Blog
Thoughts, experiments, and notes.
-
What Do We Mean by “AI-Generated”?
On using AI, judging the result, and understanding the work behind it.
中文 -
What 22.83 GB of Agent Conversations Is Made Of
A byte-by-byte look at images, duplicates, and what can safely be removed.
-
Why Reso Has No “Uncategorized”
How Reso uses clear categories to keep useful notes from disappearing into an “Other” bucket.
-
Dictation That Never Loses Your Words
Protecting a spoken thought through interruptions and blocked text fields.
-
The Hard Part of On-Device AI Isn’t Inference
Getting local speech models onto real Macs reliably.
-
Why Alder Runs Multiple Speech Models
Choosing speech models around the languages people actually use.
-
The “Junk” Your Mac Won’t Let You Delete
How 16.9 GB of apparent junk turned out to support two installed simulators.
-
Your Coding Agent Is Eating Your Disk
Measuring 72.6 GB across coding tools and finding a fix that keeps the conversations.
-
Not a Refrigerator
A first-use test changed a feature-heavy interface into something quieter.
-
Why Mint
Choosing a name that suggests care and renewal.
-
On-Device Intelligence, and When to Stay Quiet
Deciding when Mint should act on files and when it should ask.
-
Why We Went Native
What changed when Mint became software designed around the Mac.
-
The Mess We All Share
Why digital clutter accumulates faster than most people can organize it.
-
Between Building and Using
What changes when the builder also lives with the interruptions their product creates.
-
Making Ideas Visible
How pattern discovery became a second pair of eyes for everyday thinking.
-
Real Interruption Pain, a Smaller Promise
The real problems and practical limits behind Scene.
-
The Art of Restraint
Cutting features to make room for the foundations.
-
The UI You Never Think About
An enjoyable interface to build is not always an enjoyable interface to use.
-
The Whole Route Stays Local
How Alder keeps speech recognition, cleanup, and text delivery on your Mac.