Blog
Writing about software engineering, AI, building products, and the chaos of shipping fast.
AI
12 posts
I Never Open My Editor Anymore
Visual plans, visual diffs, and visualizing everything instead of reading code.
What Token Optimization Is, and What It Isn't
Saving tokens is the wrong goal. Saving time is the one that matters.
Software Is Dead, Build Experiences
The holodeck is the future of software: experiences you build and change as you go.
AI Videos with HyperFrames
Script from an LLM, custom skills, HyperFrames and storyboards.
Using a Computer Is Like Star Trek Now
Voice in and out, asking for reports, and holodecks as the future of entertainment.
Plan with the Expensive Model, Execute with a Cheaper One
Planning with Fable / Astra, executing with Opus or similar.
It's Just a Prompt
Configurations are easy and cheap now, so design decisions can be different.
AI Doesn't Make You an Engineer
AI doesn't change your skills — it amplifies the ones you already have.
Bird Watching with Claude and AI
How I use an ESP32 and a local model to track the birds in my backyard.
Rewrite, Don't Fix
Why rewriting beats fixing — especially with LLMs.
How to Vibe Responsibly
Tests, automation, test dashboards, and a vibe kanban — the missing discipline in AI-assisted coding.
The Coding with AI Toolkit
Give your AI eyes, a finish line, and the right tools — the essential toolkit for AI-assisted development.
Game Dev
7 posts
Claude Code vs Codex for Game Dev
How I choose between Claude Code and Codex when making games.
Images and References Are Very Powerful
Why images and references do so much in AI game dev.
Prototyping with Godot
Why I prototype games in Godot.
Tripo3D First, Blender Cleanup Second
Generating models with Tripo3D, then cleaning them up with Codex and the Blender MCP.
Web LLMs for Game Ideas
Using Gemini and other web LLMs to come up with game ideas.
My AI Game Dev Workflow
How I actually build games day to day with AI in the loop.
Claude Taught Me Game Dev
Why you should make games — and how AI lowers the barrier to start.
Engineering
7 posts
Don't One-Shot It
Iterations with manual or automated gates, with proper planning.
No, I Don't Know Java. I Know Claude.
Languages don't matter anymore. Asking 'do you know Java?' stopped mattering. The world changed, and a lot of us haven't caught up yet. It takes change to see change.
Your Interviews Are Broken If You're Still Asking isPrime and Data Types
isPrime on a whiteboard and 'name every primitive data type' measured memorization — a signal that died the day every candidate got an AI pair. Test judgment, not recall.
The Future of AI and Code Is Visual, Not Text
Every industry solved how to look at its data — from subway maps to blueprints. We're still staring at text files.
How to Make Your Agents Smarter
Practical ways to level up your agents — memory, feedback loops, and the habits that compound.
Why Post-Mortems With Your Agents Matter
Running retrospectives with your agents — why reviewing what went wrong is worth the time.
What to Consider When Architecting New Ideas
The questions I ask before writing a single line for something new.
Productivity
8 posts
How to Come Up with Custom Skills
Spotting when something should be a skill.
Humans Can Multitask Now
For the first time, humans can actually multitask — how to do it well.
Why You Should Run Your Whole Project from SwiftBar
A menu bar is the cheapest mission control you'll ever build — glance at every agent, build, and deploy, and control them without leaving your screen.
The Custom Skills and Commands I Actually Use
My everyday kit of custom skills and slash commands — what each one does and why it earns its keep.
Design Better Plans: Ditching .md for HTML
Why I write plans in HTML now, and how it makes them easier to design and execute.
How I Code in My Sleep
Dynamic agent workflows, agent teams, and a caffeinated computer — when to run them overnight and how.
Build Your Own Tools, Stop Paying for Them
Why I'd rather spend an afternoon building it than another month subscribing.
Custom Scripts > LLM Token Chewing
Why I write custom scripts for repetitive tasks instead of burning through LLM context.
Voice
1 post
Hardware
2 posts
AI Agents
2 posts
When to Reach for /goal Agents (and How Much Thinking to Spend)
A field guide to delegating with /goal agents — when a goal beats a one-off prompt, and how to dial thinking effort up or down for the job.
Using Multiple AI Agents Effectively
How to orchestrate multiple AI agents for complex development tasks.
Education
1 post
Lifestyle
1 post
AI Safety
1 post