Macaron: The World's First Personal AI Agent

Macaron: The World's First Personal AI Agent 127

The world's first Personal AI Agent. Unlike other work-focused AIs, it relies on deep memory and customized mini-apps to help users live a better life.

#Ai工具箱 #Ai办公效率

Macaron: The World's First Personal AI AgentIntroduction

Macaron is the world's first personal intelligent agent. Unlike other ai tools that help people with work, it relies on deep memory and customized mini-apps to help users live better lives. At its core, Macaron can generate mini-apps based on user needs and communicate with people with strong empathy. User experience has found that Macaron not only organizes schedules, but also automatically syncs chat history into relevant mini-apps, forming a personalized profile.

The team behind it has a strong technical background, having previously created the popular MidReal, with deep research in deep learning and reinforcement learning.

Macaron: The World's First Personal AI Agent

Features

  • Deep Memory: Macaron can remember important details from users' lives, providing personalized assistance that grows alongside the user.

  • Instant Mini-App Creation: Users simply describe their needs, and it can create customized tools for travel, health, relationships, and hobbies.

  • Personalized Lifestyle Enhancement: Unlike productivity tools, it focuses on improving users' daily lives and personal well-being.

  • Real-Life Solutions: It provides practical tools such as fitness planners, meal organizers, and study guides, with no manual adjustment required.

  • Adaptive Learning: It continuously improves through interaction with users, ensuring that the solutions it provides better match user needs over time.

Use Cases

  • Travel Planning: Generate personalized itineraries based on destination, time, and preferences, with thoughtful reminders and adjustment suggestions.

  • Health Management: Create exercise plans, dietary arrangements, and schedule reminders to help users develop healthy lifestyle habits.

  • Relationship Maintenance: Record important anniversaries, birthdays, and events, sending timely reminders to help users maintain relationships with family and friends.

  • Hobby Tracking: Save reading lists, cooking notes, photography plans, exercise logs, and other interest-related content.

  • Life Inspiration & Companionship: Provide small tips, spark inspiration, and respond emotionally in daily life, making life feel warmer.

  • Student Assistant: Quickly generate course assistants and club finders for students, making academic life easier to manage.

  • For Pet Owners: Remember pets' names and related details, making daily interactions more personal.

  • Entertainment Recommendations: Help users find books or games that particularly match their tastes.

How to Use

  1. Make a Request: Users simply describe what tool or solution they need in their daily life, and Macaron will immediately create a tailored mini-app.

  2. Install the App: It currently supports the Apple iOS version, and users can download the Macaron App via the link on the official website.

  3. Interactive Learning: It learns from interactions with users, remembers important details, and over time can provide more personalized assistance.

Key Questions:

❓ Question 1: What is the core difference between Macaron and other AI assistants?

Answer: Other AI assistants mainly help users with work, while Macaron is the world's first Personal AI Agent, whose core mission is to help users live better. It grows alongside users through personal testing and deep memory, remembers important things like a real friend, and focuses on creating practical tools that users truly need in their lives.

❓ Question 2: What unique technical approaches does Macaron use to handle long inputs and long outputs?

Answer: When handling long inputs, unlike traditional Long-Context LLMs that input all content directly, it uses Agentic Memory via RL, which is constrained by the maximum length of a single-turn conversation, and processes things in chunks combined with memory step by step. When handling long outputs, it also uses Agentic Memory via RL, combining the question with continuously updated memory to gradually generate thinking and the final answer. In addition, the Multi-conv DAPO framework and other techniques are also used.

❓ Question 3: What specific positive feedback have users given about Macaron?

Answer: User feedback includes: it can create course assistants and club finders within five minutes; it creates a "Beginner's Cooking Journal" for users who want to learn cooking, helping them learn to make three dishes in two weeks; it remembers users' cats and shows care for them; it can change its chat style based on user feedback; and it shows concern in a gentle way when users are tired, among other things.


Tools related to Macaron: The World's First Personal AI Agent