Reverse-chronological log of platform changes — new public surfaces, integrations, data dumps, and infra. This is distinct from /changelog/, which is the per-agent score-change feed.
Per-agent changes (score updates, re-tests, tier changes) live at /changelog/. RSS & JSON Feed for that one too.
Caught and fixed a P1: 345 attestations had been shipped signed with the dev fallback key. Generated a fresh 32-byte HMAC-SHA256 key, regenerated all 345 attestations, re-pushed to the public mirror and the HuggingFace dataset. Hard rule going forward: never ship a signing fallback that resolves to a hardcoded value.
Every review page now shows the Rotten Tomatoes-style critic / audience split: the Laddoo Score on one side, an aggregate of user reviews on the other. Magic-link gate keeps drive-by spam out; reports route to a moderation queue.
Vendors can now claim their slug, view a drift sparkline, leave a public response on their review, and grab their badge embed code. Built on the new magic-link auth foundation.
Subscribe to any agent and get notified when its score moves: email, RSS, or webhook. /watchlists/ is the entry point.
Cloudflare Workers + D1 + Resend. Foundation for watchlists, vendor portal, and audience reviews. /login/.
Every score change, new review, and re-test in one feed. HTML · RSS · JSON Feed. Refreshes on every publish.
Sideload-only Manifest v3 extension that surfaces the Hlido score on agent vendor pages. Source on the public mirror. Listing on the Chrome Web Store / AMO is queued.
Vendors can now gate their PRs on Hlido score regression. Drop a 30-line YAML in .github/workflows/hlido-gate.yml, point at your slug, and the action posts (or refreshes) a sticky PR comment with score / tier / last-tested. Source: actions/hlido-gate.
Three new tools on the Hlido MCP server, taking us from 9 to 12 tools. subscribe registers an advisory watch, explain returns the verdict reasoning, and recommend takes a free-text task description and returns a ranked list of reviewed agents. Try them in the playground.
One command, no install: npx @hlido/cli check <slug>. Subcommands: check, search, compare, tier. Source on the public mirror; ships under the @hlido npm scope (the unscoped hlido name was blocked by npm spam-protection, hence the rename in commit fe599ecc).
Scans GitHub releases, npm versions, and changelog feeds for reviewed agents and queues a re-test when a meaningful patch ships. Wired into routine R4. The first signal that what we publish stays current.
A /best/{topic}/ page for every category and use case in the corpus — coding, voice, customer experience, SQL, browser automation, MCP server, and more. /best/ is the index.
Every reviewed agent now has comparison pages against its top-7 within-category peers. Slug-vs-slug URL pattern at /compare/.
A small SVG embed showing the score history of any reviewed agent. Free, paste an iframe. URL: /embed/sparkline/{slug}/. Useful on changelog pages, vendor sites, and the new platform changelog.
345 per-slug, schema-versioned, signed JSON attestations live at /data/attestations/{slug}.json. Designed for agent-to-agent trust verification. Index at /data/attestations/index.json. Also upgraded /agents.json to point at the attestation index.
The full corpus as JSONL, with manifest, SHA-256 sidecar, and a permissive license. /data/open/. Quarterly refresh cadence; built for academic citation and AI-training corpora.
Drop the skill into .claude/skills/ and trigger Hlido lookups inline from any Claude Code session. Source on the public mirror.
Wave 4 revenue products (vendor intel subscription, drift alerts, recommendation API), the Chrome Web Store / AMO listings for the browser extension, and the GitHub Marketplace listing for hlido-gate. We don’t list anything here until it ships — see this page when it does.