A newsletter by Isaac Oresanya
I build AI applications in Python and write about how they work, the decisions behind them, and what I learn when I test them.
2,800+ subscribers on LinkedIn and Substack.
Start here
I turned my newsletter posts into an assistant you can ask questions. It searches my writing, answers only when the posts cover the question, and shows exact quotes that Python checks against the original text.
In testing, a separate check on whether the posts really cover a question made a clear difference. It refused 9 of 10 tricky, close-to-topic questions with the check on, and only 5 of 10 with it off.
Why do treemaps make ranking hard?
Treemaps make ranking hard because they show values through area, not length. Boxes can look different because they are wider or taller, even when their areas are close. The reader must compare boxes that may be far apart, so the biggest box may stand out, but the second, third, and later places take more effort to find. For showing a top 10, Isaac says a sorted bar chart is clearer and faster.
Treemaps do not work that way. They use area.The Problem With Treemaps
The reader has to move around the chart and compare boxes that may not even sit near each other.The Problem With Treemaps, “The Order Is Not Easy To See”
Explain the Data began with data analysis and how to explain it. These days I write mostly about building AI applications. A few posts to start with:
28 Jul 2026Data analysis
Your data can guide decisions without promising an exact result.
14 Jul 2026Data analysis
A simple rule for tracking customers you can actually trust.
7 Jul 2026Data analysis
It has nothing to do with charts.
23 Jun 2026Data analysis
Before you reach for machine learning, make sure you know the basics.
I'm an applied AI engineer. I build Python applications that answer questions from documents and turn text into structured records. Explain the Data is where I share what I'm building and what I learn from it.
I started this newsletter as a data analyst who wanted to explain findings to people who don't work with data. That habit carried over. When I build AI tools, I want people to be able to see where an answer came from.
I help teams build AI applications they can check and trust.
AI features built in Python: assistants that answer from your own documents and show where each answer came from, and pipelines that turn messy text into structured records.
Hands-on sessions for teams starting with AI applications: retrieval, structured outputs, and how to test whether the answers hold up.
Looking for the practice datasets and templates? They're still here: three datasets with practice questions, and nine guides and checklists for data projects.
Browse data resources