Krzysztof NowakThe AI Feature We Shipped, Measured, and Then Quietly Turned Off
Our smart search summaries looked great in demos. Real users scrolled right past them.
Artificial Intelligence
All 202Technology 8Programming 16Artificial Intelligence 12Data Science 7Design 8Startups 15Business 7Marketing 7Money 9Productivity 5Self Improvement 10Health 12Relationships 6Culture 7Politics 10Science 6Travel 6Food 13Books 9Sports 11Poetry 6Life 12
Krzysztof Nowak
Ethan ColeYour First AI Feature Should Be Boring
The AI features that survive contact with real users are rarely the flashy ones. Here is why I start with the dull jobs.
Elena MarchettiTeaching a Model to Spot Mould on Aged Cheese Was Harder Than It Sounds
Some mould is a defect, some is a feature, and the cheesemakers could tell the difference by smell.
Krzysztof NowakI Let an AI Coding Assistant Write My Tests for a Month
It wrote 412 tests. I kept 260. The ones I threw away taught me more than the ones I kept.
Krzysztof NowakI Counted How Often My AI Assistant Was Confidently Wrong
Over 300 questions in six weeks, it was wrong 1 time in 9. It sounded sure almost every time.
Ethan ColeEvals Before Vibes: How We Actually Test LLM Features
Clicking around in a playground is not testing. A small, honest evaluation set will tell you more than a week of demos.
Ben CarterMy LLM Bill Dropped 68% When I Stopped Doing These Four Things
Small, unglamorous changes to context, caching and model choice cut my monthly AI costs without users noticing any difference.
Ethan ColeHow We Cut Our LLM Bill by 70% Without Users Noticing
Large model costs creep up quietly. A few unglamorous changes brought ours down sharply while quality held steady.
Elena MarchettiLabelling 9,000 Tomato Photos Showed Me My Own Bias Before the Model's
I was sure the model was wrong about ripeness. Then I checked my own labels from Monday mornings.
Krzysztof NowakRunning a Small Language Model on My 2018 Laptop Was Slower and Better Than I Expected
Eight tokens per second, no internet, no bill, and a surprising amount of usefulness for boring tasks.
