-
The Tutor Problem
One-on-one tutoring produces enormous learning gains. AI tutoring products mostly replace the wrong part of it, and the evidence for the part they do replace is older and better than the evidence for the part they advertise.
-
Which Mac Should You Buy in 2026? An Honest Decision Tree
A no-nonsense 2026 Mac buying guide for normal people: a decision tree from MacBook Air to Mac Studio, the RAM and storage truth nobody tells you at the Apple Store, and honest picks by who you actually are, student, writer, developer, creator, or family.
-
Cline and Roo Code: open-source agents teach you more than closed ones
Cline and Roo Code let you read the exact prompt, the tool definitions and the token bill behind every agent turn. Closed tools hide all three, which is why they're worse to learn from.
-
First Day With a New Mac: 12 Things to Change Before You Touch Anything Else
A practical first-day setup guide for a new Mac in 2026: the twelve macOS settings I change on every machine before I do real work. Trackpad, Finder, keyboard, FileVault, screenshots, and a Time Machine backup you set up before you need it.
-
Why Benchmarks Lie More Than Marketing Claims
Synthetic benchmarks arrive dressed as objectivity: numbers, graphs, decimal places. Here's how they mislead without ever lying, and the five questions that make them useful again.
-
What EdTech Companies Are Actually Selling
EdTech raised $20B in 2021 and mostly failed. AI EdTech is making the same promises with better vocabulary and almost nobody is asking why the last round collapsed.
-
JetBrains AI Assistant: not every green inspection is an agent task
IntelliJ's inspections come with quick-fixes that are correct by construction. Knowing which warnings deserve an agent, and which deserve Alt+Enter, is most of the value of AI Assistant.
-
Apple Doesn't Optimize for Speed. It Optimizes for Flow.
Benchmarks measure how fast a task finishes. Nobody experiences that. What people experience is the delay between touching the glass and the screen reacting, and Apple has been optimising that number for twenty years.
-
The Academic Integrity Crisis That Wasn't
The AI cheating panic runs on a sixty-year script: photocopiers, pocket translations, the web, now language models. Each round reveals what schools actually grade for, and it isn't learning.
-
The Hidden Cost of Power User Features Nobody Uses
Feature bloat is usually blamed on product managers chasing power users. The real cost is structural: testing surface, cognitive tax, and engineering time that never reached the core.
-
Windsurf: the value of a pause button in an otherwise relentless agent
Windsurf's Cascade agent is built to run long. Interrupting it costs less than letting a bad run finish. Field notes on the three signals worth stopping on, and what state an interrupt leaves behind.
