The Download: a startup has a solution for AI’s groupthink problem

2026-07-02 · Source: MIT Technology Review · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation, Gaming & Interactive Media · Depth: Fundamental Awareness, long

Summary

Australian startup Springboards has developed an LLM named Flint to address the "groupthink" problem prevalent in mainstream large language models like Claude, ChatGPT, and Gemini. These models often exhibit predictable and less creative responses, such as consistently returning "7" when asked for a random number between 1 and 10. While this predictability is acceptable for tasks like coding, it hinders brainstorming or creative planning. Flint is specifically trained to generate a wider variety of responses to open-ended questions, for instance, "Where should I go in Europe?", aiming to overcome the rut of common LLM outputs and foster more diverse and imaginative interactions. This initiative seeks to enhance the utility of AI for tasks requiring broader ideation.

Key takeaway

For AI developers and product managers designing creative applications, you should prioritize models that demonstrate response diversity over mere predictability. If your current LLMs, like Claude or ChatGPT, consistently yield similar outputs for open-ended prompts, consider exploring specialized alternatives such as Springboards' Flint. This will help you avoid AI groupthink, ensuring your applications provide genuinely varied and imaginative results for tasks like brainstorming or content generation, thereby enhancing user engagement and utility.

Key insights

Mainstream LLMs suffer from groupthink, limiting creativity; new models aim for diverse, less predictable outputs.

Principles

• LLM predictability hinders creative applications.
• Diverse training can counter AI groupthink.
• Open-ended queries reveal LLM limitations.

Method

Springboards built Flint, an LLM specifically trained to produce a wider variety of responses to open-ended questions, moving beyond the predictable outputs of current models.

In practice

• Test chatbots with "random number" prompts.
• Use specialized LLMs for creative brainstorming.
• Evaluate AI for response diversity, not just accuracy.

Topics

• Large Language Models
• AI Groupthink
• Generative AI
• AI Creativity
• Springboards Flint
• AI Diversity

Best for: AI Engineer, Machine Learning Engineer, NLP Engineer, Tech Journalist, General Interest, Consultant

Related on AIssential

• 🔮 Is AI immune to groupthink? — LLM councils, like human committees, can suffer from groupthink, losing unique valuable ideas despite improving average answers.
• LLMs are stuck in a groupthink rut. This startup is trying to get them out. — Mainstream LLMs exhibit significant response homogeneity, which Springboards' Flint model addresses by selectively increasing output variety.
• Argument Collapse: LLMs Flatten Long-Form Public Debate — LLMs flatten public debate by converging on a narrow range of arguments and structures, even when prompted for diversity.
• Diversity Collapse in Multi-Agent LLM Systems: Structural Coupling and Collective Failure in Open-Ended Idea Generation — Multi-agent LLM systems for ideation suffer diversity collapse due to interaction structures, not model insufficiency.
• Nous: An Attempt to Extract and Inject the Cognition Behind Prediction-Market Behavior — LLM agent ensembles face cognitive monoculture; prompt-level injection of human behavioral profiles fails to induce diversity.
• Where Do Large Language Models Fail on Competitive Programming? A Taxonomy of Failures by Algorithm Type and Difficulty Rating — Chain-of-Thought prompting can degrade LLM code generation performance and instruction adherence in competitive programming.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by MIT Technology Review.