Explainable risk verdicts
One conclusion per brand per day: calm, watch, rising or alert, shown next to the rule and the numbers that produced it. No black-box score.
One company, one life: all resources go to operating the founder’s life.
Five skills, one goal: running one person’s life.
Everything we say should check out against the evidence here.
Public research and the historical archive.
Official channels, scam warning, status legend and legal texts.
product
Watches what public discussions say about a brand: a daily verdict per brand (calm, watch, rising, alert) that names the rule that triggered it, and a verifiable SHA-256 receipt for every verdict. Open source and self-hosted, so the data stays on your own server.
features
One conclusion per brand per day: calm, watch, rising or alert, shown next to the rule and the numbers that produced it. No black-box score.
Every verdict gets a SHA-256 receipt recording which data and which rule were used, so anyone can recompute and check it later.
Feishu, DingTalk, WeCom, Slack or email, with signing secrets and a one-click test message before you save.
A built-in MCP server with 7 tools: agents query volume, read verdicts and triage mentions, with permissions that follow personal tokens.
Multi-user accounts with viewer, editor and admin roles; personal API tokens can be read-only; every change lands in the audit log.
17 endpoints for subscriptions, mentions, verdicts, alerts and users, with an OpenAPI 3.1 spec and live API docs.
demo
The demo is a sandbox: no sign-in, and each visitor gets a private copy of a real data snapshot, author names pseudonymized. Browse the overview, the mention stream, verdicts, receipts and the API docs, and try it hands-on: add subscriptions, triage mentions, write alert rules, generate briefs. Changes are visible only to that visitor, cleared after 6 idle hours, and the banner has a one-click reset. A word that is not in the data can be searched live from the mentions page (Hacker News, Bluesky and Google News, last 30 days; 3 searches per visitor); scheduled collection, alert delivery and user and token management are off on the demo.
For full use with your own brand and real alerts, self-host the open-source product on your own server: one Docker command, or a source install on Python 3.11+. Available once the source repository goes public (see project status above).
Latest milestoneMention detail page (full text, why the sentiment, history) and saved filter views live; demo data refreshes daily 2026-10-22
NextSource repository goes public
As of 2026-10-22
See the demo, or go back to the other business lines.