AI tools and automation
AI tools and automation I've built to take repetitive work off people's plates. Some I use every day, some started as experiments in the first months the AI APIs were available. Together they show how I approach automation for clients.
Early LLM tools (2023)
I wrote my first AI-assisted production code in December 2022 and started building my own tools on the API in March 2023.
- Daily report pipeline. A scheduled Python job that runs 18 GPT-4 prompts, renders the results to HTML and PDF, and emails a daily report through the Gmail API.
- Commit-message writer. A tool that splits large code changes into chunks by token count and summarizes them with GPT-4. I've used it across more than 8 repositories.
- Playlist summarizer. A small Flask web app that pulls a YouTube playlist and summarizes each video's transcript.
Agent Loop Harness
A reusable runner for autonomous AI coding agents, written in bash. It works through a list of small user stories one at a time, each in a fresh session, with a cap on turns and spending per run. A change only counts when the project builds and its tests pass. I use it across more than 6 iOS and web projects, and it's how I deliver large amounts of work without giving up quality.
AI Provider Kit
A SwiftUI reference app for iPhone, Vision Pro and Apple Watch that puts 11 AI providers behind one common interface, covering chat, image and video generation, vision and voice. It tracks token usage and cost per provider, estimating when a provider doesn't report it, so it's easy to pick the right provider for the job and see what it costs.
Desktop automation suite
A cross-platform browser-automation app in Python, about 7,800 lines. It has a Tkinter control panel for running several accounts, stores credentials in the operating system's keychain instead of plain files, and uses Ed25519-signed licenses that work offline and are tied to a device. It has a pytest test suite and builds into one-click Mac and Windows apps with PyInstaller.
Deposit Estimator
An iPhone and Apple Watch app that quotes trip fees for mobile service providers. It uses MapKit routing to work out the drive, applies configurable time and distance pricing tiers, and shows the weather at the destination. Watch the demo on YouTube.
The same habits run through all of these: real logging, tests, secrets kept out of code, and costs you can see. That's what I bring to automation and AI work for clients. See business automation and AI for your business.
Where it stands today: the agent runner and the AI provider kit are tools I use in my own work every week, on more than 6 iOS and web projects.
Have a task you do every day by hand?
Tell me about it on a free call. If a script or an AI tool can do it, I'll tell you how.
See the AI service Next: the workforce platform Back to all work