Loop · Product
Close the loop from customer signal to shipped bet.
Interviews, tickets, and reviews land as typed insight — the loop closes when a shaped bet ships against them.
The discovery loop closes on the outcome (a shipped bet mapped back to the signal that motivated it), not the activity (an interview scheduled). Every call transcript, support ticket, and review lands as a typed insight on the same graph as the bets you’re already shipping. Agents cluster themes and draft candidate bets; a human gates the read-through and the shaping. After ship, the loop re-checks the signal so you know whether the bet actually retired the pain. The same signal-to-bet pattern applies on the retention side: see the same signal-to-bet pattern applied to churn. Shipped bets are also the raw material for the LinkedIn content loop, which turns what shipped this fortnight into posts the account owner approves.
For product teams whose research folder never quite makes it into the roadmap.
Or email ai@maskin.io with subject “Waitlist: product discovery loop” and we’ll send it the moment it drops.
What this loop is built with
Named the same on every loop page so the shape is portable. Below the fold on purpose — the loop is what closes; the primitives are just the parts. Inside a loop listing, agents / MCPs / skills appear as “built with” components of the loop, never as top-level marketplace sections of their own.
Loop
A closed cycle: something fires it (a signal, a schedule, a click), agents do the work, a human gates the taste-sensitive parts, and it ends on a concrete outcome — not on done for now. Every listing on the Marketplace is a loop. Maskin loops carry typed outcomes and verdicts that persist across tools, not recurring workflow execution inside one product.
Agent
A named role a model performs — an SEO analyst, a discovery synthesizer, an outbound writer. Not an AI teammate: a role with a job description, tools it can use, and gates it has to pass before its work leaves the workspace.
MCP
The wire an agent uses to reach the tools your team already uses — the CRM, the docs, the inbox, the analytics. If a human on your team can log in and click, an MCP-native agent can be given the same access under the same gates.
Skill
A packaged capability you attach to an agent — draft a follow-up in our voice, summarize a customer call against our discovery template. Skills are the reusable pieces of judgment your team is already building without realising it.