Writing
Notes on backend systems, performance, and the tools I lean on.
- My Local LLM On A 32 GB M1 Pro
How I adapted an M1 Max guide to half the machine by handing it to pi and letting it grill me first, plus the exact, reproducible steps that came out the other end.
- Django ORM: From sync_to_async Threads to Native psycopg3
A deep dive into how Django's ORM became truly async — why the old thread-pool trick was never real async, what changed in Django 6.0 with psycopg3 and ContextVar, and why asyncpg was left out.
- How I Optimized My Zsh Startup Time
An audit of my zsh config turned up shell forks, redundant plugins, and eager evals. Here are the eight fixes that shaved an estimated 400-700 ms off cold start.
- The Great Fork: How Redis Lost Its Soul and Valkey Found It
The story of Redis's controversial license change in March 2024 and how the open-source community responded by creating Valkey—a fork that's now outperforming the original on speed and efficiency.
- Default Apps 2025
Here are my Default Apps of 2025.
- Notes on Understanding Hebbian Learning
A beginner-friendly exploration of one of neuroscience's most elegant principles - how neurons that fire together, wire together.
- Cut Django Database Latency by 50-70ms With Native Connection Pooling
Deploy Django 5.1's native connection pooling in 10 minutes to cut database latency by 50-70ms, reduce connection overhead by 60-80%, and improve response times by 10-30% with zero external dependencies.
- ty: Astral's Take on Python Type Checking
Notes on ty, the pre-alpha type checker from the team behind Ruff and uv, and what its speed, its missing plugin system, and its open business model actually mean.
- Handling Django Authentication Redirects in HTMX Applications
A clean solution for handling Django authentication redirects in HTMX applications, preventing login forms from appearing in the wrong place when sessions expire.