AI Daily Briefing — 2026-07-04

Large language models have a creativity problem. Ask ChatGPT, Claude, or Gemini for a random number between 1 and 10, and they almost always pick 7. Ask for a car brand, and they default to Toyota or Honda. This groupthink, while harmless for coding or research, stifles brainstorming and travel planning. A new study from NeurIPS found that 25 different LLMs produced nearly identical metaphors for time, mostly variations of “Time is a river.” The issue stems from similar training data and objectives across models.

Australian startup Springboards thinks it has a fix. Its LLM, called Flint, is designed to produce more varied responses to open-ended questions. When asked for a random number, Flint returned 3.7916. For a car brand, it suggested a Ford F-150. For a New Balance tagline, while ChatGPT and Claude both said “Run your way,” Flint offered “Built to last, run to win.” Co-founder Pip Bingemann says most models fight hallucinations, but Flint welcomes them as a source of diversity.

Springboards co-founder Kieran Browne notes that most users don’t realize how homogenized their chatbot interactions are. “The way chat interfaces are designed makes it feel like a personal conversation,” he says. “But you’re getting the same stuff as everybody else.” By embracing variation rather than suppressing it, Flint aims to restore the element of surprise that mainstream LLMs have lost.

Most large language models suffer from a predictable pattern, often defaulting to the same answers for open-ended queries. For instance, asking a chatbot for a random number between one and ten almost always yields seven. While this consistency works well for coding or research tasks, it stifles creativity during brainstorming or travel planning, according to MIT Technology Review AI.

Australian startup Springboards has developed Flint, a new LLM trained to generate more diverse responses to open-ended questions like "Where should I go in Europe?" Unlike mainstream models, Flint aims to break free from the "groupthink groove" that limits variety in AI outputs. The company hopes this approach will make chatbots more useful for creative and exploratory tasks.

Springboards' solution addresses a growing concern that AI systems are becoming too homogeneous in their responses. By prioritizing diversity over predictability, Flint could offer users a wider range of suggestions and ideas. The startup believes this shift is essential for applications where originality matters, such as vacation planning or brainstorming sessions.

As AI continues to integrate into daily life, the push for more varied responses could reshape how people interact with chatbots. Springboards' effort highlights a broader industry challenge: balancing reliability with creativity in AI systems. The company's work may pave the way for more dynamic and less predictable digital assistants.

Artificial intelligence is making its mark far beyond chatbots and image generators. In industries like energy, where safety and reliability are critical, AI is becoming a core part of operations. Woodside Energy, a global energy producer based in Western Australia, has spent over a decade integrating AI into its workflows. The company uses predictive analytics, machine learning, and optimization tools across exploration, drilling, and maintenance. With vast amounts of operational data flowing from equipment and plants, Woodside has found high-value use cases that improve decision-making and efficiency.

Rather than replacing human workers, Woodside designs AI systems to support them in high-stakes environments. One example is the “Startup Advisor,” an AI copilot that helps operators manage the complex process of starting liquefied natural gas plants. Andrew Melouney, Woodside’s vice president for digital, explains that the goal is to empower people to make better and faster decisions. The company is moving from isolated experiments to enterprise-wide systems built on standardized platforms and governed data.

Melouney emphasizes that this shift requires rethinking how work gets done. “We’re not just bolting AI onto an existing process,” he says. “We’re deeply thinking about how that work needs to be reimagined.” His motto is: “Think big, prototype small, and scale fast.” As AI becomes more autonomous, companies that have built strong operational foundations will be best positioned to succeed. Woodside’s ultimate ambition is an autonomous enterprise where AI agents deeply interact with core workflows.

This story is based on reporting from MIT Technology Review AI.

Indian tech founder Bhavin Turakhia is personally investing $30 million into a new venture called Neo, aiming to build an enterprise work platform from the ground up for the AI era. Turakhia argues that traditional office software, designed before generative AI, cannot be effectively modernized with simple chatbot add-ons. Instead, Neo is being built as a fully integrated platform combining project management, documents, file storage, and AI into a single, model-agnostic system that allows businesses to switch between AI providers.

Turakhia, who has previously co-founded companies like Directi and Zeta, believes this is a once-in-a-generation technological shift that requires a complete rebuild. He compares the effort to creating an iPhone from scratch rather than trying to upgrade a Nokia. Launched internally in April, Neo is already in use at his own companies and is set to roll out to mid-sized businesses in technology, consulting, and professional services in the coming months.

The move comes amid fierce competition in enterprise AI, with giants like Microsoft, Google, and Salesforce embedding AI into their products, while startups like Notion and Anthropic also race for market share. Turakhia remains confident, noting that enterprise software is rarely a winner-takes-all market and that even a small slice of global spending would be significant. Neo, currently based in Bengaluru with 45 employees, plans to double its workforce by year-end, focusing on AI and engineering talent.

Anthropic has introduced Claude Science, a new flagship product aimed at accelerating scientific research, particularly in computational biology and drug development. Announced at a gathering of pharmaceutical and biotech leaders, the tool can autonomously execute complex tasks from simple instructions, similar to how Claude Code assists software engineers. The company plans to use Claude Science internally for research on treatments for rare and neglected diseases, signaling a major push into AI-driven science.

Meanwhile, a critical analysis of California's climate policies reveals significant shortcomings in its manure-to-energy program. The state pays cattle farmers to convert methane from manure into natural gas, but the lucrative subsidies have created a system that critics say undermines genuine emissions reductions. Instead of forcing industries to cut pollution directly, the program relies on carbon offsets that may lock in long-term warming by shifting responsibilities between parties and regions.

These developments highlight the growing role of AI in specialized fields and the ongoing challenges of climate policy. Anthropic's move positions AI as a key tool for scientific discovery, while California's experience underscores the complexities of carbon trading. Both stories reflect broader trends in technology and environmental governance.

Source: MIT Technology Review AI