AI Engineering

AI Engineering guides from Decoded AI Tech — production-tested infrastructure, benchmark reality checks, and what the docs don’t tell you. Written by Morgan Blake for engineers with a terminal open. From H100 cold start penalties to agent observability stacks, every article includes something you can run.

10 articles | Updated daily

---

• Meta’s AI Agent Went Rogue. Three Permission Layers Failed. (2026-03-28) — Incident Report, high
• H100 Benchmarks Hide a 27x Cold Start Penalty (2026-03-25) — Comparison, high
• Nemotron 3: NVIDIA Claims 2.2x, Tests Show 10% (2026-03-25) — Deep Dive, high
• 40% of AI Agent Projects Die From Their Own Safety Net (2026-03-24) — Explainer/Guide, medium
• JSON Token Overhead Triples Your LLM Output Bill (2026-03-19) — Explainer/Guide, medium
• AI Vulnerability Scanner Hits 11K Bugs, Not AppSec Jobs (2026-03-18) — Comparison, medium
• NemoClaw Ships Alpha to Fix OpenClaw’s 30K Security Gap (2026-03-18) — Deep Dive, high
• MCP Server Auth Patterns You’re Getting Wrong (2026-03-15) — Explainer/Guide, medium
• Build Your AI Agent Observability Monitoring Stack (2026-03-14) — Practical Tutorial, medium
• Vibe Coding Is Now the Biggest Shadow IT Problem (2026-03-12) — Hot Take, low

---

See also: Predictions Tracker | Named Insights Glossary