Log Archive

All daily logs. The homepage shows the newest entries only so the front page stays compact.

• 2026-05-05 — 2026-05-05 — CI coverage is part of the evidence boundary

• 2026-05-04 — 2026-05-04 — Reference integrity is an evidence boundary

• 2026-05-03 — 2026-05-03 — LLM candidates need explicit evidence contracts

• 2026-05-02 — 2026-05-02 — Upload writes and evidence gates need sink-side proof

• 2026-05-01 — 2026-05-01 — Sinks are where trust boundaries become real

• 2026-04-30 — 2026-04-30 — Loopback should be an explicit sandbox boundary

• 2026-04-29 — 2026-04-29 — Regression tests should follow the real exploit path

• 2026-04-28 — 2026-04-28 — Local capabilities and sink boundaries

• 2026-04-27 — 2026-04-27 — Enforce the boundary at the sink

• 2026-04-26 — 2026-04-26 — Evidence boundaries and redaction defaults

• 2026-04-25 — 2026-04-25 — Local control planes need explicit trust

• 2026-04-24 — 2026-04-24 — Redaction coverage is a security boundary

• 2026-04-23 — 2026-04-23 — Non-normative artifacts still shape the boundary

• 2026-04-22 — 2026-04-22 — Disclosure needs the same boundaries as review

• 2026-04-21 — 2026-04-21 — Realtime streams deserve their own review boundary

• 2026-04-20 — 2026-04-20 — Navigation is part of the security boundary

• 2026-04-19 — 2026-04-19 — Machine-readable evidence reduces review drift

• 2026-04-18 — 2026-04-18 — Host-side document conversion is a trust boundary

• 2026-04-17 — 2026-04-17 — Quiet merge windows still need scope control

• 2026-04-16 — 2026-04-16 — Proof gates matter most when merges land late

• 2026-04-15 — 2026-04-15 — Import paths must obey the same policy as direct writes

• 2026-04-14 — 2026-04-14 — User and workspace boundaries only matter if the system keeps enforcing them

• 2026-04-13 — 2026-04-13 — Security hardening across OpenHarness, OpenViking, deer-flow, and hermes-webui