Page 76
12 articlesThe Vendor-Portability Tax: Why 'We Can Swap Models' Is a Quarterly Cost Line, Not a Checkbox
Behavioral portability across LLM providers decays the moment you stop funding it. A breakdown of the quarterly burn rate — eval subscriptions, prompt-as-function-of-model routing, contract leverage — that turns 'we can swap models' from a slide into a real option.
Your Provider's 99.9% SLA Is Measured at the Wrong Boundary for Your Agent
A vendor's 99.9% availability is measured per call; your agent makes 12 per task. The arithmetic, the missing contract clauses, and the divergence alarm that catches incidents before users do.
Why Your Voice Agent Feels Rude: Turn-Taking Is a Latency Budget You Never Wrote Down
Why voice agents feel rude: a four-stage latency budget, hybrid turn detection, full-duplex audio, and a preemption contract that protects state.
Your Agent's Outbox Is Your Next Deliverability Incident
An agent fans out 80,000 emails before breakfast and the password-reset domain reputation is gone for six weeks. The subdomain, DKIM, and rate-limit discipline you need before the first send.
The 95% Reliability Illusion: Why Your 10-Step Agent Fails 40% of the Time
A ten-step agent at 95% per-step reliability succeeds end-to-end only 60% of the time. Verification placement, redundancy patterns, and shorter chains are the architectural levers that bend the curve.
Why AI-Generated Comments Rot Faster Than the Code They Describe
AI agents emit a fluent docstring on every function, paraphrasing the code rather than encoding intent. The moment the code shifts, the comment lies — and the next reader trusts the lie over the code. A code-review discipline for the AI-assisted era.
AI Reviewing AI: The Asymmetric Architecture of Code-Review Agents
When the author and reviewer agents share a base model, code review becomes a confidence amplifier rather than a quality gate. The asymmetric architectures, multi-pass critics, and eval discipline that turn AI review into actual signal.
Your APIs Assumed One Human at a Time. Parallel Agents Broke the Contract.
Internal APIs were designed for human-paced sessions. When users spawn parallel agents, the rate limits, idempotency assumptions, audit-log schemas, and CSRF flows all break at once.
Your Model Update Is a Breaking Change: The Behavioral Changelog You Owe Your Integrators
When a vendor silently rolls a minor model update, every downstream prompt becomes a contract that nobody honored — what a behavioral changelog should contain, why nobody publishes one, and what consumers should instrument while waiting.
The kWh Column Missing From Your Inference Span: Carbon Attribution Per Request
Sustainability disclosure is moving from corporate aggregate to product-level granularity. Engineering teams that measure cost per token but not energy per request are about to discover the dashboards they built solve the wrong problem.
Content Provenance for AI Outputs: C2PA, SynthID, and the Audit Trail You Will Soon Owe
By August 2026, a generative model's output is a signed artifact, not a string. Here's the architecture C2PA and SynthID demand — and why retrofitting it later costs more than building it now.
The Contestability Gap: Engineering AI Decisions Your Users Can Actually Appeal
Production AI agents turn refund denials, content removals, and verification rejections into final answers by accident. Build the durable record, the appeal endpoint, and a real second-look pipeline before regulators or angry users force you to.
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.
