Essays

Long pieces, one a month, each with its sources in the margin.

• 2026-10-13 Where I stand with AI, and where a reader might One year from doubting that AI assistants were any use to me and worrying about what they cost in energy, to using one every working day, told in order from my own record, with the doubts that stayed.
• 2026-10-03 Testing a skill with planted errors and a run without the skill Claude built a unit checking skill for me and tested it on a calculation sheet with four planted errors, on two models, with and without the skill.
• 2026-08-23 Prediction and interpretation are two different uses of a fitted model A fitted model can forecast an output or be asked which inputs it relies on, and this is the vocabulary for the second use with the limits its own documentation gives.
• 2026-02-15 What the DOE generative AI policy asks of the person using the tool I read the US Department of Energy policy on generative AI to write a short post about it, and this is what it says about the user, where my own summary was wider than the text, and why I think others will adopt it.
• 2026-02-10 What a SHAP value says and what it does not A SHAP value splits one prediction of one fitted model among its inputs, so it describes the model, and there are four common readings of the plot that it does not support.
• 2026-02-08 Ask where the numbers came from before charting them An assistant's first analysis of chemical engineering employment came with data it had made up, so I pulled the real series from the BLS and Census sources myself, and that took the evening.
• 2026-02-02 Telling an assistant to stop agreeing Over about a week I had to tell Claude several times to stop agreeing with me, and these are the things I told it: search and argue, here is what the term means, ask me questions first.
• 2026-01-16 Git and GitHub for one person What commit, branch, push and .gitignore mean for one person working alone, and how I ran into each of them.
• 2025-11-25 Kernel models in plain terms What a kernel is, what C, epsilon, gamma and the length scale mean in support vector regression and Gaussian process regression, why the inputs are scaled first, and what the uncertainty from a Gaussian process rests on.
• 2025-11-25 Machine learning vocabulary in terms a process engineer already knows Twelve words from the scikit-learn workflow, each explained by a run table or a calibration curve, with the name the library gives it.
• 2025-11-25 Tree ensembles in plain terms What a decision tree, a random forest, gradient boosting and XGBoost are, what their main settings mean, and why none of them can predict beyond the range of the runs they were trained on.
• 2025-11-19 A small neural network in plain terms What the units, layers, activation and epochs of a multilayer perceptron are, why its inputs are scaled first, and why the fit on a small table changes from one start to the next.
• 2025-11-19 Starting Python and machine learning from zero with an assistant as the teacher The first days of learning Python and machine learning as a chemical engineer with no coding, with the questions the assistant asked, the plan it gave and what two evenings covered.
• 2025-11-19 Starting Python twice as an engineer who had never programmed How I started Python twice, first in a hosted notebook with nothing installed and then on my own computer where the setup took most of the effort, with the errors I hit and the order I would install things in now.
• 2025-11-16 Colab, Jupyter or VS Code, and which one for which job What a notebook, a cell, a kernel and a virtual environment are, for an engineer who has never programmed, and what I ended up using Colab, Jupyter and VS Code for.
• 2025-11-06 Checking the numbers an assistant reads out of a long PDF I gave Claude a long technical document of my own and it reported a number I still have to check, and later I caught things missing from what it put together from the same material.