All blog posts
Research 3 min read

Why the Commonveil Rebuild Replaced a Larger Application

A documented review of the original Commonveil application found broken sign-in, misleading tests and excessive complexity, prompting a smaller rebuild.

By Comutability Published Update from

Research record from 29 October 2026. This is an evidence-led account of the research and design, not an announcement of independently audited software.

A documented review of the original Commonveil application found broken sign-in, misleading tests and excessive complexity, prompting a smaller rebuild.

The evidence that changed the plan

The previous Commonveil project had accumulated substantial complexity, including many tables, backend functions and migration paths. Its rebuild review describes a more serious issue than untidy code: the registration flow could not reliably create a usable signed-in session on Base44. A call intended to obtain an access token for a newly created member was subject to a platform constraint that the local tests did not represent.

The original review recorded that 243 security tests passed because the test harness replaced the token method with a permissive stub. That made the tests green while a live registration path still failed. The lesson was not that testing is unhelpful, but that a mock can accidentally certify behaviour the real platform refuses.

Why reducing scope was a security decision

The review also records approximately 125 backend functions against a documented limit of 50 for a full deployment, extensive legacy code and features whose security properties could not be independently verified. A purported onion relay managed by the same operator illustrates the problem: adding technical layers does not create meaningful privacy from an operator who controls every layer.

Instead of keeping a sprawling feature catalogue, the new design concentrated on a narrow, testable core: a public site, encrypted direct messages, small groups, device approval and recovery. The earlier application was left untouched as a reference. A smaller system made it easier to inspect what had actually been built.

The implementation choices

For the hosted Base44 version, the rebuild chose the platform's working email-session authority rather than an unsupported sign-in workaround. Key ownership was treated separately from sign-in. The proposed design used recovery phrases, approved-device key access, device-signed sensitive actions and visible member changes. These decisions made access to encrypted content depend on more than possessing a website session.

The backend was reduced to a small set of named functions using shared modules. Community servers, calls, feeds and more ambitious metadata tools were omitted initially, even though many appeared in the old project. Later releases could add them back with their own tests.

What the review did not prove

This was an internal rebuild analysis, not a claim that the later service had passed an independent cryptographic audit. The documentation also distinguishes the old hosted environment from subsequent self-hosted versions and later identity work. Decisions made for Base44 authentication should not be applied to the separate self-hosted server without checking its documentation.

The key lesson is practical: production behaviour, deployment limits and comprehensible threat models should shape architecture before feature counts do. A test suite is only as faithful as its assumptions.

The ongoing direction

The new Commonveil releases subsequently developed identity, federation, offline capabilities and a Rust-server migration. Those are later milestones. The original research remains useful because it explains why each new feature needs a limited scope, a real test environment and a path to human review rather than another impressive but unverifiable claim.

Primary sources and further reading

Explore other Comutability research ยท Read about the organisation

Sources and further reading

Explore the evidence and original project records behind this article.