine-learning), [#llm](https://daily.dev/tags/llm), [#ai-agents](https://daily.dev/tags/ai-agents), [#reinforcement-learning](https://daily.dev/tags/reinforcement-learning) [View this post on daily.dev](https://daily.dev/posts/can-skills-learned-in-games-transfer-to-real-world-work--hoz6ln9so) ```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":"Can Skills Learned in Games Transfer to Real-World Work?","url":"https://daily.dev/posts/can-skills-learned-in-games-transfer-to-real-world-work--hoz6ln9so","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/can-skills-learned-in-games-transfer-to-real-world-work--hoz6ln9so"},"datePublished":"2026-09-15T20:17:55.184Z","dateModified":"2026-09-15T20:18:19.870Z","description":"Good Start Labs, spun out of Every with $3.6 million in funding, trains AI models on games like Diplomacy and the railroad strategy game 1830 to teach...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/fa7eb47d2ca631144c18047375d99807?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/fa7eb47d2ca631144c18047375d99807?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Latent Space","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":"Latent Space","logo":"https://media.daily.dev/image/upload/s--DnJ9laFj--/f_auto,q_auto/v1780213196/logos/latentspace?_a=BAMAMiWQ0","url":"https://daily.dev/sources/latentspace"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/can-skills-learned-in-games-transfer-to-real-world-work--hoz6ln9so","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":2},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"machine-learning,llm,ai-agents,reinforcement-learning","timeRequired":"PT7M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Latent Space","item":"https://daily.dev/sources/latentspace"},{"@type":"ListItem","position":3,"name":"Can Skills Learned in Games Transfer to Real-World Work?"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/can-skills-learned-in-games-transfer-to-real-world-work--hoz6ln9so#faq","mainEntity":[{"@type":"Question","name":"Does training an AI model on a board game actually improve its performance on unrelated real-world tasks?","acceptedAnswer":{"@type":"Answer","text":"Yes, but only with the right training design. Good Start Labs fine-tuned a 30B model on the strategy game 1830, which has a stock-market mechanic, and tested it on financial research tasks. A single-turn question-answering training approach did not improve performance on the Finance-Agent benchmark, but a multi-turn terminal-agent design that used tools to explore, plan, and adapt did improve it. Anyone evaluating training methods for agentic AI can follow ongoing coverage of these experiments on daily.dev."}},{"@type":"Question","name":"Why did Claude Opus 4 lose consistently to OpenAI's o3 model when both played the game Diplomacy?","acceptedAnswer":{"@type":"Answer","text":"OpenAI's o3 won all the games by planning future betrayals, a core strategic element of Diplomacy, while Claude Opus 4 refused to lie and consequently got destroyed in play. This divergence, observed in a 2025 Twitch stream of frontier models playing Diplomacy, revealed that different frontier models exhibit distinct personality traits like betrayal, collaboration, and theory of mind when facing game scenarios. Developers comparing model behavior across vendors can track these personality-trait differences via daily.dev."}},{"@type":"Question","name":"What does Good Start Labs actually sell to AI companies?","acceptedAnswer":{"@type":"Answer","text":"Good Start Labs sells reinforcement learning data and learning environments, primarily to frontier AI labs. The data comes in two forms: trajectories of agents playing games (recording observations, decisions, actions, and outcomes) and custom data generated by agents playing live inside specific publishers' games. The learning environments are full games models can play end to end, anonymized and stripped of personally identifiable information before sale. Teams evaluating RL data vendors for model training can keep up with this space through daily.dev."}}]} ```