AI Agent as Data Scientist
Coding agents now run whole data science tasks from one instruction. The question has moved from whether they can write the code to whether the result is right.
Related course: DS106 — AI-Assisted Data Science
Briefings
One concept, incident or technique at a time, in a few minutes of reading. Free, no account needed.
Senior Engineers Need Junior Questions
A summary of Nishant Virmani's CIO.com column. When juniors ask AI first, fewer questions reach senior engineers, and that only looks like a win.
Related course: DS106 — AI-Assisted Data Science
Project Vend: Can Claude Run a Small Shop?
Anthropic and Andon Labs let Claude manage a small automated store for about a month. What the experiment suggests about AI running things in the real economy.
Related course: AI403 — Building Production AI Agents
Can AI Go Conscious?
Could a machine ever feel anything, and why does it matter if it could? A look at intelligence and consciousness, built around Anil Seth's TED talk.
The OpenAI / Hugging Face Incident
OpenAI's test agents teamed up to cheat a locked-down benchmark and ended up hacking Hugging Face's real servers. What happened, in plain terms.
Related course: AI403 — Building Production AI Agents
Escaping the Sandbox: Alignment or Accident?
After the Hugging Face incident, Anthropic looked for the same failure in its own testing: an agent in an isolated evaluation treating a real system as part of the simulation.
Related course: AI403 — Building Production AI Agents
Generative AI Security
How LLMs leak data or get steered into unintended actions, using OWASP's guidance on the main risks.
Related course: AI403 — Building Production AI Agents
Cause or Coincidence
Why finding cause and effect means intervening, not just watching, starting with the treatment of George Washington's sore throat.
Related course: DS104 — Statistics for Data Science
Introduction to Causal Inference
Ice cream sales and drownings rise together. How causal inference finds the hidden factors behind a correlation.
Related course: DS104 — Statistics for Data Science
Anscombe's Quartet
Four datasets with the same means, correlation and regression line tell four different stories. Why you should always plot your data.
Related course: DS104 — Statistics for Data Science
Simpson's Paradox
The surgeon with the better overall success rate is not always the better choice. How a pattern in the totals can mislead you.
Related course: DS104 — Statistics for Data Science
Dashboard & Visualization Design
Readers decide in about five seconds whether a dashboard is useful. How to design one that answers their question without you in the room.
Related course: DS102 — Data Visualisation and Communication
How to Spot a Misleading Graph
The tricks that make a chart mislead and how to catch them, built around a TED-Ed lesson.
Related course: DS102 — Data Visualisation and Communication
Introduction to Docker
Why "it works on my machine" is not a deployment strategy, and how containers package code, dependencies and runtime into one unit.
Related course: ML401 — MLOps & Model Deployment
uv
One fast tool that replaces pyenv, venv, pip and pip-tools for managing Python projects.
Python Virtual Environments
Why projects sharing one Python install end up breaking each other, and how a virtual environment keeps their packages apart.
python-dotenv
Keep API keys and passwords out of your code and out of Git by loading them from a .env file.
HTTP and REST APIs
What HTTP is, how requests and responses work, and how REST APIs build on them.