Page 61
12 articlesThe AI Feature Deprecation Playbook: Shutting Down LLM Features Without Destroying User Trust
A practical guide for engineers and PMs on how to deprecate LLM-powered features cleanly — covering data lifecycle teardown, behavioral migration testing, user trust dynamics, and communication strategy.
What 'Done' Means for AI-Powered Features: Engineering the Perpetual Beta
AI-powered features never reach a stable 'done' state — model drift, world drift, and expectation drift create continuous iteration pressure. Here's the engineering and governance infrastructure that makes 'stable but evolving' feel like quality rather than incompleteness.
The AI-Generated Code Maintenance Trap: What Teams Discover Six Months Too Late
Teams adopting coding agents see dramatic velocity gains in months one through three. By month twelve, many find themselves unable to ship features without understanding their own systems. Here's the failure pattern — and how to avoid it.
AI Infrastructure Carbon Accounting: The Sustainability Cost Your Team Hasn't Measured Yet
AI inference now produces 2.5–3.7% of global emissions and is growing 15% annually. Here's how to measure your team's contribution and why it will become a compliance concern whether you plan for it or not.
Choosing a Vector Database for Production: What Benchmarks Won't Tell You
Benchmark leaderboards measure the wrong things. Here's the evaluation framework that actually predicts whether your vector database will hold up in production.
AI Oncall: What to Page On When Your System Thinks
How to design alerting for non-deterministic AI systems, what an AI incident looks like vs. a traditional failure, and runbook structures that actually help an on-call engineer at 2am.
When Everyone Has an AI Coding Agent: The Team Dynamics Nobody Warned You About
When every engineer on your team has an AI coding agent, individual productivity gains can quietly destroy collective code ownership, accelerate knowledge silos, and break code review culture — here's what to do about it.
The AI Product Metrics Trap: When Engagement Looks Like Value but Isn't
How teams measure session count and completion rate while missing what actually predicts value — and why the first 30 days of AI feature metrics are almost always wrong.
AI for SRE Log Analysis: The Tiered Architecture That Actually Works
Real-time frontier model analysis of streaming logs is financially and latency-untenable. Here's the tiered approach—fast anomaly detection gating selective LLM calls—that actually works in production.
AI Succession Planning: What Happens When the Team That Knows the Prompts Leaves
When the engineer who wrote your system prompt leaves, the reasoning behind every phrasing decision leaves with them. Here's how to build AI systems that survive personnel changes.
AI User Research: What Users Actually Need Before You Write the First Prompt
Most AI features fail not because the technology is wrong, but because teams asked users what they wanted instead of observing what they actually do. Here's how to run user research that produces reliable behavioral signal before you build.
The Alignment Tax: Measuring the Real Cost of Shipping Safe AI
Every safety layer you add to a production AI system has a measurable cost in latency, tokens, and user friction. Here's how to instrument that cost and make principled tradeoffs.
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.
