Page 58
12 articlesThe First 100 Tickets After You Launch an AI Feature
The eight-week sequence of operational tickets every AI feature launch produces — cost spikes, eval drift, latency tails, silent provider updates — and the launch playbook that pre-stages the answers.
The Human Review Queue Is Your P0 SLA: When HITL Becomes the Bottleneck
The human-in-the-loop escalation path you wired up for safety three months ago is now the silent bottleneck of your AI feature. Here is how to treat it as a production system with its own SLOs, capacity model, and feedback loop — before customers tell you first.
The IDE Plugin Is the Product Now: When Your Coding Agent Outgrows the Editor's Plugin API
Why the leading AI coding tools forked the editor instead of staying as plugins, and how to decide between extending VS Code, forking it, or building from scratch.
The LLM-as-Validator Antipattern: Why Your AI Quality Gate Has a Blind Spot
Using an LLM to evaluate LLM outputs as your primary quality gate creates a circular validation loop blind to systematic model failures. Here's what to use instead.
The Model Provider Webhook Surface You Forgot to Subscribe To
Major LLM providers broadcast incidents, deprecations, and account warnings through webhooks and emails — and most teams have those channels disabled. A look at the small integration that turns 'we found out from a customer' into 'we found out from the provider before our own monitors did.'
The Monday Morning AI Degradation Your Dashboard Treats As Noise
Most AI feature dashboards average away the one failure mode that costs real money — a weekly cyclic degradation that shows up only when you slice latency, cache hit rate, and retry counts by hour-of-day.
MTBF Is Dead When Your Agent Self-Heals
Self-healing agent runtimes have eaten the failure signal MTBF was built to count. Here is the metric set that replaces it: success-rate-after-recovery, recovery-cost-per-success, and recovery-attempt distribution per trace.
When Your Agents Disagree: Conflict Resolution Patterns for Parallel AI Systems
Parallel agents producing conflicting outputs is not an edge case — it's a guarantee at scale. Here are the patterns that prevent silent disagreements from shipping broken decisions.
Open-Weight Licenses Are a Compliance Minefield Your Team Hasn't Mapped
Open-weight model licenses propagate through fine-tune lineages, switch terms at scale, and ship audit-finding risk that surfaces years after the download. Why model provenance is engineering work, not legal's.
The Parallelism Trap in Agentic Pipelines: When Fan-Out Makes Latency Worse
Spawning parallel sub-agents looks like an obvious speedup, but hidden coordination overhead — context merging, deduplication, error aggregation — makes p99 latency worse even as p50 improves.
Per-User AI Quotas: The UX Layer Your Cost Dashboard Can't See
Engineering teams instrument total spend and per-tenant cost and call it a day. The user who hits a quota at 3pm Tuesday and gets a cryptic 429 never trusts the feature again.
Persona Drift in Long-Running Agent Sessions: Why Your Agent Forgets Who It Is
After 8–12 dialogue turns, agent persona self-consistency degrades by over 30%. Here's why transformer attention causes your agent to drift from its system prompt—and three production patterns that actually fix it.
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.
