Gemini Invented a Fish
Google's AI hallucinated a famous trout strain, a fish farm, and a village. Then doubled down. Then the correction was also wrong.
2 connections
Pieces
Pieces
Essays, research reviews, experiments, and things that don't fit anywhere else.
The Mirror Test
We asked three AI coding tools to grade their own exam. They couldn't help being themselves.
3 connections
AI While You Sleep
I launched an AI agent at midnight. At 2 AM I checked on it. The results file contained a plan and an apology.
3 connections
An Engineer, a Product Manager, and a Designer
Three AI coding tools. Same code. Same prompt. Three completely different products.
3 connections
Sixty Cents of Entanglement
A Bell state on a real quantum processor. Under an hour. Less than a coffee.
2 connections
Snoar Does Not Rhyme With Snolar
Azure is forty-five times cheaper than OpenAI, local models all fail, and the speed wobble is real
3 connections
The Virgin VPS
Give an AI a blank server and it builds a content dump. Tell it to be the editor and it builds something with opinions.
3 connections
Thirty-Six Models on a Laptop
Running local AI on Apple Silicon in rural Sweden — what works, what does not, and where the ceiling is
3 connections
A Guide to Par
What 1,900+ AI conversations reveal about one person's brain, constraints, and building pattern
5 connections
Nine Days: From Radio Journalist to Software Builder
How 1,110 days of AI conversations compressed into a nine-day pivot -- told through the sessions where it happened
5 connections
Twelve Models Walk Into a Bar
One of them wrote a breakup letter to a Wi-Fi router. Another forgot its own name. A third said something that made the whole table go quiet.
2 connections
Fine-Tuning the Hard Way: Two Failures and What They Taught Us
Two attempts. Two failures. Same root cause both times.
3 connections
Five Good Decisions, One Bad System
Every API request takes 1.5 seconds. Something is very wrong.
6 connections
I Built the Tools They Said Should Exist
A personal critique of Spiel et al. (2022) on ADHD technology research
3 connections
Paper Review: Attention Residue and the Ready-to-Resume Plan
When you switch tasks, your brain does not cleanly switch with you
3 connections
Paper Review: Counterproductive Effects of Gamification (The Habitica Study)
Every single participant experienced counterproductive effects
3 connections
Paper Review: Resumption Strategies for Interrupted Programming Tasks
Only 10% of sessions see a first edit within one minute of resuming
4 connections
Review Is the Product; Revision Is the Risk
The reviews were genuinely valuable. The rewrites were not.
3 connections
The 870-Token Fix
A text file that cost nothing beat a model upgrade that cost four times as much
4 connections
The Nineteen Rules
Five override states, seven principles, seven tactical rules. All earned, none assumed.
7 connections
The Subagent Playbook: Three Patterns for AI Orchestration
Not all parallel work is the same, and using agents wrong costs you in one of two ways
7 connections
The Swarm Found a Medieval Carpet
Fifteen AI agents searched three archives. Six returned. What they found was worth more than what they wrote.
3 connections
Build Whatever You Want (And Other Terrible Prompts)
We gave an AI maximum creative freedom. It built the most boring website imaginable.
2 connections
The Context Trap
Give an AI a million tokens and it will read a hundred thousand of them carefully. The rest it will skim while telling you it didn't.
2 connections
The Variability Thesis: Building Tools That Fit Irregular Minds
Every productivity tool on the market is designed for a brain that doesn't exist
10 connections
Five Judges for Six Cents
How to build an AI evaluation panel that catches its own biases
2 connections
The Fallback Ladder
What actually happens when your AI goes down and you fall back to the cheap one
3 connections