Page 12
12 articlesThe Pointer Your Agent Mistook for a Value: Reference vs Value in Tool Outputs
Tool outputs that hand back IDs, paths, or URLs ask the agent to dereference them. The model's policy for when to resolve and when to compose-as-if-resolved is implicit, inconsistent, and silently wrong. Type the indirection.
The PR-Bot That Never Sleeps: When Your Reviewers Become the Rate Limiter
AI coding agents now open pull requests faster than humans can read them, turning reviewers into the rate limiter. Risk-tiered auto-merge, review budgets, and AI-on-AI triage are how teams keep throughput honest without rubber-stamping code into production.
The PR Description Your Coding Agent Cannot Write
Agents ship clean PRs with empty descriptions, and async review breaks because the rationale lived in the prompt the harness threw away.
The Prompt Log Is the Product Roadmap You Threw Away
Your AI feature's prompt log is the highest-resolution product discovery signal you have — and the one nobody on your product team is reading. Here's how to mine it for unmet demand.
The Redaction Layer Your Agent Cannot Reason Through
Privacy redaction can preserve classification accuracy while quietly destroying the entity continuity multi-step agents depend on. The fix lives in how placeholders are scoped, not whether they exist.
The Sandbox Your Agent Didn't Notice Was Real
A staging agent sent a real customer email because one tool in its registry held a production credential. Why sandbox is now a per-tool property, and the attestation pattern that catches credential-tier drift before it ships.
The Typo Your Agent Learned to Honor
Fine-tuning teaches a model to behave like your corpus — including the misspellings, hedges, and one rep's verbal tics. Here is how that inheritance happens and the curation pass that catches it.
The Verifier Loop That Couldn't Converge
Worker-critic agent loops promise convergence to quality but rarely deliver it — the verifier is a stochastic policy, the max-iterations cap is a budget gate dressed as a quality gate, and the patterns that restore termination treat the satisfaction surface as the real architectural problem.
When Safety Training Collapses the Operator Into the User
Safety-tuned LLM agents refuse legitimate operator requests because the model can't tell an on-call engineer from an anonymous user. The fix is architectural — signed runbooks, capability tokens, and operator-mode channels — not retuning refusal calibration.
The Agent That Scheduled Itself Into the Maintenance Window
Agents execute multi-step plans into deploy freezes, active incidents, and red status pages because they cannot read the side channels humans absorb for free. Here is how to fix it.
The Token Budget That Ran Out Mid-Conversation: Why Free-Tier Users Think Your Model Got Dumber
Per-user token budgets bite hardest mid-conversation, where silent truncation, dropped tool calls, and model fallbacks read as a quality regression — and the upgrade conversation never happens.
The Deflection Metric That Lied: When AI Support Success Hides User Churn
Deflection rate counts silence, not help. The same number can mean a resolved customer or a churned one — and the dashboard can't tell which until the cohort report arrives.
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.
