Page 12
12 articlesThe Standup Is Lying: Coordinating Work When Agent Fleets Run Overnight
Daily standups and burndown charts assume a human doing one task at a time. Agent fleets running overnight break that unit of work, and the coordination artifacts quietly stop describing reality.
Approval Fatigue: How Human-in-the-Loop Gates Decay Into Rubber Stamps
Human-in-the-loop approval gates quietly decay into reflexive rubber stamps under volume. Why automation bias turns oversight into theater, the security hole it opens, and how to design gates that stay meaningful.
The Broken Apprenticeship: Rebuilding the Skill Ladder When AI Eats the Grunt Work
AI automated the grunt work that quietly trained junior engineers into seniors. Here is why the skill ladder is losing its bottom rungs — and how to rebuild the junior-to-senior path on purpose.
The Documentation Renaissance: Your README Is the Agent's Primary Context Surface
For two decades documentation was where good intentions went to die. Coding agents flipped that: your README is now executable context, and its quality directly drives task success.
The Carbon Line Item: Budgeting the Energy Cost of Inference
Two identical inference requests can differ in carbon footprint by 5x at the same dollar cost. Why spend is a bad proxy for emissions, and how to put energy on the dashboard next to latency.
Comprehension Debt: The 2 A.M. System No Human Understands
AI agents ship clean code no human ever modeled, and the bill comes due at 2 a.m. during an incident. Why comprehension debt is the defining operational risk of the agentic era, and how to keep human understanding in sync.
Your Internal Platform's New Primary Customer Is an Agent
AI agents are becoming the dominant consumer of internal APIs, and they fail in ways humans never do. How to design tooling for a customer that can't read your docs or file a ticket.
The Velocity Mirage: Why AI Teams Ship More PRs but Deliver Value Slower
AI coding tools push PR counts and commits up while end-to-end lead time stalls. Here's why activity and outcome came unbolted, and which metrics expose the slowdown.
The Two-Speed Roadmap: Planning AI Features When the Model Floor Moves Every Quarter
Frontier models now ship roughly every four weeks, deleting the workarounds you built last quarter. Split your roadmap by what compounds when models improve and what gets obsoleted, and plan each at its own speed.
You Didn't Choose a Model, You Married One: The Prompt-Level Lock-In Nobody Budgets For
Swapping your LLM endpoint is one line of config; making your prompts work on the new model is a quarter you didn't plan for. How prompt-level lock-in accumulates and how to measure your real migration cost before you're forced to pay it.
The ML Engineer You Hired Isn't the AI Engineer You Need
Staffing an AI product team out of your ML org is the most common org mistake of the year. The craft that trains models and the craft that ships reliable systems on top of someone else's model barely overlap — here's the role split nobody put in the ladder, and what to actually screen for.
Where Your Tokens Sleep at Night: Data Residency for LLM API Calls
Hosting LLM inference in an EU region is not the same as keeping your prompts out of foreign jurisdiction. A practitioner's guide to residency, the CLOUD Act, and fail-closed gateway architecture.
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.
