The Engineering Notebook — page 4 of 38
Notes on building with AI, agents & the modern stack.
Deep dives on AI/ML, RAG systems, agent engineering, and senior-engineer architecture decisions.
AI and Machine Learning Mercury 2.5 770 tok/s Benchmark: The Production Playbook [2026]
Mercury 2.5 is clocking ~780 tok/s on Artificial Analysis. At that speed, decode stops being your bottleneck. Here’s what actually changes in production: batching, streaming UX, context packing, cost controls, and evals that don’t lie.
Cybersecurity Claude Code AGENTS.md Telemetry Bug: The Control Lesson [2026]
Claude Code briefly gated local AGENTS.md instruction loading behind a remote flag that didn’t fetch with telemetry off. Here’s what broke, what was fixed, and the controls teams need so agents can’t silently ignore repo policy.
Cybersecurity How to Stop Repo Prompt Injection in Coding Agents [2026]
Repo-level prompt injection turns “clone and ask the agent” into a supply-chain compromise. Here’s the threat model, a safe demo, and practical mitigations.
AI and Machine Learning How to Use MLflow LLM Evaluation Tracing [2026] (Spans → Gates)
Turn MLflow traces into living eval datasets, rubric scorecards, and CI/CD regression gates for tool-calling agents. Stop debugging with random logs.
Cybersecurity MCP Server Security Best Practices: Checklist + CI Linter [2026]
A practical MCP server security test plan: authn/authz models, tool allowlists, prompt-injection regression tests, rate limits, audit logs, and a CI-friendly permissions linter.
AI and Machine Learning Jev Models Explained [2026]: Faster Routing, Reranking, JSON
Jev-style decision models are non-autoregressive models for routing, reranking, and structured outputs. Here’s how they work, where they fail, and how to evaluate them like a builder.
Cybersecurity RatHat Android Malware “AI‑Powered” Claims: What to Detect [2026]
RatHat is getting labeled “AI-powered,” but the real story is the Android stealer kill chain: accessibility abuse, permission traps, and noisy C2 loops you can actually detect.
Developer Tools LLM Knowledge Base GitHub Template [2026]: Starter Kit Repo
Ship a forkable LLM knowledge base starter kit: a clean repo layout, ingestion scripts, GitHub Actions freshness checks, agent-readable templates, and retrieval evals that don’t rot.
Cybersecurity How to Secure Local LLM Inference [2026]: Sandbox + Egress
A practical blueprint for secure local LLM inference: sandbox inference hard, default-deny outbound network, stage allowlisted downloads, scan artifacts, and isolate tools.
Developer Tools KoboldCpp GGUF Setup Guide [2026]: When It Beats Ollama
Run your existing .gguf models in KoboldCpp in under 15 minutes. I’ll show the exact settings for context, GPU layers, streaming, and when Ollama still wins.
AI and Machine Learning 7 AI Agent Swarm Coordination Patterns [2026]: Failure Modes Included
AI agent swarm coordination patterns decide whether your multi-agent system converges or spirals. Here are the topologies, stop conditions, budget caps, and failure modes most teams don’t test in 2026.
Developer Tools How to Prevent AI Coding Assistant Repeating Mistakes [2026]
Your assistant isn’t “forgetful.” Your workflow is. Here’s a repo-first system: mistake memory, patch-based retrieval, and regression gates that stop relapses.