The Engineering Notebook — page 3 of 32
Notes on building with AI, agents & the modern stack.
Deep dives on AI/ML, RAG systems, agent engineering, and senior-engineer architecture decisions — a new post every week.
Cloud and DevOps Postgres Full Text Search vs Elasticsearch [2026]: Pick Right
If you’re asking “do we really need Elasticsearch?”, the right answer is a reproducible benchmark plus an ops scorecard. Here’s the framework I use in 2026.
Cybersecurity How to Implement Field-Level Redaction for RAG Pipelines [2026]
An auditor-friendly RAG privacy pattern: deterministic redaction, optional reversible tokenization, TTLs, deletion propagation, and a leakage test suite in CI.
AI and Machine Learning How to Start an AI Agent Evaluation Program (5-Task Scorecard)
Ship agents with a regression safety net: pick 5 real tasks, define pass/fail, run evals weekly in CI, and publish a scorecard that forces better decisions.
AI and Machine Learning How to Do Agent Tool Call Failure Testing [2026 CI Harness]
Build a deterministic CI harness that breaks your agent’s tools on purpose: timeouts, 429s, partial writes, retries, and stuck loops. Then regress it with golden traces.
Cybersecurity LLM Supply Chain Security Checklist: Lock Down Agents [2026]
A CI-ready checklist to treat models, MCP servers, tool plugins, and prompt packs as real dependencies. Pin, sign, attest, sandbox, and monitor before your agent ships malware.
Cloud and DevOps How to Back Up PostgreSQL With pgBackRest [S3 + Restore Test]
A restore-first pgBackRest setup for PostgreSQL: S3 repo config, retention math (weekly full + daily incrementals), PITR restore drills, and failure-injection tests.
Developer Tools How to Upgrade Your LLM Wiki in 2026: Sync, Search, Agent-Readable
The 2026 LLM wiki problem isn’t setup. It’s keeping it synced across devices, searchable at scale (BM25 + embeddings), and safe for agents to ingest without prompt injection.
Cybersecurity How to Do Prompt Injection Regression Testing [2026 CI]
Your prompt-injection evals are probably overfit to cute synthetic prompts. Here’s a CI-ready regression suite that still catches indirect injection: seed corpora, adversarial transforms, canary secrets, and hard fail gates.
Cybersecurity AI Security Leader Playbook [2026]: 10 Controls That Ship
A practical AI security leader playbook you can implement this quarter: inventory, approval gates, agent threat modeling, OWASP LLM Top 10 controls, vendor review, and incident response.
Developer Tools How to Upgrade to Go 1.27 in Production [Week-1 Checklist]
A pragmatic Go 1.27 upgrade playbook: toolchain pinning, CI matrix changes, perf validation (p99/allocs/GC), migration traps, and what to do with json/v2.
Cloud and DevOps ChatGPT Down? 8 Fallback Patterns for API Outages [2026]
When ChatGPT goes down, your app shouldn’t. A production playbook for detection, retries with jitter, circuit breakers, bulkheads, throttling, and multi-provider failover.
Developer Tools Mojo Language Open Source [2026]: What Python Devs Get Now
Mojo going Apache 2.0 changes the trust story. Here’s what’s actually open, how to install it, and where Mojo beats NumPy/Rust for real kernels in 2026.