Watch & save
Visuals
15 frameworks and field notes — each one published as a LinkedIn post and built to be screenshot, saved, and reused in a design review.
From workspace silos to governed data
Your development workspace may be able to read production data. Unity Catalog did not block it. By default, catalogs can be accessible from every workspace attached to the same metastore, subject to privileges.
Seven Files That Never Ship to Production
AGENTS.md, rules, MCP, commands, skills, subagents, hooks, ignore — the seven files that decide whether Cursor knows your repo or guesses at it.
5 AI Failure Patterns That Look Fine in a Demo
Confident hallucination, stale RAG, prompt injection, silent fallback, runaway agent — every one of them passes a demo. A demo proves the happy path, not that the system is trustworthy.
The AI Failure Rings
Most AI apps don't fail at the model — they fail one ring out. Model, prompt, architecture, product, and trust: diagnose which boundary actually broke.
The model is not your AI agent
Eight layers and one control plane — the practical stack that turns an impressive agent demo into a system people can trust.
10 Cursor rules that stopped the AI breaking my code
Boundaries beat prompt tricks: plan first, protect secrets, commit before agent runs, verify APIs, and make “done” a real gate—not the model’s confidence.
AI in Action retrospective: four apps, one lesson
Four live apps, four runtime AI postures — what AI accelerated, what stayed human, and the architecture principles that connect the series.
AI in Action #4: I Used to Lie to My Habit App
Streak theater teaches fake completion. HabitCheck keeps Facts in code, puts Coach behind consent, and asks what happens after you miss a day.
AI in Action #3: Is your AI system ready for production?
RetireCheck, SleepCheck, and Readiness Advisor — why demos are easy, trusted AI is harder, and hard gates keep the model from talking past the score.
Six layers that make AI production-ready
Most AI diagrams put the LLM at the center. Production systems shouldn't — business value, trusted data, model strategy, security, evaluation, and CI/CD decide whether the system deserves trust.
Engineering interviews are stuck between two eras
A hand-drawn visual on why both LeetCode-era and "AI allowed" interviews miss the same thing — and the 45-minute exercise that tests engineering judgment instead.
AI coding → production cheat sheet
An animated hand-drawn cheat sheet covering the full AI-assisted delivery workflow — spec, generate, review, test, secure, gate, ship, watch, roll back — with real commands to copy.
Vibe coding gets you to a demo. CI/CD gets you to production.
An infographic on the 7-step path from idea to production — where AI accelerates, CI/CD validates, and engineers own architecture, risk, and outcomes.
What I learned from the first two AI in Action apps
A short LinkedIn video on AI-assisted development across RetireCheck and SleepCheck — where AI accelerates delivery, and where engineering judgment still shapes the product.
AI in Action: building RetireCheck
A short LinkedIn video introducing the AI in Action series through RetireCheck — practical app-building with AI-assisted engineering.