Page 128
12 articlesLLM Provider Lock-in: The Portability Patterns That Actually Work
Most LLM lock-in advice stops at API wrappers, but the real lock-in hides in prompts, tool-calling assumptions, and behavioral quirks. Portability patterns that address what abstraction layers cannot.
The MCP Composability Trap: When 'Just Add Another Server' Becomes Dependency Hell
The MCP ecosystem hit 10,000+ servers and 30 CVEs in sixty days. How dependency sprawl, supply chain attacks, and tool conflicts turn composability into a liability — and the operational patterns that prevent it.
Open-Weight Models in Production: When Self-Hosting Actually Beats the API
A practical decision framework for self-hosting open-weight models like Llama, Mistral, and Qwen versus using frontier APIs — covering real cost breakdowns, compliance triggers, operational burdens, and the hybrid architecture most production teams actually need.
The Post-Framework Era: Build Agents with an API Client and a While Loop
Why 80% of production AI agents need nothing more than a prompt, a tool list, and a while loop — and how framework complexity becomes the bottleneck it promised to eliminate.
The 10x Prompt Engineer Myth: Why System Design Beats Prompt Wordsmithing
Production data shows the first 5 hours of prompt work yield 35% improvement while the next 40 hours add just 1%. The real leverage in LLM applications lies in retrieval quality, task decomposition, output validation, and evaluation infrastructure — not prompt wordsmithing.
The Agent Debugging Problem: Why Printf Doesn't Work When Your Code Thinks
Agent bugs don't throw exceptions — they return confident, wrong answers with a 200 status code. A practical guide to trace-based debugging, replay workflows, and the tooling gap holding back production AI agents.
The AI-Legible Codebase: Why Your Code's Machine Readability Now Matters
Codebase structure is the biggest lever on AI-assisted development velocity. Learn the refactoring patterns, file organization strategies, and context engineering techniques that help LLM-powered agents navigate and modify your code correctly on the first try.
The Alignment Tax: When Safety Tuning Hurts Your Production LLM
RLHF and safety alignment training can degrade LLM task performance by 15–17 F1 points and cause up to 91% false refusal rates on benign prompts. A measurement methodology and recovery patterns — from null-space optimization to structured output schemas — for reducing the alignment tax without compromising safety.
The Internal AI Tool Trap: Why Your Company's AI Chatbot Has 12% Weekly Active Users
Most internal AI chatbots die at 12% weekly active users because they're built as standalone destinations instead of workflow intersections. The integration patterns — IDE plugins, Slack bots at decision points, CLI tools — that actually drive adoption, and the metrics that separate vanity dashboards from real usage.
The Model Deprecation Cliff: What Happens When Your Provider Sunsets the Model Your Product Depends On
Forced model migrations expose hidden dependencies in production AI systems. A practical guide to regression harnesses, canary rollouts, and building systems where the model is a replaceable component.
Token Budget as Architecture Constraint: Designing Agents That Work Under Hard Ceilings
Fixed token budgets force fundamentally different agent designs than unlimited-budget prototypes. Learn budget allocation strategies, dynamic reallocation patterns, and constrained-first architectures that keep production agents reliable under hard ceilings.
The Tool Explosion Problem: Why Your Agent Breaks at 30 Tools
Agent tool selection accuracy drops from 96% to under 15% as tool counts grow. Three architectural patterns — Tool RAG, hierarchical routing, and the STRAP consolidation pattern — keep agents reliable past 30 tools.
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.
