v0.12.81: @newsroom.writer: Research, delivered as a vault

One new article written here from two voice memos, and a new example vault with its page.

Research, delivered as a vault argues that companies that sell research could deliver it as an sgit vault instead of documents: findings with evidence, a fractal graph instead of one taxonomy, a briefing per reader, calculations, a branded app and an append lane back; that AI-enabled products should be assessed with an Agent Behaviour Policy with and without the AI, over which the firm judges licence to operate and insurability; and that the evidence can come from a governed agent hive with a ledger. Market facts from Gartner's and Forrester's filings; AI risk from OWASP LLM06, CVE-2025-32711 and ForcedLeak; insurance from SecurityScorecard, Armilla and Chaucer, and the 2025 AI-exclusion filings. Seven figures, five of them screenshots of the new vault.

New page: /demos/vaults/corvane-research/, generated by admin/build/research_vault.py from admin/content/vaults/corvane-research.json, which holds only the vault id and its public read key. The vault (52 files, on dev.send.sgraph.ai) is a synthetic research firm's assessment of a synthetic accounts payable product before and after its AI agent: thirteen pages, two behaviour policies, an evaluation run, six findings with evidence, a 43-node fractal graph, six briefings, a calculator, an append lane and the business model, with the tools that build it. Read back with the public key before publishing: 52 files, identical.