Page 25
12 articlesThe GUI Agent That Clicked the Right Button on the Wrong Screen
Computer-use agents fail not because they ground the wrong element but because the screen moves between observation and action. A look at screen-state drift and how re-grounding fixes it.
Hiring for AI Roles That Have No Career Ladder Yet
The title 'eval engineer' didn't exist two years ago, so there's no leveling rubric and no resume that matches. Define the role around a real failure, screen for judgment over tools, source sideways from QA and platform, and write the ladder before you make the offer.
MCP Server Sprawl: The Unbounded Tool Surface Nobody Owns
Every team wires in another MCP server, and the agent's tool surface grows with no owner, no budget, and a token tax on every turn. How sprawl breaks selection accuracy and what curation discipline fixes it.
The Postmortem Where the Root Cause Was a Prompt Nobody Owned
A production incident traced back to a system prompt edited weeks earlier with no PR, no reviewer, and no owner. Why prompts keep escaping change management, and how to put them back under review without killing iteration speed.
Your Retry Logic Is Teaching the Agent the Wrong Lesson
Retry policies built for flaky networks assume the caller is correct. Agents make the caller the unreliable part, and a blind retry quietly launders real bugs into green checkmarks.
Your Synthetic Training Data Is Collapsing Toward the Mean
Synthetic data generated to fill dataset gaps quietly contracts toward the mean, erasing the rare cases you needed. Why per-example quality checks miss it, how to measure set-level diversity, and how to anchor generation to real data.
The Token Budget Is a Product Decision, Not a Config Value
The line that sets your context window split — history versus retrieval versus tool output — is a product decision hiding in an f-string. Here is how to surface it, measure it, and give it an owner.
Your Voice Agent Trusts Every Transcription Error as Fact
Speech recognizers return guesses, but the handoff to your language model relabels them as facts. How to preserve ASR uncertainty and design voice agents that confirm before they act.
When RAG Should Have Been a JOIN
Routing a relational question to a vector index turns exact answers into plausible ones. How to spot hard predicates, aggregations, and joins — and send them to a query, not cosine similarity.
The Agent Feedback Loop You Never Built
Every thumbs-down, every silent abandonment, every rephrased question is a free labeled failure case — and most teams discard it. How to build the pipeline from user signal to triaged failure to permanent eval case.
The Agent That Memorized Your Bug: Why a Fix Is a Memory-Invalidation Event
When you fix a bug, an agent that memorized the workaround keeps applying it against a world that no longer exists. Why a fix is an unscheduled memory-invalidation event, and how to treat agent memory like a cache.
The Agent That Narrated a Number It Should Have Computed
An LLM agent will narrate a plausible number instead of computing one, and fluent prose hides the missing tool call. How to force tool use and attach provenance to every figure.
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.
