AI-assisted code generation was used in building MetaLoom. This page states when that started, what it covers, and what it does not change. It exists because you should not have to reverse-engineer the provenance of a codebase you are considering deploying.
Timeline
| Period | AI code generation | What that means |
|---|---|---|
2023 – 2025 |
❌ None |
The platform was designed and written by hand. The architecture, the data model, the pipeline mechanism, the REST and GraphQL surfaces and the original node implementations all originate from this period. |
2026 – today |
✅ In use |
AI-assisted code generation was introduced as a regular part of development. It works on top of the existing architecture rather than replacing it. |
The dividing line is a change in tooling, not in authorship or ownership. The project has had the same maintainer throughout, and every change — hand-written or AI-assisted — passes through the same review and the same test suite before it lands.
What AI Is Used For
Since 2026, AI assistance is applied to:
-
Implementation work inside an existing design — filling in nodes, endpoints, DAO methods, client bindings and UI components against interfaces and specifications that already exist.
-
Tests — endpoint, permission, DAO and end-to-end tests, which the project’s definition of done requires for every change.
-
Documentation — the specification tree that drives development, and large parts of this customer-facing documentation site.
-
Refactoring and migration — mechanical, repetitive changes across many modules.
What It Is Not Used For
-
No unreviewed code lands. AI output is treated exactly like a patch from a contributor: it is read, corrected and accepted or rejected by the maintainer.
-
No architectural decisions are delegated. The system boundary between Loom and Cortex, the pipeline model, the permission model and the data schema are human decisions. AI assistance implements against them.
-
No generated code is trusted on the strength of looking plausible. Behaviour is verified by the test suite, not by review alone.
-
No customer data is involved. Code generation happens during development, against the source tree. It is unrelated to the models a running MetaLoom deployment uses — those are covered on the Model Licenses page.
Review and Ownership
Every commit in the repository is authored and reviewed by the project maintainer. AI assistance is disclosed here at project level; it is not tracked per commit, so do not read the absence of a marker on an individual commit as evidence that it was written without assistance — or with it.
The obligations that apply to a change are the same either way: the code must compile, the tests must pass, the matching specification must be updated in the same change, and customer-facing behaviour must be documented on this site.
Licensing Impact
None on the license you receive. MetaLoom is licensed under the Apache License 2.0, unchanged before and after 2026. The use of AI assistance does not add any term, restriction or third-party claim to the source you get.
Two clarifications that are worth stating explicitly:
-
The models used to build MetaLoom are not the models MetaLoom runs. Development-time assistants leave no runtime dependency in the product. What a deployment loads at runtime — and the licenses attached to it — is inventoried on Model Licenses, including the models that are not cleared for commercial use.
-
Apache-2.0 already governs contribution and distribution. Nothing on this page narrows the grant. If your organisation applies a policy to AI-assisted code, this disclosure is what you need to evaluate it against.
Why Disclose This
Procurement and compliance reviews increasingly ask how a codebase was produced. Answering it up front — with the date the practice changed and the boundaries it operates under — is cheaper for everyone than answering it per enquiry, and it lets you make your own call rather than guess.
See Also
-
Legal & Licensing — section overview and the platform license
-
Model Licenses — the models a running deployment pulls in
-
Chat & AI Agent — the AI features inside the product, which are a separate topic from how the product was built