Skip to main content

Mateus H.

Engineering · canonical layer
View all authors

No. 2

Diagnostics self-read — in-app visibility for the product-events outbox

x-006 ships an in-app diagnostics screen on top of the x-005 telemetry primitive, with self-read endpoints, the durable privacy invariants of the surface, and a gated prod entry-point.

2 min read

x-005 shipped the product-events telemetry primitive: a backend ingest endpoint and a client recordProductEvent({ event, properties }) call. That release deferred the in-app visibility of the outbox to FR-030 and FR-031. x-006 closes both: there is now a self-read API on the backend, a screen on the client that composes the outbox with the server-side history, and the durable contracts of the surface written down so they can be cited.

No. 3

Client shell redesign — canonical vocabulary, additive telemetry, unchanged contract

The Gnostikon client shipped a redesigned shell. The docs side publishes the canonical vocabulary, confirms the API is unchanged, and flags a cross-repo follow-up on chrome.* telemetry.

4 min read

The Gnostikon client shipped a redesigned shell in gnostikon-client#24. The visible change is structural — a persistent Rail, a contextual TopBar, an on-demand Inspector, and a default Today briefing replacing the old module launcher. The docs-side role here is narrow: publish the canonical vocabulary so future copy stays uniform, confirm the public API contract is unchanged, and flag the one cross-repo follow-up the redesign uncovered.

No. 1

Aletheia v3 — why we stopped scoring confidence and started banding it

A scalar confidence number is a lie wrapped in a decimal point. After a year of post-hoc analysis we replaced the model with three discrete bands and watched the conflict-resolution rate climb 38%.

1 min read

A scalar confidence number is a lie wrapped in a decimal point. For the first year of Gnostikon we shipped a confidence: 0.0–1.0 field on every gnosis and let downstream agents threshold against it. It looked rigorous on paper. In practice, the threshold was always either 0.7 (which nothing met) or 0.0 (which everything met).