Mallow, a Paris-based startup founded in September 2024 by Flore Cousin and Cédric O, has launched a screenless gaming speaker designed for children aged 4 to 10 years. The product, previously known as "The Marshmallow Project," has raised 11 million euros from investors including Balderton Capital, Daphni, and Motier Ventures. The company now employs more than 35 people and has established new offices. The speaker allows children to interact with audio games and stories through voice commands, using figurines that are placed on the device. Each figurine, equipped with an NFC chip, triggers different content. The speaker features four game formats: Adventures, where the child is the hero of a multi-episode story; Creatures, where the child teaches a character words; Mini-Games, which focus on reasoning and vocabulary; and Quizzes, which test knowledge on various subjects. The initial figurine, Zorp, provides access to a 1.5-hour adventure titled The Creation of the Mallow Squad, which includes one episode of each game format. Additional figurines cost 19.99 euros each, with more expected to be released in the coming months. Mallow has partnered with French publisher Nathan for a quiz on knights and with the Miraculous franchise for a mission involving Ladybug that teaches children to manage emotions. A third partnership is anticipated in the near future. The startup claims to have tested its product with over 300 children and their families for 1.5 years, resulting in 6,000 sessions and 900 hours of play. The product is currently available on the Mallow website and Amazon for nearly 120 euros and is manufactured in China, though the company hopes to bring some production back to Europe. The speaker uses a combination of generative AI technologies, including Mistral AI, with which Cédric O is familiar, having been a shareholder and participant in its launch. Mallow has implemented several safeguards to ensure the AI's responses are appropriate, including avoiding suggestions of dangerous activities and asking children to rephrase inappropriate statements. The AI also adapts to speech defects, such as stuttering, and can understand children who speak with a pacifier in their mouth. The product is set to expand internationally, with plans to enter the U.S. market by the end of next year. Currently, the speaker only supports French, with English support expected in 2026. A bilingual version is under development, which will be more complex due to the need for the AI to understand French-speaking children who may struggle with English pronunciation. The speaker requires a Wi-Fi connection to function fully, though an offline mode allows for listening to previously earned content. A headphone jack is preferred over USB-C by parents to avoid the need for an additional charging device.