When an automation runs, but the business result doesn't show up
Source events, workflow runs and downstream receipts are different kinds of evidence. Includes the free local Outcome Reconciler and its important limitations.
Moruneva / Journal
Ideas worth
testing in public.
Guides, methods and transparent experiments at the intersection of digital operations and creative production. Each field note links back to a concrete tool, example or an explicit research question.
GuidesToolsEvidence notesField notes / October 2026
Start with the problem
Six fictional failures, one reproducible experiment
Downloadable source and results from a synthetic reliability simulation. A demonstration, not a customer success story.
What counts as an accepted automation handoff?
Define cases and uncertainty before the final sign-off, using the client handoff checklist.
Companion checklist →Shopify order missing from an n8n workflow?
Distinguish a missing trigger, execution record and downstream receipt before retrying an action.
Companion outcome tool →Before sending a work message: what should the reader do?
A fictional before-and-after example, six useful checks, and a local utility for finding missing actions, owners and evidence.
Companion message checker →From our research desk · not publicly released tools
Experiments we haven't finished.
We explore more than workflow reliability. These are short public descriptions of draft work; they are not production-ready services or claims that the underlying rights or outputs have been verified.
Does an AI task's cost tell you if it helped?
Per-task AI cost is observable; whether the work accomplished the intended goal is a separate question. We are testing a local spreadsheet comparison that keeps failed attempts and unknown outcomes visible. A provider invoice or a cheap average cannot prove the task was valuable.
Browse currently available tools →What evidence should travel with generated audio?
Tools such as Suno and ElevenLabs have terms and permission requirements that depend on how material is created and used. We are investigating a versioned handoff checklist, not a licensing certificate. Original provider rules and human review remain necessary.
See the current toolkit →One scene, many shot versions.
A scene asset might intentionally change between shots. A metadata preflight can flag a reference mismatch or a missing render label, but it cannot see whether two clips look consistent or decide which creative change is right.
Explore existing tools →Research first. Tools second. Better decisions third.
Moruneva is actively exploring where small independent software utilities help creators and operators do reliable work. Tool coverage will broaden where there is evidence of a real task—not just to fill a catalog.
Browse our tools →