--- title: LLMs are stuck in a groupthink rut. This startup is... description: Most large language models exhibit a striking homogeneity in their outputs — ask any major chatbot for a random number and you'll almost always get 7. Research... canonical: https://daily.dev/posts/llms-are-stuck-in-a-groupthink-rut-this-startup-is-trying-to-get-them-out--2otelcdhy twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: LLMs are stuck in a groupthink rut. This startup is trying to get them out. | daily.dev og:description: Most large language models exhibit a striking homogeneity in their outputs — ask any major chatbot for a random number and you'll almost always get 7. Research... og:url: https://daily.dev/posts/llms-are-stuck-in-a-groupthink-rut-this-startup-is-trying-to-get-them-out--2otelcdhy og:image: https://api.daily.dev/og/posts/2oTeLcDHy.png og:image:alt: LLMs are stuck in a groupthink rut. This startup is trying to get them out. 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. # LLMs are stuck in a groupthink rut. This startup is trying to get them out. **[MIT Technology Review](https://daily.dev/sources/technologyreview)** · 9 min read · 1 upvotes · 0 comments ## Summary Most large language models exhibit a striking homogeneity in their outputs — ask any major chatbot for a random number and you'll almost always get 7. Research presented at NeurIPS 2025 confirmed this 'artificial hivemind' effect, showing that 25 different LLMs converged on nearly identical metaphors when given open-ended prompts. Australian startup Springboards built Flint, an LLM fine-tuned on Alibaba's Qwen 3, that targets specific decision points in its output to inject controlled randomness rather than blindly cranking up the temperature parameter. The approach aims to serve creative professionals in advertising and marketing who need genuinely novel brainstorming rather than statistically average responses. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### Why do ChatGPT, Claude, and Gemini all give similar answers to open-ended questions like naming a metaphor for time Mainstream LLMs converge on similar high-probability responses because they are trained in similar ways on similar data for similar tasks. A NeurIPS best-paper-winning study called 'Artificial Hivemind' found that when 25 different LLMs were asked 50 times each to write a metaphor about time, over half of 1,250 responses were versions of 'time is a river' and the rest were versions of 'time is a weaver'. _Developers evaluating LLM creativity and variety can track findings like this on daily.dev._ ### Does raising the temperature setting fix repetitive or predictable LLM outputs Not reliably, because temperature is a blunt instrument that raises randomness across an entire response rather than at the specific points where variety matters. Springboards found that maxing out temperature on an OpenAI model caused it to produce incoherent output, switching from English into code mid-sentence, so it trained its Flint model to selectively boost randomness only at key output points. _Anyone tuning LLM parameters for more creative output can follow practical findings like this on daily.dev._ ### What is Flint, the LLM built by the startup Springboards Flint is a language model built by Australian startup Springboards on top of Alibaba's open-source Qwen 3 model, specifically trained to produce more varied answers to open-ended questions than mainstream chatbots like ChatGPT, Claude, and Gemini. Rather than adjusting temperature globally, it identifies specific points in its output where more variety is useful and inserts more random word or phrase choices there. _Teams comparing LLM options for brainstorming and ideation can track new entrants like this on daily.dev._ ## Similar posts on daily.dev - [LLMs are a failure. A new AI winter is coming.](https://daily.dev/posts/llms-are-a-failure-a-new-ai-winter-is-coming--ehtpgyz3i) · Hacker News · 54 upvotes · 30 comments - [Researchers discover a shortcoming that makes LLMs less reliable](https://daily.dev/posts/researchers-discover-a-shortcoming-that-makes-llms-less-reliable-wiiiapmcn) · MIT News · 3 upvotes · 0 comments - [LLMs are bullshitters. But that doesn't mean they're not useful](https://daily.dev/posts/llms-are-bullshitters-but-that-doesn-t-mean-they-re-not-useful-fctkoubaq) · Lobsters · 1 upvotes · 0 comments - [Meet the new biologists treating LLMs like aliens](https://daily.dev/posts/meet-the-new-biologists-treating-llms-like-aliens-i8raaa1kt) · MIT Technology Review · 1 upvotes · 0 comments --- Tags: [#llm](https://daily.dev/tags/llm), [#qwen](https://daily.dev/tags/qwen) [View this post on daily.dev](https://daily.dev/posts/llms-are-stuck-in-a-groupthink-rut-this-startup-is-trying-to-get-them-out--2otelcdhy) ```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":"LLMs are stuck in a groupthink rut. This startup is trying to get them out.","url":"https://daily.dev/posts/llms-are-stuck-in-a-groupthink-rut-this-startup-is-trying-to-get-them-out--2otelcdhy","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/llms-are-stuck-in-a-groupthink-rut-this-startup-is-trying-to-get-them-out--2otelcdhy"},"datePublished":"2026-07-01T14:43:46.717Z","dateModified":"2026-09-14T07:59:41.436Z","description":"Most large language models exhibit a striking homogeneity in their outputs — ask any major chatbot for a random number and you'll almost always get 7. Research...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/eaa0d77d8d000451ba13ac2ce477f042?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/eaa0d77d8d000451ba13ac2ce477f042?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"MIT Technology Review","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":"MIT Technology Review","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/a8d03889722f4f22b3986e0b395ff076","url":"https://daily.dev/sources/technologyreview"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/llms-are-stuck-in-a-groupthink-rut-this-startup-is-trying-to-get-them-out--2otelcdhy","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"llm,qwen","timeRequired":"PT9M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"MIT Technology Review","item":"https://daily.dev/sources/technologyreview"},{"@type":"ListItem","position":3,"name":"LLMs are stuck in a groupthink rut. This startup is trying to get them out."}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/llms-are-stuck-in-a-groupthink-rut-this-startup-is-trying-to-get-them-out--2otelcdhy#faq","mainEntity":[{"@type":"Question","name":"Why do ChatGPT, Claude, and Gemini all give similar answers to open-ended questions like naming a metaphor for time","acceptedAnswer":{"@type":"Answer","text":"Mainstream LLMs converge on similar high-probability responses because they are trained in similar ways on similar data for similar tasks. A NeurIPS best-paper-winning study called 'Artificial Hivemind' found that when 25 different LLMs were asked 50 times each to write a metaphor about time, over half of 1,250 responses were versions of 'time is a river' and the rest were versions of 'time is a weaver'. Developers evaluating LLM creativity and variety can track findings like this on daily.dev."}},{"@type":"Question","name":"Does raising the temperature setting fix repetitive or predictable LLM outputs","acceptedAnswer":{"@type":"Answer","text":"Not reliably, because temperature is a blunt instrument that raises randomness across an entire response rather than at the specific points where variety matters. Springboards found that maxing out temperature on an OpenAI model caused it to produce incoherent output, switching from English into code mid-sentence, so it trained its Flint model to selectively boost randomness only at key output points. Anyone tuning LLM parameters for more creative output can follow practical findings like this on daily.dev."}},{"@type":"Question","name":"What is Flint, the LLM built by the startup Springboards","acceptedAnswer":{"@type":"Answer","text":"Flint is a language model built by Australian startup Springboards on top of Alibaba's open-source Qwen 3 model, specifically trained to produce more varied answers to open-ended questions than mainstream chatbots like ChatGPT, Claude, and Gemini. Rather than adjusting temperature globally, it identifies specific points in its output where more variety is useful and inserts more random word or phrase choices there. Teams comparing LLM options for brainstorming and ideation can track new entrants like this on daily.dev."}}]} ```