Outsourcing the Logistics of Staying in Touch
I have a group of girlfriends from business school who are scattered across the country, all of us over 40, most of us with little kids, and the kind of busy where months can disappear without anyone realizing nobody has texted the group chat. We landed on a small ritual to fix that...
Finally, an AI That's Actually Opened PowerPoint Before
After months of janky decks, hallucinated quotes, and layouts only a dot-matrix printer could love, Claude's PowerPoint integration felt like working with someone who had genuinely done this before.
The Instructions That Held, The Ones I Had to Keep Repeating, and the One That Taught Itself (Perhaps a 'Persuasion Bomb'?)
A few months ago I published a piece on using custom instructions to reduce hallucinations in AI research. What I didn't anticipate was how differently that protocol would behave depending on where the instructions lived, how I delivered them, and what happened when a tool decided on its own that I wasn't going to settle for guessing.
The Data Trail We Leave Behind - Part 2: Making Decisions in the Dark
In Part 1, I explored how my data lives in fragments across AI platforms, with each company holding incomplete and potentially distorted representations of who I am. The deeper problem isn't just that these fragments exist. It's that the companies controlling them are making choices that undermine the very trust they need from users...
Three AI Models Walk Into an Unsolvable Puzzle
My second grader came home with a logic puzzle worksheet. After some frustration at the kitchen table, I realized there wasn't enough information to solve it. I decided to test whether AI could confirm my suspicion.
Trapped in the Chat: Why I Rebuilt My Claude App in Gemini 3
We often talk about "Vendor Lock-in" with enterprise software, but I recently discovered a new flavor of it: "Prompt Lock-in."
Hands-on with "Nano Banana Pro": Context, Cartoons, and Guardrails
Yesterday's experiment started with a simple goal: test whether Google's latest image model understands context or just follows instructions. The results revealed both a massive leap forward in AI's a...
Custom Instructions to Combat GenAI Hallucinations
On Wednesday I shared an example of a wild hallucination that resulted from the Gamma implementation of GPT5. Today, I'd like to share some tips on how to use custom instructions in the models you use...
Data Science Deep Dive: Analyzing 3 Years of Peloton Data with Gemini 2.5 Flash
For today's GenAI experiment, I fed Gemini 2.5 Flash my personal three-year Peloton workout history, a 1,351 row CSV file and asked it to diagnose the factors behind my longest training breaks.
