Page 147
12 articlesSynthetic Data Pipelines for Domain-Specific LLM Fine-Tuning
A practical guide to building synthetic data pipelines for domain-specific LLM fine-tuning — covering distillation vs. self-improvement, quality filtering, model collapse prevention, and budget-driven strategy selection.
Structured Generation: Making LLM Output Reliable in Production
Most LLM-powered apps have a silent bug waiting to surface: brittle JSON parsing. Structured generation — constrained decoding, JSON Schema enforcement, and the validation sandwich — is the infrastructure layer that prevents an entire class of production failures.
Six Context Engineering Techniques That Make Manus Work in Production
After rebuilding their agent framework four times and serving millions of tasks, the Manus team identified six concrete techniques for managing context windows in long-horizon AI agents — and why KV-cache hit rate is the most important metric most teams ignore.
The Action Space Problem: Why Giving Your AI Agent More Tools Makes It Worse
Adding more tools to an AI agent degrades its performance through attention dilution, selection noise, and context confusion. How hierarchical action spaces and the agent-as-tool pattern fix this.
Four Strategies for Engineering Agent Context That Actually Scales
Long-running agents degrade because context accumulates unchecked. Four strategies — write, select, compress, and isolate — keep agent context sharp across hundreds of steps.
The Anatomy of an Agent Harness
A breakdown of the infrastructure layer that makes AI agents reliable in production — the execution loop, context management, error handling, safety guardrails, and state persistence that separate prototypes from shipped systems.
Context Engineering: Memory, Compaction, and Tool Clearing for Production Agents
How to prevent context drift in production AI agents using compaction, tool-result clearing, and external memory — with token budget allocation strategies, failure modes, and measurement patterns.
CLAUDE.md and AGENTS.md: The Configuration Layer That Makes AI Coding Agents Actually Follow Your Rules
A practical guide to CLAUDE.md and AGENTS.md — the instruction files that give AI coding agents persistent project context, and why getting them right matters more than model choice.
Building AI Agents That Actually Work in Production
A production-focused guide to building AI agents: six composable patterns, a decision framework for single vs. multi-agent systems, tool design principles, the seven failure modes that cause incidents, and what real observability looks like for agent systems.
Effective Context Engineering for AI Agents
Active management of the LLM context window is the top engineering challenge for production AI agents. A breakdown of the four strategies — write, select, compress, isolate — that keep agents coherent across long tasks.
Mastering AI Agent Observability: Why Your Dashboards Are Lying to You
Standard monitoring dashboards show green while your AI agents silently hallucinate, skip tools, and degrade in quality. Here's what to actually measure — and why.
The 80% Problem: Why AI Coding Agents Stall and How to Break Through
AI coding agents generate code fast — but teams adopting them see 91% longer review times and 154% larger PRs. Here's what actually separates teams that ship quality from those drowning in AI-generated complexity.
About Tian Pan
I'm Tian Pan, an engineer-founder focused on agentic engineering — building autonomous AI systems and scaling engineering teams. I write practical guides on system design, technical leadership, and shipping with AI agents. Previously an early engineer at Uber, Brex, and IoTeX.
