BOSTON, JUNE 23, 2026 — the block, archived
The Block
Campus + RevOps. B2B PM work.
Project questions and answers
What is this?
I was an APM at The Block, an institutional crypto research and media firm, and the part worth showing is that I built the tools, not just the specs for them. The clearest example for an AI role: three production Claude Skills, each with a human-review gate, that took recurring PM work off the team. Around them I built a RevOps automation pipeline that collapsed a manual deal handoff to a single Salesforce trigger, took the Polymarket sponsored-courses product from first PRD to launch, and wrote the company's internal strategy memo on agent-economy payment rails (x402). Most institutional B2B receipts get redacted before they ship, so this page keeps to what I can say out loud.
Why this approach?
At an institutional B2B firm the product is really the relationship, so a PM's week fills up with the same repeatable work: deal handoffs, client setup, status updates, course launches. I treated that repetition as something to engineer away. If a task came back every week, I tried to turn it into a Claude Skill, a Salesforce-triggered workflow, or a cleaner intake path, with a human still in the loop on anything that touched a customer. None of it was AI for its own sake. The payoff was fewer dropped balls and a PM who stopped being the bottleneck.
What would break?
Three places the work could have quietly fallen apart, and what I did about each. Closing a deal took seven manual steps across tools, each one a place the ball could drop, so I replaced it with an eleven-workflow Zapier pipeline triggered by a single Salesforce close, about three hours back per deal. The same PM chores came around every week, so I built Claude Skills for them that gave the team back roughly six hours a week. And the Block Pro 2.0 proposal could have sailed through on enthusiasm and died on capacity, so I ran a pre-mortem that surfaced the engineering-capacity and renewal-cliff problems before the pitch, not after.
What did I learn?
Two lessons stuck, one technical and one human. The technical one: a Claude Skill only saves time if people actually reach for it, and getting the other PMs onto my AI workflows took more onboarding and hand-holding than building the skill did. The human one counts just as much, because this was my first PM role. It's where I learned to lead a team's day-to-day, run the daily standups, and communicate clearly enough to keep engineering, design, and revenue pointed the same way. The systems were the fun part. Learning to move a team is the part I'm taking with me.
─ METHODS ─
| TASK | AGENT / TOOL | MODEL / COST |
|---|---|---|
| production tooling | 3 Claude Skills (etf-page-creator + stakeholder-update + jira-automation) | per-token billing |
| revops automation | Zapier + Tables (11 workflows + 10 intake forms + central DB) | SaaS |
| PRD authoring | Claude Code + Confluence | per-token + SaaS |
| AI media generation | Nano Banana Pro + Veo 3.1 / Kling 3.0 + ElevenLabs | per-credit |
| analytics | GA4 | free tier |
