Blog
Thoughts on programming, AI, and building things.
I benchmarked 8 LLMs as coding tutors for kids across 131 conversations. The cheapest model beat the most expensive ones, but only when I gave it the right prompt. Same model, different prompt: a 30-point swing.
I Create 6 Massive Code Projects (10K lines+) In 3 Months
Been playing around with Claude Code a lot lately
We actually need to not get hacked...
All sounds great, I still can't jockey-code
Introduction to jockey coding methodology - learning to command and steer AI tools effectively rather than letting them wander aimlessly.
Understanding the foundation of all apps - database design and schema as the critical source of truth for AI-assisted development.
Maximizing AI potential through strategic tool selection - balancing imperative instructions with declarative frameworks.
Getting to working code fast - using prototypes as living documentation and source of truth for AI-assisted development.
AI getting gold in IMO is not that surprising
AI has completely flattened the human informational landscape.
Coding is dead; Designing is alive;
I just entered flow state prompting Claude Code
Rust: First Impressions
Just completed 100 exercises to learn Rust, here are my initial impressions
Why LLMs are misunderstood
What does prompt engineering show us about LLMs
LLMs Evaluated
After more than 2 years of using LLMs daily, this is my verdict in 2025
Architecting In The Age of AI
Maybe we should all stop typing code
Human: Tool Choice
Are you chosing tools for yourself, or for the AI?