Pieces on product management, applied AI, and platform delivery.
-
Anthropic's Killer Product Is the Meter
Anthropic's vertical launches aren't an invasion of the industries they disrupt. They're load building - electric irons for the token meter underneath.
-
MPP: Agents can pay. It doesn't mean they should.
The Machine Payments Protocol elegantly solves payment mechanics for agents. The harder problem is everything around the payment: trust, judgment, identity, and compliance.
-
The harness frontier
The jagged frontier has a twin baked in by product design: the same model hits a completely different capability boundary depending on the harness it runs in.
-
The rise of the fragile expert
LLMs let a tool-savvy person perform close to an expert's output. The gap between output and retained understanding is fragility - and it's the cost of speed.
-
Claude, how is my driving?
A Claude Skill that reviews ceremony transcripts against Scrum.org's 8 Stances - a facilitation code review for Scrum Masters.
-
Smoke testing with Claude and Playwright
From a browser recording to CI-integrated Playwright smoke tests in roughly two hours - no QA team required.
-
The radical acceleration of human capital
AI lets employees leapfrog learning curves almost overnight. Companies that stay structurally blind to that multiplier will lose their best people.
-
Thrown Into the Machine
AI has made decades of inherited practice suddenly negotiable. That's not only a problem to solve - it's a rare window to think from first principles.
-
The DoorDash problem
AI agents are coming between platforms and their customers. Who survives depends on whether the moat is execution excellence or interface control.
-
Four hours to a custom standup dashboard
Building an internal standup tool with Claude Code and Streamlit: a meeting timer, auto-loaded action items, and the burndown chart Jira couldn't draw.