The Decision Log: How I Get My Reps In
Kevin Yien said something on Lenny's Podcast that stuck with me: to build Product Sense, you have to document your decisions. Write the date. Write the rationale. Write what you think will happen. Then come back months later and find out if you were right.
It's embarrassing how often you're not.
I've made this a habit anyway. I put myself in other teams' shoes, draft shadow roadmaps, and watch what ships. The goal isn't to be right — it's to get enough reps that being right starts to feel like instinct.
Here's what the log looks like.
LinkedIn “Tags” vs. Apple Intelligence
Recruiters and hiring managers are drowning in “Coffee Chat?” and “Referral Request?” messages. Every inbox looks identical. Nobody can triage it.
I proposed using NLP to tag message intent at the thread level — letting users filter and batch responses by motive, not sender.
Then iOS shipped Notification Summaries: AI-generated intent labels across your entire notification stack. Apple did for the whole OS what I wanted for one screen.
I'm calling that a win. Motive is the most valuable metadata in a crowded inbox. Someone clearly agreed.
The Apple CarPlay Hijack
We've all been there. You're mid-turn on an unfamiliar road and an incoming call hijacks the entire screen. Navigation disappears. You miss the exit. You add seven minutes to your commute while fielding a call from your dentist.
My call: Navigation is Priority 0. A phone call is a guest — it shouldn't redecorate the house.
Apple eventually shipped a compact call overlay that keeps the map in focus. Took them a while. They got there.
Spotify's Social Ghost Town
Spotify has always known what you're listening to. It just never let you talk about it. My friends and I traded tracks like a second language — screenshots, links, DMs across three different apps just to say “you need to hear this.”
The discovery loop kept escaping the ecosystem. My bet: Spotify would eventually close it with native messaging.
They did. The social layer was always there in the data. It just needed a front door.
🦾 The Takeaway
None of these calls were lucky. They came from making enough wrong calls first.
The rep isn't about being right — it's about building the pattern recognition to know why something will work before the market confirms it. In a world where AI can generate a hundred feature ideas in thirty seconds, that judgment is the differentiator.
These are a few pages from the log. If you've got your own — or a take on one of these — I'd love to compare notes.