Changelog: 2026-07-12
Green Flags are now publicly browsable with SEO support, while Culper gained governed investigation capabilities, first-class agent memory, and a redesigned operations console.
- Made Green Flags discoverable to everyone with public list and detail pages that show only headline-level information.
- Strengthened visibility rules so public and free users only see items intended for them across browsing, stats, and filtering.
- Updated submissions so new Green Flags start private, with a clearer publish and unpublish flow.
- Improved SEO and page performance for Green Flags with server-rendered loading and sitemap support.
- Increased chat reliability by handling budget-limited runs more gracefully.
- Refined issuer-risk research behavior with better intent routing and evidence selection, plus optional social-source signals.
- Expanded Culper with governed investigation capabilities, first-class agent memory, and a redesigned console for reviewing agent work.
- Green Flags for public readers: People can now view Green Flags without signing in, but only see a minimal, headline-style summary. Detailed evidence and other sensitive attributes remain available only to eligible tiers.
- Consistent access controls: Visibility checks were expanded across common entry points (lists, individual pages, related items, summary stats, and filter options) to ensure results match a user’s access level.
- Safer publishing workflow: Newly submitted Green Flags now default to private. Publishing is more deliberate, and authors can also revert their own public entries back to private.
- SEO + rendering improvements: Added server-rendered loaders, sitemap coverage, and shared filter parsing between browser and server so pages load consistently and hydrate correctly.
- Chat experience: When a tool-only run exceeds budget limits, the system now attempts to produce a final response instead of failing outright.
- Issuer-risk improvements: Enhanced how the product routes broad questions vs. issuer-specific research, tightened evidence matching to reduce ambiguity between similarly named entities, and added configurable social-source scanning to supplement research when enabled.
- Culper investigations: Added governed data access and capability discovery for agent-led investigations, plus versioned memory that can be recalled, reviewed, and retracted. A redesigned console makes investigations, approvals, and agent activity easier to follow.
- Product content update: Revised a public evaluation write-up to add context and use more careful, factual language.