A randomized controlled study examining how AI assistance affects skill formation in software development, finding that while AI can improve productivity, it significantly impairs conceptual understanding, code reading, …
Letta agents achieve 74% accuracy on the LoCoMo memory benchmark using simple filesystem operations instead of specialized memory tools, suggesting that agent capabilities matter more than complex retrieval mechanisms …
A comprehensive walkthrough of modern large language model applications, covering everything from basic text interactions to advanced features like voice mode, image processing, code generation, and tool integration …
A comprehensive analysis of the AI investment landscape covering market opportunities, business model innovations, infrastructure buildout, and the evolution of private markets. Discusses how AI companies are achieving …
A comprehensive tutorial on implementing a GPT-style transformer from scratch, covering self-attention mechanisms, multi-head attention, and the complete architecture needed to build a language model capable of …
A comprehensive analysis of how AI represents the latest major product cycle in technology, examining three key investment themes: traditional software going AI-native, new categories where software replaces labor, and …
Dr. Wang discusses the transformation of AI from solving artificial problems in the 1980s to addressing real-world challenges today, while sharing insights on China’s competitive AI landscape and his experience …
A comprehensive guide to structuring large software projects using modular black-box architecture, demonstrating how to build scalable systems through examples of video editors, healthcare systems, and jet fighter …
A comprehensive analysis of the AI investment landscape covering market opportunities, business model innovations, infrastructure buildout, and the evolution of private markets. Discusses how AI companies are achieving …
How to Code with AI Without Losing Control AI coding assistants accelerate development dramatically. At Anthropic, 90% of Claude Code’s codebase comes from Claude itself. Yet using LLMs effectively requires …
