-
Why Benchmarks Are Lying to You About AI Progress
AI benchmarks are contaminated, over-optimized, and poor proxies for real capability. Why the numbers in every AI press release are less meaningful than they appear.
-
The Multi-Mac AI Cluster — Insane Overkill or the Future?
Can you build an AI cluster from multiple Macs? A hands-on look at exo distributed inference on Apple Silicon, Thunderbolt bridge performance, real token-per-second numbers, and when a Mac cluster beats one big Mac Studio.
-
Agentic coding: How to review a PR an agent opened in the middle of the night
A field-tested take on how to review a PR an agent opened in the middle of the night with Agentic coding: what it rewards, where it breaks, and how to keep the workflow honest.
-
When Software Becomes Invisible: The Ideal State of Technology
Explore why the best software becomes invisible during use. Learn how technology achieves its ideal state when users focus on their goals rather than the.
-
The Next Billion AI Users Won't Speak English. That's a Bigger Problem Than You Think.
AI's English-language bias is measurable, systematic, and being systematically deprioritized. What it means for global adoption and who's trying to fix it.
-
Why I Downgraded From the Biggest AI Model — and Got Better Results
The contrarian case for small local LLMs on Apple Silicon. Why downgrading from a 70B to an 8-14B model improved real productivity through latency, iteration speed, and prompt discipline — plus the escalation rule for when big models still win.
-
MCP servers: Composing agents over a catalogue of shared MCP tools
A field-tested take on composing agents over a catalogue of shared MCP tools with MCP servers: what it rewards, where it breaks, and how to keep the workflow honest.
-
Why Apple Usually Doesn't Win on Paper, But Wins in Practice
Discover why Apple products consistently deliver better real-world experiences despite lower spec sheet numbers. Learn how integration, optimization, and.
-
The Real Reason Tech Layoffs Follow a Specific Pattern Every Single Time
Tech layoffs follow a predictable template — Tuesday announcements, immediate badge revocation, simultaneous profit reports. The structural reasons why are more revealing than the PR.
-
The AI Language Learning Method That Replaced My Duolingo Streak
How to learn languages with LLMs on a Mac. Complete AI tutor system prompt, graded reading with rewritten news articles, Anki flashcard generation, and pronunciation feedback using Whisper transcription — honestly compared to Duolingo.
-
AI coding workflows: Treating diffs as the primary artefact, chat as exhaust
A field-tested take on treating diffs as the primary artefact, chat as exhaust with AI coding workflows: what it rewards, where it breaks, and how to keep the workflow honest.
