Exploring AI, product thinking, innovation, behavior, and the future of work
Your AI assistant doesn't know what day it is. Here's why that matters and what to do about it.
I asked my AI assistant to check my calendar and block time for deep work. It queried the wrong date and proposed time blocks that had already passed. This isn't a bug. It's a fundamental limitation of how language models work, and it breaks more products than you think.
15-minute windows between classes, BJ Fogg's behavior model, and the app I built when no todo app solved the actual problem.
In an MBA, your day is a patchwork of 15-minute windows between classes, clubs, and recruiting. I kept opening my todo app, seeing 47 tasks, and losing half the window just deciding what to do. When my professor asked 'what's the Job to Be Done?' I realized every todo app solves the wrong problem. So I built one that solves the right one.
I already knew the answer. Four model tiers and 45 minutes later, Claude agreed.
I was optimizing my AI workflow at 3am when Claude gave me the answer I already had in mind. Instead of accepting it, I ran an experiment -- escalating through four model tiers to see if a 'smarter' model could do better. It flip-flopped through three positions and landed back where it started. Research shows this happens 60% of the time.
208 ideas, 6 templates, 3 deep dives. A PM's guide to structured creativity.
What if creativity wasn't magic? After applying Systematic Inventive Thinking to 9 products and generating 208 ideas, I found that 70% of breakthrough innovations follow just six patterns.
