--- title: Candy Crush, Doom, and a 1977 Atari: the unexpected... description: Three talks from Codemotion Milan 2026 explore the intersection of gaming and modern tech. Carolina Pinto (King/Candy Crush) argues that game design principles... canonical: https://daily.dev/posts/candy-crush-doom-and-a-1977-atari-the-unexpected-gaming-side-of-codemotion-milan-2026-zzu62zvhl twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Candy Crush, Doom, and a 1977 Atari: the unexpected gaming side of Codemotion Milan 2026 | daily.dev og:description: Three talks from Codemotion Milan 2026 explore the intersection of gaming and modern tech. Carolina Pinto (King/Candy Crush) argues that game design principles... og:url: https://daily.dev/posts/candy-crush-doom-and-a-1977-atari-the-unexpected-gaming-side-of-codemotion-milan-2026-zzu62zvhl og:image: https://api.daily.dev/og/posts/Zzu62ZvHL.png og:image:alt: Candy Crush, Doom, and a 1977 Atari: the unexpected gaming side of Codemotion Milan 2026 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. # Candy Crush, Doom, and a 1977 Atari: the unexpected gaming side of Codemotion Milan 2026 **[Codemotion](https://daily.dev/sources/codemotion)** · 5 min read · 0 upvotes · 0 comments ## Summary Three talks from Codemotion Milan 2026 explore the intersection of gaming and modern tech. Carolina Pinto (King/Candy Crush) argues that game design principles around onboarding, trust, and retention can teach AI product teams how to build better user experiences. Enzo Lombardi demonstrates connecting Claude to a 1977 Atari VCS via MCP and a Go-based emulator, reverse-engineering assembly code to build an AI opponent. Stefano Fiorucci (deepset/Haystack) challenges the 'bigger is better' AI narrative by training a 1-million-parameter ModernBERT model to play Doom on a CPU in milliseconds, using supervised fine-tuning and reinforcement learning. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How did someone connect Claude to an old Atari 2600 emulator? A Cloud Software Architect at Cleafy built a Model Context Protocol (MCP) server linking Claude to a Go-based emulator of the Atari VCS 2600, a 1977 console with just 128 bytes of RAM and no frame buffer. He manually reverse-engineered the assembly of an obscure 1980s game, then had Claude build an AI opponent capable of putting up a real fight. _Developers exploring MCP integrations with unconventional systems can follow similar experiments on daily.dev._ ### Can a small 1-million-parameter model play Doom effectively? Yes, a ModernBERT model with only 1 million parameters was trained to play Doom on a CPU in milliseconds, using supervised fine-tuning followed by reinforcement learning. The project, built by an engineer at deepset (makers of the Haystack framework), was designed to test when a small model can actually outperform a frontier LLM for a narrow task. _Anyone weighing small models against frontier LLMs for specific tasks can track these tradeoffs on daily.dev._ ## Similar posts on daily.dev - [If Your Tech Career Isn’t Moving, This Might Be the Nudge You Need](https://daily.dev/posts/if-your-tech-career-isn-t-moving-this-might-be-the-nudge-you-need-kb0tj4d2i) · Codemotion · 1 upvotes · 0 comments - [Codemotion Madrid 2026 recap](https://daily.dev/posts/codemotion-madrid-2026-recap-riwpqdvio) · Cassidy's blog · 3 upvotes · 0 comments - [5 Talks to Inspire \(and Level Up\) Your Tech Team](https://daily.dev/posts/5-talks-to-inspire-and-level-up-your-tech-team-kcrdeprb0) · Codemotion · 1 upvotes · 0 comments - [Former Microsoft dev trains AI to master Robotron: 2084](https://daily.dev/posts/former-microsoft-dev-trains-ai-to-master-robotron-2084-jubxpam6a) · The Register · 4 upvotes · 0 comments - [Exploring new frontiers of AI and games research](https://daily.dev/posts/exploring-new-frontiers-of-ai-and-games-research-ks90xxvar) · DeepMind · 2 upvotes · 0 comments --- Tags: [#llm](https://daily.dev/tags/llm), [#game-development](https://daily.dev/tags/game-development), [#mcp](https://daily.dev/tags/mcp), [#reinforcement-learning](https://daily.dev/tags/reinforcement-learning) [View this post on daily.dev](https://daily.dev/posts/candy-crush-doom-and-a-1977-atari-the-unexpected-gaming-side-of-codemotion-milan-2026-zzu62zvhl) ```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":"Candy Crush, Doom, and a 1977 Atari: the unexpected gaming side of Codemotion Milan 2026","url":"https://daily.dev/posts/candy-crush-doom-and-a-1977-atari-the-unexpected-gaming-side-of-codemotion-milan-2026-zzu62zvhl","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/candy-crush-doom-and-a-1977-atari-the-unexpected-gaming-side-of-codemotion-milan-2026-zzu62zvhl"},"datePublished":"2026-07-13T20:53:09.792Z","dateModified":"2026-09-14T08:19:20.806Z","description":"Three talks from Codemotion Milan 2026 explore the intersection of gaming and modern tech. Carolina Pinto (King/Candy Crush) argues that game design principles...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/c15abb45df283ed7507540116041f083?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/c15abb45df283ed7507540116041f083?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Codemotion","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":"Codemotion","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/008c4c0fe1ba4cd3b6a6740d442d242b","url":"https://daily.dev/sources/codemotion"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/candy-crush-doom-and-a-1977-atari-the-unexpected-gaming-side-of-codemotion-milan-2026-zzu62zvhl","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"llm,game-development,mcp,reinforcement-learning","timeRequired":"PT5M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Codemotion","item":"https://daily.dev/sources/codemotion"},{"@type":"ListItem","position":3,"name":"Candy Crush, Doom, and a 1977 Atari: the unexpected gaming side of Codemotion Milan 2026"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/candy-crush-doom-and-a-1977-atari-the-unexpected-gaming-side-of-codemotion-milan-2026-zzu62zvhl#faq","mainEntity":[{"@type":"Question","name":"How did someone connect Claude to an old Atari 2600 emulator?","acceptedAnswer":{"@type":"Answer","text":"A Cloud Software Architect at Cleafy built a Model Context Protocol (MCP) server linking Claude to a Go-based emulator of the Atari VCS 2600, a 1977 console with just 128 bytes of RAM and no frame buffer. He manually reverse-engineered the assembly of an obscure 1980s game, then had Claude build an AI opponent capable of putting up a real fight. Developers exploring MCP integrations with unconventional systems can follow similar experiments on daily.dev."}},{"@type":"Question","name":"Can a small 1-million-parameter model play Doom effectively?","acceptedAnswer":{"@type":"Answer","text":"Yes, a ModernBERT model with only 1 million parameters was trained to play Doom on a CPU in milliseconds, using supervised fine-tuning followed by reinforcement learning. The project, built by an engineer at deepset (makers of the Haystack framework), was designed to test when a small model can actually outperform a frontier LLM for a narrow task. Anyone weighing small models against frontier LLMs for specific tasks can track these tradeoffs on daily.dev."}}]} ```