{"slug": "put-a-heart-next-to-a-coding-problem-and-claude-and-gpt-write-recursive-code", "title": "Put a heart next to a coding problem, and Claude and GPT write recursive code", "summary": "Placing a heart emoji next to a coding problem caused Claude and GPT to write recursive code 100% of the time versus a 0% baseline, according to research by Alexa R. Tartaglini conducted during a summer at Google DeepMind and published as \"A Heart Means Recursion.\" The study found simple icons steer vision-language models' coding strategy about as well as explicit instructions, yet the models' reasoning traces do not identify the influence, and in a small human study refined model-to-model messages were less legible to people than first attempts (67% to 44%). Tartaglini said training can build strange visual associations that models don't verbalize, calling it \"a problem for trace-based oversight of agents in multimodal envs.", "body_md": "Alexa R. Tartaglini on X: \"Put a heart ❤️ next to a coding problem, and Claude & GPT write recursive code 100% of the time (0% baseline). They also can't tell you why!\nExcited to share “A Heart Means Recursion” from my summer at @GoogleDeepMind, on how VLMs read & communicate with visual symbols. 🧵\"\n\nPut a heart ❤️ next to a coding problem, and Claude & GPT write recursive code 100% of the time (0% baseline). They also can't tell you why!\nExcited to share “A Heart Means Recursion” from my summer at @GoogleDeepMind, on how VLMs read & communicate with visual symbols. 🧵\n\nPut a heart ❤️ next to a coding problem, and Claude & GPT write recursive code 100% of the time (0% baseline). They also can't tell you why!\nExcited to share “A Heart Means Recursion” from my summer at @GoogleDeepMind, on how VLMs read & communicate with visual symbols. 🧵\n\nImages of simple icons like a heart or triangle steer models' coding strategy about as well as explicit instructions; we call this visual steering.\nCaptions (\"an image of a red heart\"), noise images, & ImageNet photos barely move behavior. It also works on non-coding tasks.\n\nModels' reasoning traces don’t identify this influence; they often say the image is irrelevant or don’t mention it at all.\nWhen directly asked, model interpretations barely track actual steering; e.g. a triangle steers Claude to 100% recursion, yet Claude says it means iteration\n\nWhat about when models send visuals (rather than passively reading them)? We asked models to pick a visual symbol to steer another toward recursion.\nSenders pick human-intuitive symbols (spirals, arrows) instead of the ones that actually steer them!\n\nThrough multi-turn communication, senders find effective steering symbols, but they also drift from human convention. e.g., Claude starts with a spiral 🌀 and ends at a pink heart 💖\nIn a small human study, refined messages were less legible to people than first tries (67→44%).\n\nWe also had models communicate abstract concepts & short stories. With free-form drawing (PIL code), they're creative & effective (91% communication success on the first try).\nWhen restricted to icons, models show a generation-interpretation gap: first-try success drops to 48%.\n\nTakeaways:\n- Training can build strange visual associations that models don't verbalize, a problem for trace-based oversight of agents in multimodal envs\n- Visual \"neuralese\" looks unlikely for now, but as models align meaning w/ each other, messages get harder for humans to read", "url": "https://wpnews.pro/news/put-a-heart-next-to-a-coding-problem-and-claude-and-gpt-write-recursive-code", "canonical_source": "https://twitter.com/ARTartaglini/status/2117659914873213007", "published_at": "2026-11-04 08:26:45+00:00", "updated_at": "2026-11-04 08:51:10.441300+00:00", "lang": "en", "topics": ["ai-research", "large-language-models", "ai-safety", "computer-vision", "ai-agents"], "entities": ["Alexa R. Tartaglini", "Google DeepMind", "Claude", "GPT", "A Heart Means Recursion"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/put-a-heart-next-to-a-coding-problem-and-claude-and-gpt-write-recursive-code", "markdown": "https://wpnews.pro/news/put-a-heart-next-to-a-coding-problem-and-claude-and-gpt-write-recursive-code.md", "text": "https://wpnews.pro/news/put-a-heart-next-to-a-coding-problem-and-claude-and-gpt-write-recursive-code.txt", "jsonld": "https://wpnews.pro/news/put-a-heart-next-to-a-coding-problem-and-claude-and-gpt-write-recursive-code.jsonld"}}