Tom Dickson's blog

What 19 uses of Claude Code's /goal command taught me about pointing an agent at an epic and walking away: six ingredients that separate the runs that finished from the ones that stalled, and two template variants for the self-feeding work loop.

Why I built an MCP-native project management tool on Cloudflare Workers.

I built a browser game for learning the Solent racing marks: a nautical chart, a highlighted buoy, and a multiple-choice question against the clock.

Bermuda's weather for sailors: the subtropical synoptic picture, how to read the forecast models, and the live observations — with the sources I rely on.

A small Typst package that turns a year and month into a grid you can write on.

A wing-foil gear calculator that shows its working.

Most AI agents are stateless.

How I built an interactive knowledge graph connecting blog posts through semantic similarity and shared tags.

The OWASP Top 10 for LLMs focuses on commercial applications.

A practical guide to the MOD's Dependable AI directive.

I built a Claude Code plugin that applies The Economist's editorial standards to writing.

Applying Burnside's lemma and Polya's enumeration theorem to quantify the number of patterns I can choose to arrange the baubles on my Christmas tree.

I implemented an automated system to find related blog posts using embeddings and k-means clustering.

I investigate how it's possible to apply optimisation techniques and group theory to the ancient game of Terni Lapilli.

An interactive calculator built on the Aghion-Howitt model (2025 Nobel Prize in Economics).

How long does it take a ship to stop?

I adapted a naval combat model to Warhammer 40k and built interactive calculators using iterative expected values, Markov chains, and Monte Carlo simulation.

This article demonstrates how to enhance Python error handling using the traceback module while leveraging uv's support for PEP 723 to create reproducible scripts.

A short snippet for using `uv` to iterate on changing Python project dependencies.

How to add goatcounter analytics to a Next.js website.

Troubleshooting GPU configuration on GKE can be painful.

I've integrated Large Language Models (LLMs) into my workflows to automate developer activities.

Useful resources for sailing, specifically in the Solent.

I demonstrate how to serve a TensorFlow model from Scala by serializing it to the ONNX format, allowing it to be utilized through the Java ONNX runtime.

In this blog post, I explain how I implemented a GitHub Actions workflow to automate the updating of my Jekyll blog by building it and transferring files to my Droplet using `rsync`.

I explain how to perform metaprogramming in Scala using Scalameta to generate case classes based on a dataset schema defined in a CSV file.

Writing simple tests in Postgres without requiring extra dependencies.

Using org mode to generate documentation and infrastructure.

Literate programming with org-mode and sql.

I share a Python script that safely upserts Pandas DataFrames into a Postgres database using psycopg2, highlighting the importance of handling potential SQL injection risks.

Using scikit-learn and some Python features to set up a machine learning pipeline to quantify the efficiency of different classifers for classifying the movement of a stock price.

I wanted to use lower level linear algebra constructs in Python to calculate a covariance matrix.

A script illustrating how the memory usage of a pandas DataFrame can be reduced through altering data types.

Some notes on useful SQL commands and when they might be useful.

A review of different sources on when JSONB is suitable for use.

Useful resources for learning about data science.

Recovering point locations from vectors.

A review of the MIT 18.065 Spring 2018 opencourseware course, instructed by Gilbert Strang.

Some resources that I use to practice for passing coding interviews.

A review of numerical linear algebra for machine learning.

A review of mathematical concepts helpful for data science.

A review of the literature on uncertainty and sensitivity analysis when applied to optimisation models.