Reading Notes 2026 Mar - Apr

My Medium Articles!

Let me start with my new article – in the last two months, I published two articles on TDS:

1. Beyond Code Generation: AI for the Full Data Science Workflow: In this article, I used Codex + MCP to connect Google Drive, GitHub, and BigQuery, and go from raw Apple Health data to a stakeholder-facing report in about 30 minutes — including data retrieval, XML parsing, BigQuery loading, and analysis. The bigger shift is not just faster coding, but AI participating in a much more end-to-end workflow, while data scientists still provide the problem framing, technical judgment, and domain context.
2. Beyond Prompting: Using Agent Skills in Data Science: I have been making one visualization every week since 2018. In my new article, I shared how that long-running habit became a practical use case for agent skills in data science. And more importantly, how skill can be used in broader data science use cases.

Reading List in Past Two Months

Now, let’s talk about the great articles I came across in the past two months.

Data Science & Analytics

1. Beyond Code Generation: AI for the Full Data Science Workflow: AI can now do much more than writing code and docs – it can automate the full data science workflow. In this article, I shared how I used Codex + MCP to connect Google Drive, GitHub, and BigQuery, and go from raw Apple Health data to a stakeholder-facing report in about 30 minutes.
2. Beyond Prompting: Using Agent Skills in Data Science: I have been making one visualization every week since 2018. In this, I shared how that long-running habit became a practical use case for agent skills in data science and how to use skill in real work environment.
3. Does Causal Inference Overly Rely on Assumptions?: Every causal estimate rests on assumptions – ignorability (no unmeasured confounders), overlap (treatment and control groups actually comparable), and correct model specification (no hidden functional form mistakes).
4. MMM Isn’t Causal: This is How Meta and Google Fix It: The central question that MMM is trying to answer is “What would happen to sales if we changed spend?”, which is a causal problem. Meta and Google use experimentation to establish causality.
5. Causal Inference Is Different in Business: Focus on the big business question, find easier alternative solutions if possible, and do 80/20 depends on what it takes to the decision.

Data Career

1. **From Dashboards to Decisions: Rethinking Data & Analytics in the Age of AI **: Decision Intelligence that can guide decisions is the next evolution, beyond traditional dashboards.
2. AI Is Coming for Data Science Jobs — But Not in the Way You Think: AI is automating the janitor work. Which means data scientists are finally being redirected toward work requiring creativity, strategic thinking, ethical judgment, and interpersonal skills — the stuff humans are actually good at.
3. The Ultimate Guide to Future-Proofing Your Data Science Career (2026–2027): How to stay competitive as a data scientist by being AI-powered, learn ML & ML Ops, and becoming a human-first strategist.
4. How to Run 1-on-1s That Your Team Actually Looks Forward To: A manager’s role in 1-on-1 should be to listen, ask questions, and find ways to help.
5. Good Data, Bad Decisions: The Analytics Paradox Nobody Talks About: The common reasons of why a clean pipeline doesn’t produce better decisions.
6. Future of Data Analytics: What’s Next After AI (2026 and Beyond): Why data analytics still matters and major shifts with AI.

AI and LLM

1. How to Make Claude Code Improve from its Own Mistakes: Using generalize knowledge command, daily reflections, and skills to make Claude Code working better for you every day.
2. Inside Meta’s Home Grown AI Analytics Agent: 77% of Meta’s Data Scientists and Data Engineers use Analytics Agent on a weekly basis.
3. Is MCP Dead? The Context Crisis That Broke Naive Tool Loading: MCP gives agents a uniform way to connect to and learn about APIs without bespoke client integration every time, but also created a context crisis – and how does Agent Skills and CLIs help here.
4. How to Build a Production-Ready Claude Code Skill: What are skills and the best practices to build one.
5. AI-Ready Data vs. Analytics-Ready Data: Analytics-ready data and AI-ready data are not two maturity levels of the same thing. They are optimised for entirely different consumers and punished by entirely different failure modes.
6. I Pay for 5 AI Tools as a Product Manager. Here’s Why Every Single One is Worth It.: The author shared his experience with 5 AI tools: UXpilot, Miro, Grammarly, Cursor, and ChatGPT.
7. I Built an AI Agent That Monitors Our Power BI Dashboards 24/7. It Caught a $180K Error at 3 AM.: A Python-based AI agent that monitors Power BI dashboards 24/7 and catches errors that humans and built-in tools miss.