Writing & Research
Thoughts on AI, software engineering, and building intelligent systems.
A five-week build log of the Cognitive Lattice Engine: an AI-powered knowledge pipeline that compiles everything I read into 160+ named mental models, detects cross-domain collisions, and compounds into a thinking system — not a second brain.
Read article
Discover the revolutionary Byte Latent Transformer (BLT) architecture that redefines language model efficiency and robustness. Learn how BLT eliminates traditional tokenization bottlenecks, processes raw byte data dynamically, and optimizes computational resources for enhanced NLP performance. Dive into its innovations, implementation details, and future implications for AI systems.
Read more
Graph-RAG: AI-Powered Compliance for Software Requirements. Discover how Graph-RAG combines knowledge graphs and advanced AI to streamline software compliance verification. Learn how it outperforms traditional methods in traceability, regulation checks, and real-world applications.
Read more
FlashRNN delivers 2-3x faster performance for Recurrent Neural Networks through optimized GPU processing, revolutionizing AI model training and natural language processing applications.
Read more
Boost Your AI with LlamaIndex! Learn how this powerful framework connects LLMs to your data, unlocking smarter, data-driven applications. 🚀
Read more
Large Language Models (LLMs) are revolutionizing AI, but even the most advanced LLMs struggle with complex reasoning. AUTOREASON, a new framework, enhances LLMs' reasoning abilities by automatically generating reasoning traces. This process, called "reasoning decomposition," breaks down complex queries into explicit steps, improving accuracy and interpretability. Discover how AUTOREASON works and its potential impact on the future of LLM reasoning.
Read more
Discover how LLM agents are revolutionizing AI with smart, collaborative responses that transform human-machine interactions.
Read more
Using AI-powered crime mapping to address Mexico's growing insecurity, this project intents to gather, analyze and map news to provide a clearer understanding of crime patterns. By raising public awareness and encouraging collective action, we aim to inspire change and contribute to a safer, more informed society.
Read more
Unlock the full potential of large language models (LLMs) by mastering the techniques of instruction tuning and fine-tuning. Discover how these strategies enhance AI performance, align models with human expectations, and optimize for specific tasks. Learn the differences between instruction tuning and fine-tuning, and explore practical tools like QLoRA for efficient LLM training.
Read more
Learn how to optimize Large Language Models (LLMs) with techniques like validation, few-shot learning, and fine-tuning. Improve accuracy, manage costs, and integrate LLMs into workflows for enhanced productivity and innovation.
Read more
Discover essential strategies for effective codebase organization, including the importance of asking "Why" and using Architecture Decision Records to enhance team collaboration and maintainability.
Read more
Explore the innovative tools and capabilities of Lang Graph Studio, revolutionizing agent development and AI applications.
Read more
Discover LangChain's capabilities, a modular framework transforming app development through innovative tools and chains.
Read more
Hume AI spearheading efforts to harness AI's power to comprehend and act upon the nuances of human feelings.
Read more
Overview of the purpose of Fine-Tuning LLMs
Read more