Skip to content

Field report: Strategy Kernel + Third Finder produced breakthrough our frameworks alone missed #1

@bizbuilderai

Description

@bizbuilderai

We run a system with a similar philosophy to Anty — autonomous agents executing strategy for a B2B SaaS product. Multiple strategic frameworks in play: positioning (April Dunford-derived), relationship management, signal processing, and a Kill/Pivot/Double Down decision loop.

Saw the Rumelt Strategy Kernel in your framework list and wanted to share something we found.

The gap we hit with Strategy Kernel alone:

Rumelt's Diagnosis → Guiding Policy → Coherent Actions is excellent at analyzing what's wrong and forming a response. But for positioning specifically, it didn't help us find what we call the "cornerstone" — the thing ALL competitors rejected that becomes your unique foundation.

We added a step inspired by Kurpatov's non-substantive methodology (he calls it "Third Finder"): instead of analyzing competitors' strengths and weaknesses, you map the hidden shared assumption that every competitor holds. Then you build on what they all threw away.

Concrete example: In one project, three competitors in the space all shared one assumption — that a certain activity was a "task to be executed." One automated it fully, another gave experts better tools, the third taught people to do it themselves. The hidden binary was Control vs Trust. The "Third" that dissolved it: it's not a task at all — it's a signal system. Agents run micro-experiments; the founder reads signals and decides what to do next.

That reframe didn't come from Strategy Kernel analysis. It came from finding what every competitor rejected and asking whether they were wrong to reject it.

Suggestion (not a feature request, more a framework observation):
Your PMF detection module could be the natural home for this. The question isn't just "do we have PMF?" but "are we looking in the right place?" — and the "right place" is often where everyone else stopped looking.

Curious whether you've hit similar framework gaps in real-world usage, or whether the Rumelt+Goldratt combination covers this through a different path.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions