McGinnis, Will

I wrote recently about why uv changed how I think about Python projects.

I’ve been writing Python for a long time, and for most of that time, packaging has been the worst part.

Every company has a closet full of old code.

I’ve been on both sides of the tech debt conversation.

I’ve written twice now about Makefiles: once about why they’re useful and once about advanced patterns.

I wrote previously about why Makefiles are useful for Python projects.

I’ve been hiring data scientists for the better part of a decade now, and the job has changed more in the last two years than in the previous eight combined.

Navigating a Career in Data Science I’ve been in and around data science for over a decade now, and the field has changed dramatically in that time.

I’ve been building and using MCP servers for about a year now, and at some point I realized they’d quietly grown to cover most of the software development lifecycle.

Over the past few months, we’ve taken a deep dive into the world of rating systems through the lens of Elote, my Python library for implementing and comparing competitive rating algorithms.

mutmut-mcp wraps the mutmut mutation testing tool so an agent can run mutation tests and act on the results without you doing anything.

writing-tools-mcp is the MCP server I’ve written about the most, starting with the initial deep dive and then the MCPB packaging update.

todolist-mcp was the first MCP server I built that wasn’t wrapping an existing tool.

Quick one today: ContextSwitch is now free on the Mac App Store.

We’ve reached the penultimate post in our rating systems series!

I’ve been using uvx (the tool runner from uv) to run Python CLI tools from PyPI for a while.

I shipped idea.log to the App Store recently.

If you connect enough MCP servers to your AI client, you eventually start wondering: is this combination of tools actually safe?

We’ve covered quite a range of rating systems in this series so far: Elo, Glicko-1, Glicko-2, TrueSkill, ECF, and DWZ.

Most MCP servers are passive.

GitHub branch protection is one of those features that most people either ignore entirely or enable with whatever defaults seemed reasonable at the time.

When I first released stargazers last year, installation meant cloning a repo and running pip install.

Our tour of rating systems continues!

I’ve been using makefile-mcp as part of my standard coding workflow for a while now.

I maintain a few CLI tools - rampart, vanity, stargazers - and I wanted them all installable with brew install.

I have a graveyard of ideas.

We’ve explored Elo and the Glicko family of rating systems so far.

hugo-frontmatter-mcp has been one of my most quietly useful MCP servers.

If you’ve ever had more than one GitHub account - a personal one and a work one, or an old employer’s account you can’t log into anymore - you know the feeling.

I maintain a handful of open source Python libraries.

In this final post of our series on bankroll management strategies in Keeks, we’ll bring everything together with a comparison of all the approaches we’ve explored.

Two small but useful additions to rampart landed this week: automatic default branch resolution and HTML report output.

I have about 140 ideas in idea.log right now.

I should say upfront: for single-user utilities, many or even most of the things people build as MCP servers should probably just be skills.

In our previous posts, we’ve explored sophisticated bankroll management strategies from the Kelly Criterion to Dynamic Bankroll Management.

Note: This article represents a practitioner’s interpretation of the relevant rules and regulations in place at the time of writing.

I have too many GitHub repos with no branch protection at all.

When you publish an open source library, you have some idea of who might use it.

Following my fall reading list, here’s what I’ve been reading this winter.

In our previous posts, we’ve explored various bankroll management strategies, from the mathematically optimal Kelly Criterion to the capital-preserving CPPI.

Every productivity app eventually becomes Notion.

Last year I wrote a comprehensive guide to Python library development covering everything from project setup to PyPI publishing to security auditing.

I’ve been working on a trading algorithm that uses reinforcement learning to make buy, hold, and sell decisions.

In our previous posts, we’ve explored various bankroll management strategies, from the mathematically optimal Kelly Criterion to the straightforward Fixed Fraction approach.

When Ruff announced it had reimplemented Bandit’s security rules in Rust, my first thought was skepticism.

Every researcher has a graveyard of old code.

In our previous posts, we’ve explored mathematically sophisticated bankroll management strategies like the Kelly Criterion, Fractional Kelly, Drawdown-Adjusted Kelly, and OptimalF.

I’ve been writing about decision frameworks for a while now.

My software development workflow has gotten messier since I started using AI tools heavily.

In our previous posts, we explored the Kelly Criterion, Fractional Kelly, and Drawdown-Adjusted Kelly strategies.