OpenAI Build Hours
Comprehensive series covering advanced AI development concepts including prompt caching, memory management, and integration patterns.
- Prompt Caching - Optimize API costs and latency
- Memory Management - Maintain context in long conversations
- Advanced Prompting - Complex multi-step tasks
- Integration Patterns - Production deployment best practices
Mario Zechner's Blog
Independent developer with deep expertise in ML, compilers, and coding agents. Essential reading for understanding agent architecture.
- Pi Coding Agent - Building focused coding agents
- MCP Alternatives - Tool integration approaches
- MCP vs CLI - Performance benchmarking
BAML Podcast: "AI That Works"
Weekly interactive sessions with @hellovai & @dexhorthy building real AI applications. Live code, Q&A, and production techniques.
- Claude Agent Skills - Deep dive + PII redaction
- Agentic Backpressure - Production patterns
- Bash vs MCP - Token-efficient tooling
- SSE Streaming - Real-time AI applications
OpenClaw Architecture
Production-ready agent gateway with advanced session management, streaming, and multi-agent routing capabilities.
- Streaming & Chunking - Real-time response streaming
- Session Pruning - Efficient memory management
- Session Management - Long-lived agent sessions
- Multi-agent Routing - Intelligent agent orchestration
Anthropic Engineering
Daily technical blog from Anthropic's engineering team. Essential reading for production AI systems and agent development.
- Agent Infrastructure - Scaling managed agents
- Claude Code - Auto mode and security patterns
- Tool Design - Writing effective tools for agents
- Evals & Testing - Infrastructure noise quantification
LangChain Blog
Deep insights into agent evaluation, harness engineering, and skills development from the LangChain team building Deep Agents.
- Evaluating Skills - Building skills for coding agents
- Deep Agents Evals - How to build effective evaluations
- Harness Engineering - Improving agents with Terminal Bench 2.0
- Agent Behavior - Measuring and improving reliability
Efficient Inference with SGLang
DeepLearning.AI course on LLM inference optimization. Learn caching optimizations, KV cache, and RadixAttention for faster text and image generation.
- Inference Fundamentals - How LLM inference works under the hood
- KV Cache Optimization - Reduce memory usage and latency
- RadixAttention - Advanced attention mechanism optimization
- Text & Image Generation - Multi-modal inference techniques
Meta-Harness
End-to-end optimization of model harnesses. Automatically optimizes code determining what to store, retrieve, and present to LLMs.
- TerminalBench-2 - Harness evolution and optimization
- Text Classification - Surpasses hand-designed systems
- Math Reasoning - Advanced problem-solving capabilities
- Agentic Coding - TerminalBench-2 evaluation results
NeMo Agent Toolkit
DeepLearning.AI course on making agents reliable with NVIDIA's NeMo Agent Toolkit. Turn proof-of-concept demos into production-ready systems.
- Agent Observability - Monitor and debug agent behavior
- Evaluation Framework - Comprehensive agent testing
- Deployment Tools - Production-ready agent deployment
- Reliability Patterns - From demo to production systems
Deep Agents GitHub
Agent harness built with LangChain and LangGraph. Equipped with planning tool, filesystem backend, and sub-agent spawning capabilities.
- Agent Harness - Core tool calling loop with built-in tools
- Planning Tool - Plan before task execution
- Filesystem Backend - Shell and filesystem access
- Sub-agent Delegation - Isolated task execution
Criticism of Coding Harnesses
A must-watch critique of current coding agent harnesses. Exposes fundamental limitations and blind spots in how we build and evaluate coding agents today.
- Harness Limitations - What current harnesses get wrong
- Evaluation Gaps - Blind spots in agent benchmarks
- Design Flaws - Architectural weaknesses exposed
- Better Approaches - Paths forward for improvement