Highlights
Top Insights
Using AI like ChatGPT at the start of brainstorming can reduce originality and make people feel less confident in their own ideas.
When people think on their own first, and then use AI, they come up with more ideas and feel a stronger sense of ownership and control.
To get the best results from AI tools, teams should delay AI input until after some independent thinking—this boosts creativity and user satisfaction.
Source: Timing Matters: How Using LLMs at Different Timings Influences Writers’ Perceptions and Ideation Outcomes in AI-Assisted Ideation (CHI 2025)
Top News
1. Mistral AI has launched Mistral Medium 3, a enterprise-ready language model at a fraction of the cost of larger models.
2. Google’s new AI Max for Search campaigns is a one-click suite of AI-powered tools that enhances ad targeting, creative optimization, and campaign insights.
3. Figma unveiled Figma Make, Sites, Grid, Draw, and Buzz, designed to unify design and production, and empower creative expression.
4. Anthropic has launched a web search feature on its API.
5. FutureHouse, a nonprofit, has launched a suite of AI tools to aid scientific research.
Additional Insights
1. The Right Way to Launch an AI Initiative (Harvard Business Review)
AI projects often fail—about 80% of the time—because they are inherently probabilistic and more complex than traditional IT projects, making outcomes unpredictable. Many fail due to poor project selection, misalignment with business goals, or a lack of feasibility and ethical planning from the outset. Trust plays a central role in adoption; even effective AI tools may go unused if users don’t understand or believe in them. Success requires involving intended users early, continuously testing key hypotheses, and building infrastructure for rapid but responsible experimentation. Ultimately, AI initiatives must be evaluated not just for technical performance but for real-world impact and unintended consequences, requiring ongoing auditing and refinement.
2. Why the Humanoid Workforce Is Running Late (MIT Technology Review)
Despite investor enthusiasm and flashy demos, experts warn that humanoid robots are far from ready for widespread use in industry. Roboticists like Daniela Rus highlight that current humanoids lack common sense and physical reliability, making them impractical for many real-world tasks. Companies like Figure AI have attracted massive investment, but their claims about capabilities and partnerships—such as with BMW—have faced scrutiny and legal threats over alleged exaggeration. While advances in AI and robotics are real, deploying humanoids at scale will be slow, industry-specific, and face significant technical, regulatory, and safety hurdles. The gap between investor hype and engineering reality underscores the challenges in translating lab prototypes into truly useful workforce robots.
3. How to Stay Human-Centered in the Age of AI and Emerging Tech (IDEOU)
In the face of rapid technological advancement, Grace Hwang emphasizes the importance of staying rooted in human-centered design to ensure innovation serves real human needs. She advocates for using emerging tech like AI and AR as creative assistants rather than ends in themselves, stressing the importance of prototyping, feedback loops, and ethical responsibility—particularly around data privacy and user safety. Grace urges teams to embed safety measures from the outset and to design for inclusivity, noting that edge cases often predict future mainstream needs. She also reassures leaders that it’s never too late to bring in human-centered practices, even mid-development, and that small, nimble teams can yield transformative insights. Ultimately, keeping people at the core of design not only builds trust and better products but also delivers stronger business outcomes.
4. Meaningful Resistance (IDEO)
The article emphasizes the value of meaningful friction—the idea that intentionally adding challenges can enhance enjoyment, engagement, and personal growth. While much of today’s technology aims to remove friction for the sake of speed and convenience, doing so can strip life of its richness, especially as seen in the behaviors of Gen Z, who are actively pushing back against digital overload. Through examples like friction-based games, limited-use children’s apps, and AI tools that support rather than replace meaningful work, the author illustrates how carefully designed obstacles can encourage reflection, creativity, and human connection. The goal isn’t to reject efficiency entirely, but to apply it selectively, preserving effort where it creates joy and value. Ultimately, designing for friction is presented as a way to foster more authentic, rewarding experiences in a world increasingly driven by automation.
Innovation Radar
1. AI Model Releases and Advancements
Google has released Gemini 2.5 Pro Preview (I/O edition), featuring major upgrades in coding, especially for building interactive web apps, with enhanced UI, video understanding, and tool-calling reliability (Google).
Israeli startup Lightricks has launched LTX VIDEO-13B, an open-source, cost-effective, high-speed generative AI video model (CTech).
Mistral AI has launched Mistral Medium 3, a highly efficient, enterprise-ready language model that delivers near state-of-the-art performance—especially in coding and professional tasks—at a fraction of the cost of larger models, with broad deployment and customization capabilities (Mistral).
2. AI Tools and Features
A new AI-powered headphone system called Spatial Speech Translation can translate multiple speakers’ voices in real time by detecting their direction and vocal traits, enabling natural, multilingual group conversations (MIT Technology Review).
Pinterest has enhanced its visual search with AI-powered features—like generative visual language models, search refinement tools, and broader app integration—to help users better explore and describe visual content such as styles and aesthetics (TechCrunch).
Google’s new AI Max for Search campaigns is a one-click suite of AI-powered tools that enhances ad targeting, creative optimization, and campaign insights to expand reach, boost relevance, and drive more conversions at lower costs (Google).
At Config 2025, Figma unveiled a suite of new tools—including Figma Make, Sites, Grid, Draw, and Buzz—designed to unify design and production, empower creative expression, and harness AI to help teams build and ship high-quality products more efficiently from concept to launch (Figma).
Hugging Face has released a free, cloud-based AI agent called Open Computer Agent that can perform basic computer tasks via a virtual machine, showcasing the growing potential of open-source agentic AI despite current limitations like slowness, errors, and CAPTCHA issues (TechCrunch).
Anthropic has launched a web search feature on its API, enabling Claude-powered applications to access real-time online information for more accurate, up-to-date, and verifiable responses across various use cases (Anthropic).
3. AI for Science and Medicine
An experimental Google AI chatbot, AMIE, has outperformed human doctors in diagnosing rashes and interpreting medical images by integrating visual and clinical data, though it remains untested in real-world settings (Nature).
For the first time, researchers used generative AI to design synthetic DNA that precisely controls gene expression in healthy mammalian cells, opening new possibilities for targeted gene therapies (Phys Org).
A man with ALS using a Neuralink brain implant now communicates faster and more naturally with help from generative AI like Musk’s Grok, raising questions about identity and authorship in human-AI interaction (MIT Technology Review).
FutureHouse, a nonprofit backed by Eric Schmidt, has launched a suite of AI tools to aid scientific research, claiming they can accelerate discovery despite skepticism over AI’s current reliability and lack of proven breakthroughs (TechCrunch).
4. Other
Kawasaki has unveiled plans for Corleo, a robotic horse powered by electric motors and hydrogen fuel, designed to mimic equestrian movement with AI and sensor technology, joining a growing trend of rideable robotic animals despite high costs and limited near-term availability (The Economist).
Sam Altman’s biometric ID project, called World, has launched in six U.S. cities using iris-scanning devices called Orbs to verify human identity and distribute cryptocurrency, amid new partnerships with Visa and Match Group and ongoing privacy concerns (CNBC).
Amazon unveiled a new warehouse robot named Vulcan that can “feel” items using AI-powered touch sensors, aiming to assist rather than replace human workers by handling hard-to-reach tasks and reducing injury risk (CNBC).
Leave a Reply