pi-fusion: Local inference-time scaling for coding agents
Why I built a local pi extension for OpenRouter Fusion-style planning panels, with the parts exposed.
leblancfg.com
Data Science, Geospatial Python, Space Stuff
Latest posts
I Spent My Sabbatical Building a Power Meter for Sledgehammers
One month, one prototype. The first Intensity Pad, and why I think striking movements deserve a power meter of their own.
Trying out the Absurd queue for AI Workloads
Testing out Absurd, an experimental durable execution system that uses only Postgres, for AI workloads with expensive LLM calls.
Tracking AI Model Deprecations with Git Scraping: Introducing deprecations.info
How I built deprecations.info using Simon Willison's git scraper pattern to track AI model shutdowns and prevent service disruptions
Addicted to My Power Meter
What happens when a CrossFit guy gets a smart trainer, discovers power data, and can't stop chasing the numbers.
When's the best time to surge when cycling up a hill?
A simple cycling simulator that calculates the time it takes to cycle up a hill with a surge of power.
I miss vim
A short reflection on neovim's extensibility vs simplicity tradeoff
First steps with Rust
Documenting my first steps with Rust, from installation to running a simple script.
TIL: Automating Command Execution Across All Tmux Panes
Learn how to automate the propagation of configuration changes across all tmux panes, saving time and enhancing productivity.
Being a good UNIX neighbour
A few tips to be a good UNIX neighbour and make your scripts more portable.
TIL: NERDTree Bookmarks
NERDTree Bookmarks is a plugin that allows you to bookmark files and directories in NERDTree.
The Data Bus: a helpful pattern for data engineering
The Data Bus is a pattern that can help you tidy up data flows, and helps to keep track of the preprocessing steps taken. It is a simple, language-agnostic pattern that can be used to make sense when multiple transform steps are involved.
Benchmarks for writing pandas DataFrames to SQL Server (ODBC)
In this article, we benchmark various methods to write data to MS SQL Server from pandas DataFrames to see which is the fastest. We compare multi , fast_executemany and turbodbc , and find the best candidate.
CrossFit Open 2019 Analysis - Exercise Statistics
In this article, we take a deeper dive into the benchmark exercise times and weights for the 2019 CrossFit Open athletes, including sprinting, running, squat, deadlift, clean & jerk, and snatch.