Skip to content
HN On Hacker News ↗

The net · how every FogCam frame gets caught · Fog-Bank

▲ 16 points • 10 comments • by rdmuser • 6d ago • HN discussion ↗

Pangram verdict · v3.3

We believe that this entire text is AI.

89 %

AI likelihood · overall

AI
0% human-written 100% AI-generated
SEGMENTS · HUMAN 0 of 1
SEGMENTS · AI 1 of 1
WORD COUNT 255
PEAK AI % 89% · §1
Analyzed
Oct 4
backend: pangram/v3.3
Segments scanned
1 windows
avg 255 words each
Distribution
0 / 100%
human / AI fraction
Verdict
AI
Pangram v3.3

Article text · 255 words · 1 segments analyzed

Human AI-generated
§1 AI · 89%

FogCam publishes one picture at a time and throws it away when the next one lands. Fog-Bank keeps every one of them, and when something goes wrong (on my side or FogCam's), the record shows exactly when, for how long and whose fault it was. Four catchers on three networks ask FogCam for its picture: a poller on my own server every 7 seconds, fogline on Cloudflare every 15 seconds (it keeps a 30-day buffer), a Cloudflare worker every 20 seconds that writes straight to permanent storage, and a machine at my house every 15 seconds. Each one pushes every frame it gets into the archive, and sweeps every 5 minutes and every night pull back anything a push dropped. Every request asks for the real file (a unique query string plus no-cache headers), because in August I found FogCam's web server cache handing out copies minutes old. Each frame is named by FogCam's own upload time, fingerprinted with sha-256, stamped without re-compressing, stored as lossless JPEG XL and copied to Cloudflare R2 the same minute. A status board checks every stage from the camera to YouTube every few minutes and pages me on Telegram when one stops. The page lists every outage so far, in order, marked as mine or FogCam's: FogCam's server outage on Aug 13, its 64-hour freeze on Sep 8 to 10, my server's 40-hour outage inside that freeze (a full disk), and the rest. More: the live archive · how the backup works · the forensic record · the story