Home / Articles / Almost working / Versions / v1.2.0
Almost working: what changed in v1.2.0
From v1.1.0 (2026-10-24, 264ddad06) to v1.2.0 (2026-10-27, db67c5ce1), paragraph by paragraph.
0 paragraphs added, 0 removed, 3 changed in place, 51 unchanged. About 0 words added and 94 removed. Insertions are marked like this, deletions like this; unchanged runs are folded to one line; figures appear as their file names.
3 unchanged paragraphs
!shot wk-pipeline.webp | images/ | Three agents, three stores, one site: the researcher writes the source watch (oq9gflln, raw captures); the curator promotes what holds intowatch, the evidence vault (3ytzxm3g, 227 curated sources);and the journalist updates the site from it. The schedule vault (weic8imn, public) records who runs, when, and what each run did.vault. | architecture
27 unchanged paragraphs, under In short {#in-short}, Slowly, and bit by bit {#slowly}, The problem: updatability {#updatability}…
!shot wk-handshake.webp | images/ | One handshake, then a schedule: the agent makes aagent's session key and pinskey, the root;editor's shows its fingerprintcheck and QR; the editor checks it and seals the vault key to itsealed on pki.sgit.ai; the agent opens it, clones the vault and deletes the file; a scheduled task wakes the same session.pki.sgit.ai. | diagram
10 unchanged paragraphs, under A feature, and a vulnerability {#feature-and-vulnerability}, The schedule vault: control is the queue {#schedule-vault}, Where it stands {#state}
!shot wk-state.webp | images/ | Almost working, on the evening of 23 October: what works (the researcher's 40 recorded runs, the handshake, the schedule vault, the journalist's views, the last-checked stamp),works, what is set up but not yet proven (the curator's wakes, run records written by the agents themselves, the researcher's schedule),proven, and what is missing. | comparison
11 unchanged paragraphs, under Next time, much easier {#next-time}, What this does not claim {#limits}, Where this comes from {#sources}