Docs

What We Are Building

An overview of Meraki, Meraki Core, Studio and the evidence-to-guidance learning loop.

There are three names. They should not be confusing.

Catalyst#

Catalyst is the lab.

It asks whether AI systems can improve through use without turning corrections into hidden, permanent rules.

Meraki#

Meraki is the user-facing product.

A person records a correction, choice or outcome. Meraki shows the proposed rule, where it applies, what happened on a later run and whether the rule should stay.

The public product is still early. Pricing paths describe intent, not a finished self-serve service.

Meraki Core#

Meraki Core is the engine.

It owns:

• immutable evidence and provenance;
• versioned, scoped profile rules;
• task-specific guidance packs;
• agent run records;
• baseline and targeted-removal evaluations;
• approval, edit, rescope, weaken, split, revoke and rollback commands;
• Studio read models for inspection.

Core is deliberately stricter than a memory store. Evidence, interpretation and action are separate stages.

The first proof#

The smallest useful demo is:

user corrects one result
-> candidate rule is visible but inactive
-> user approves or edits it
-> next matching run changes
-> unrelated run does not change
-> removing the rule removes the improvement
-> user can revoke it

The local code tests that path. The live hosted path is not finished yet.