- Ainsider
- Posts
- Weekly Ainsider Newsletter #65: Last week in AI
Weekly Ainsider Newsletter #65: Last week in AI
Very powerful new models, tools and especially - new kind of AI Agents: summary of last week in AI

Welcome to this week’s AI Insider newsletter. We’ve combed through the latest software, models and agents added Ainsider.tools - all to bring you a curated lineup of new AI agents, models, and development frameworks.
Table of Contents
See all below, and follow me at X: https://x.com/piotrmacai to be with AI, Agents and LLMs technology always up-to-date!
Last week in AI: New AI Agents, Models and Tools

Higgsfield Product-to-VideoNow you can drop any product/object into any environment/scene 🎥 And it'll be perfectly integrated with existing video & the actor can even interact with 👀 Best way to create production quality videos. |
|
Macaron
Macaron is the first truly personal AI agent, crafted to learn from your conversations and adapt its responses to mirror your style. It stores context across interactions, enabling richer, more meaningful dialogues that evolve over time.
Ultra-personalized conversational AI
Persistent memory for ongoing context
Customizable persona and tone
Cross-platform integration
Parlant 3.0
Parlant 3.0 is an open-source engine for modeling agentic behaviors in large language model agents. It provides a modular framework that orchestrates multiple sub-agents, enabling them to collaborate, delegate tasks, and make decisions autonomously.
Agent-of-agents orchestration
Modular behavior templates
Real-time decision-making workflows
Community-driven extensibility
UI-TARS-desktop
UI-TARS-desktop brings multimodal AI agents to your terminal, desktop, browser, and products with a unified interface. It combines GUI components, vision modules, and text-based agents to create versatile and interactive AI workflows across environments.
Multimodal input support (text, image, GUI)
Terminal, desktop, and browser integration
Custom GUI agent component library
Extensible vision-based modules
Blink.new
Blink.new is a powerful no-code builder that helps you create software and AI-driven applications with simple drag-and-drop tools. It accelerates development with prebuilt templates, collaborative features, and seamless one-click deployment options.
Drag-and-drop AI app builder
Prebuilt AI templates and connectors
Real-time collaboration features
One-click deployment
Open Lovable
Open Lovable is the community-driven, open-source version of the Lovable conversational AI platform originally developed by Firecrawl. It offers a flexible framework for building chatbots with a plugin-based structure and a developer-friendly API.
Self-hostable conversational AI framework
Plugin-based architecture
Rich domain-specific language support
Developer-friendly API
Yupp
Yupp is an AI model discovery platform that lets you explore, compare, and test over 500 machine learning models in a unified interface. Its recommendation engine and in-browser demos help you quickly identify the best model for your project.
Unified AI model catalog
Side-by-side model comparisons
Instant in-browser model demos
Personalized recommendations
Jan V1
Jan V1 is a 4-billion-parameter open-source language model optimized for web search and information retrieval tasks. It offers an open alternative to Perplexity Pro, with both local and cloud deployment options to ensure privacy and scalability.
4B-parameter open-source LLM
Optimized for search query tasks
Local and cloud deployment options
Privacy-preserving inference
MemU
MemU introduces an agentic memory layer for language model applications, allowing developers to implement contextual memory storage, retrieval, and manipulation in their AI workflows. This enhances long-term coherence and enables agents to reference past interactions seamlessly.
Contextual memory management
API for memory reads and writes
Fine-grained memory controls
Plugin integration for agents
Kuse AI
Kuse AI is an all-in-one AI canvas that transforms chat interactions with files, links, and videos into actionable insights and tasks. It streamlines collaborative analysis by automatically extracting key points, tagging content, and suggesting next steps.
Unified chat interface for multiple media
Automated insight extraction
Actionable task generation
Collaborative workspace
Graphiti
Graphiti is a specialized framework for constructing and querying temporally-aware knowledge graphs tailored to AI agents. By mapping evolving information over time, agents can perform decisions based on both historical trends and current data.
Temporal knowledge graph construction
Agent-consumable graph queries
Integration with popular LLMs
Developer SDK for customization
Airtop
Airtop provides AI-powered cloud browsers for automating web tasks such as scraping, form filling, and site control. Its secure, no-code interface lets you schedule automated workflows, manage credentials, and run parallel jobs with ease.
Cloud-based AI-driven browser control
No-code web scraping workflows
Scheduling and parallel execution
Secure credentials management
Agno UI
Agno UI offers a minimalist yet powerful chat interface for interacting with Agno Agents. It focuses on seamless streaming responses, customizable themes, and quick setup, making it easy to embed conversational agents into your products.
Minimalist conversation UI
Real-time message streaming
Customizable themes and layouts
Seamless Agno backend integration
/