GEO

Architecture · LikeC4

GEO

Measures and optimizes how generative engines surface a target product; reports share-of-voice, recall, and misrepresentation, and gates model-update drift

Open the interactive explorer → View on GitHub
built evolving planned research risk

The whole picture

System landscape

The system in context, then opened up into its containers.

GEO — system landscape

GEO — system landscape

index
GEO — containers (built vs planned)

GEO — containers (built vs planned)

geoSystem

Measures and optimizes how generative engines surface a target product; reports share-of-voice, recall, and misrepresentation, and gates model-update drift

Risks & open questions

Risks & open questions

risks
Planned & research work

Planned & research work

planned

Inside each box

Containers & components

Each part decomposed into the components that implement it. Every box links to its source in the interactive explorer.

Shared measurement pipeline — components

Shared measurement pipeline — components

sharedContainer

Corpus -> inference -> 3-layer scoring -> storage -> report. The engine all experiments reuse.

Coding-agent test harness — components

Coding-agent test harness — components

agentsContainer

Measures whether a coding agent discovers and recommends the product while solving a realistic task in a generated project

Model-update regression harness — components

Model-update regression harness — components

regressionContainer

Treats an engine version change as a deployment: snapshot metrics, compare to a baseline, and gate drift for CI

Intervention experiments — implemented vs scaffolded

Intervention experiments — implemented vs scaffolded

interventionsContainer

Each module measures one GEO tactic — an edit to the public content estate — against the shared pipeline. Some are implemented trackers; most are methodology + data scaffolds.

How it runs

Walkthrough flows

Dynamic views — the narrative spine of the system, step by step.

Walkthrough — one measurement run end-to-end

Walkthrough — one measurement run end-to-end

measureRun

Load the corpus, fan out prompt x lane x repetition against the engine under test, score deterministically (L1/L2), persist every call, then build the share-of-voice report. The paid L3 judge is opt-in.

Walkthrough — a coding-agent session

Walkthrough — a coding-agent session

agentRunFlow

A realistic project fixture is generated, the agent solves the task over multiple turns with simulated tools, the transcript is collected, then analyzed for product vs competitor signals using the shared text scorers.

Walkthrough — model-update regression gate

Walkthrough — model-update regression gate

regressionFlow

verify treats a version change as a deployment: run a fresh experiment, snapshot its metrics, and compare to a frozen baseline. A metric is only flagged when both effect size and a two-proportion z-test agree; the exit code drives CI.

Walkthrough — measuring one intervention tactic

Walkthrough — measuring one intervention tactic

interventionFlow

An implemented tracker edits/observes the public content estate, runs its probe set through the shared client, scores the outcomes, and reports the tactic-specific metric (e.g. recall vs content age for freshness decay).

Where it runs

Deployment

What runs where, and the process & data boundaries between the pieces.

Deployment — where each piece runs

Deployment — where each piece runs

deployment

Explore it live

The figures above are static exports. The interactive explorer lets you pan, zoom, follow relationships, and jump from any box to the source.

Open the interactive explorer →