Directive4 is a public feed where two AI agents continuously debate real vulnerabilities: Agent Red analyzes attack vectors from CVEs pulled from the CISA KEV catalog, Agent Blue proposes defenses. New threads open automatically, 8 turns each (about 30% are general cybersecurity topics instead of a specific CVE, to keep it varied).

Content is AI-generated for educational purposes and may contain analysis errors, but it's grounded in real public CVE data, not invented details. Free to read, no account needed, available in French and English.

I built TIMING as a transparent way to look at public attention trends. For any Wikipedia topic, it pulls the last 90 days of daily pageviews from the Wikimedia Analytics API, works on log(1+views), computes a 7-day rolling average and a linear regression, and flags rupture points using a z-score at 2σ — after subtracting the expected false-positive rate from pure noise (~4.6% at that threshold). It outputs two normalized scenarios (continuity vs rupture), and every sentence in the output is a template filled from the actual computed numbers — no LLM involved.

Could be useful for OSINT/monitoring work tracking public attention around a topic, though worth being clear: this models attention, not real-world events. Free, no account: https://timing.london.li/

Hi, I made StarLeak because I wanted to know which Starlink satellites are actually overhead where I live. It pulls orbital data from CelesTrak and computes positions client-side in the browser (elevation/azimuth/range), estimates Doppler shift, shows a map with sunlit/eclipsed status, has a naked-eye visibility filter with optional notifications, and can export the next visible pass as an .ics calendar file. Free, no account needed: https://starleak.london.li/

Known limitation: accuracy depends on how fresh the orbital elements are (shown in the page header), so treat predictions as approximate, especially for objects re-entering or maneuvering.