Writing & observations.
On product decisions, AI, and the way teams build.
On Spotify's product loop
Most takes on AI-assisted software development currently revolve around shipping code faster or replacing engineers with super-PMs. Spotify just published a refreshing take on how modern teams should adapt to produce value, hinting at the real changes the AI revolution is requiring from organizations.
The Builder PM trap
AI makes building tempting for PMs, but the Builder PM path is a trap: it undercuts engineers, creates unsustainable delivery overlap, and pulls PMs away from the problem space only they can own. Instead of shipping MVPs, PMs should use AI to master their actual job: research, framing, and stakeholder alignment.
Turning AI speed into leverage
Engineers now code 5-10x faster with AI, but velocity is a team effort and the bottleneck has simply shifted upstream. How can product teams re-calibrate their operating model so the new AI speed becomes real leverage and value, not just more tickets shipped?
An experiment with GPT-4 vision and Excalidraw
Sketching flowcharts on paper is quick, but sharing or editing them afterwards is painful. Here's how I wired together GPT-4 vision, Mermaid, and Excalidraw to turn hand-drawn diagrams into editable digital charts, and what worked, what broke, and what I'd do next.