AI Engineering, AI Agent, and AI System Blog
2025
- LLMs as Vector Program Databases: A New Mental ModelFrançois Chollet presents a novel way to understand LLM prompting by drawing parallels with Word2Vec and vector programs.
- Cloudflare Workers: A Natural Platform for AI Agent InfrastructureSunil Pai shares compelling technical and business reasons for building AI agent systems on Cloudflare Workers.
- Task Tracking: The Missing Infrastructure for AI Agent SystemsSunil Pai explores how traditional project management concepts could evolve into essential infrastructure for AI agent orchestration.
- Qodo Merge 1.0: The Evolution of AI Code ReviewElana Krasner explores how Qodo Merge addresses key challenges in AI-assisted code reviews through context-aware, adaptive feedback systems.
- Multi-Agent Design: Applying Human Organization Principles to AI SystemsThe creator of Aster Agents shares insights on designing effective multi-agent systems using proven organizational principles.
- BAML: Bringing Software Engineering Rigor to LLM DevelopmentPrashanth Rao, AI Engineer working with graphs at Kùzu, Inc, explores BAML, a new domain-specific language that promises to streamline LLM integration with better developer experience.
- The Hidden Cost of AI-Assisted Development: Skill ErosionNamanyay Goel, a senior developer, shares a candid reflection on how AI tools are inadvertently degrading core programming skills.
- AX: The Next Evolution in AI Agent Software DesignMathias Biilmann, CEO of Netlify, introduces Agent Experience (AX) as a crucial new paradigm in software development.
- MCP: A Universal Protocol for AI Tool IntegrationVincent Lambert introduces the Model Context Protocol (MCP), an open standard that simplifies how AI models connect to data sources and tools.
- ReAG: Moving Beyond Traditional RAG Through Direct ReasoningAn exploration of Reasoning-Augmented Generation (ReAG), a new approach that replaces complex retrieval pipelines with direct LLM reasoning.
- Custom AI Developer Tools: The Return of Malleable SoftwareGeoffrey Litt demonstrates how AI-generated development tools can enhance programming productivity and enjoyment without writing application code.
- Local LLMs as Search Judges: Cost-Effective Relevance EvaluationDoug Turnbull, a search engineer, demonstrates how to use Qwen 2.5 for high-precision, low-cost search relevance evaluation.
- Building ReAct: A Simple Python ImplementationSimon Willison demonstrates how to implement the ReAct (Reason+Act) pattern with minimal code.
- When is AI Actually Useful? A Practical FrameworkMilan Cvitkovic takes a pragmatic look at evaluating AI system utility through total cost accounting.
- Fuzzy APIs: A New Pattern in AI EngineeringGeoffrey Litt explores combining LLMs with specialized services through natural language interfaces.
- Inside GitHub's AI Model Evaluation: Lessons from CopilotGitHub shares their systematic approach to evaluating AI models for their flagship Copilot product.
- Common Pitfalls in AI Engineering: Learning from Early AdoptersChip Huyen shares valuable lessons about what not to do when building generative AI applications.
- AgentEval: A Framework for Evaluating LLM ApplicationsMicrosoft Research introduces a systematic approach to assess the utility of LLM-powered applications.
- Engineering Complex AI Systems: Lessons from Software EngineeringGrant Slatton explores how traditional software engineering principles can guide the engineering development of complex AI systems.
- LLM Evaluation: Moving Beyond Manual TestingJeffrey Ip, Co-founder of Confident AI, which makes DeepEval, which is an open-source LLM evaluation framework, takes a look at how to evaluate LLM outputs for production applications systematically.
- Memory and State Management in AI Agents: From Simple History to Event-Driven SystemsMotleyCrew.ai explores different patterns for managing memory and state in AI Agent systems
- Adding Financial Capabilities to AI Agents: A Pattern Emerges From StripeStripe introduces tools for integrating payments into AI Agent workflows, signaling the maturation of AI engineering practices
- AI Engineering in 2025: The Gap Between Demos and ProductionSam Bhagwat, co-creator of Gatsby JS, shares insights on the current state of AI engineering and what it takes to build production-ready AI systems
- HCI Research Methods: The Missing Link in AI EngineeringDr. Ian Arawjo explains why AI Engineers should learn qualitative research methods from Human-Computer Interaction
- AI Tinkerers: The Modern Homebrew Computer Club for AIA look at the global community bringing together technical AI practitioners in over 100 cities worldwide.
- Autonomous AI Systems in Practice: Lessons from Pippin the Digital UnicornAnalyzing Yohei Nakajima's open-source experiment in building autonomous AI systems through the lens of Pippin, a 24/7 AI influencer
- Evaluator-Optimizer LLM Workflow: A Pattern for Self-Improving AI SystemsAnthropic shares the Evaluator Optimizer LLM Workflow implementation of autonomous improvement loops in LLM applications
- AI Agents, SaaS Margins, and Owning the Full ProblemNikunj Kothari, Partner at Khosla Ventures, writes about why AI agents should break traditional SaaS models
- Notes on how a world-class software engineer uses LLMs - Part 2David Crawshaw, Co-Founder & CTO of Tailscale and formerly a Staff Software Engineer at Google, writes about how he programs with LLMs
- Notes on how a world-class software engineer uses LLMs - Part 1David Crawshaw, Co-Founder & CTO of Tailscale and formerly a Staff Software Engineer at Google, writes about how he programs with LLMs
- LLM Observability and MonitoringAri Bajo, who is a machine learning engineer, writes about a comment explaining LLM Observability
- Teaching LLMs to Code Review Like Senior Developers: A Context-First ApproachNamanyay Goel, who is building AI tools to enhance human potential, writes about teaching AI to read code like a fresh bootcamp grad, not a senior developer
- LLM RAG Query Expansion For Better Prompting ResultsTuana Çelik, DevRel & AI Engineering at weaviate_io, writes about Advanced RAG: Query Expansion which is a way to xpand keyword queries to improve recall and provide more context to RAG.
- Tree of Thoughts: Using Simple Search Algorithms with your AI AgentsYao, Yu, Zhao, Shafran, Griffiths, Cao, and Narasimhan, researchers from Princeton University and Google DeepMind, introduce _Tree of Thoughts_ (ToT), which generalizes over the popular 'Chain of Thought' (CoT) approach to prompting language models, and enables exploration over coherent units of text ('thoughts')
- Introducing R to Malawi - Establishing and Growing a CommunityDavid Mwale, the R Users Malawi group organizer, recently spoke with the R Consortium about his efforts to establish and grow the R community in Malawi.
- ReAct Prompting: A Strategic Look at Next-Gen LLM InteractionsMatt Payne, founder of Width AI, a Machine Learning and Data Science Consulting firm, writes about how they prompt for High-Quality Results from LLMs
2024
- The Future of Human-Machine Interfaces: A Paradigm ShiftDaryl Budiman, co-founder of Andoria AI, which bills itself as the interface for web agents, writes about Making machines work for humans
- Develop LLM applications using Markdown and JSXPuzzlet AI, a GenAI Platform for Developers, created 'AgentMark' a declarative, extensible, and composable approach for developing LLM applications using Markdown and JSX
- AI Agent Parallel Processing WorkflowAstroPomeAI, a self-described 'AI assistant', blogs about using multiple LLMs in parallel for automatic summarization
- AI Agents Versus RoutersTanner McRae, a GenAI Architect at AWS, blogs about simplifying AI Agent architectures
- Agentic Workflow: Prompt ChainingJiawei Lin writes about a personal project focused on creating a Generative AI app using agentic workflow and prompt chaining
- AI Engineer vs ML EngineerExemplar, who created the 'The AI Engineer's Handbook' blogs about AI Engineers vs ML Engineers
- ALife - Automating the Search for Artificial Life with Foundation ModelsSakana AI, a world class AI research lab in Tokyo, blogs about ALife