Lighthouse

Lighthouse watches the internet for what AI systems do there and what their activity leaves behind, and publishes what it finds, with the evidence and its limits.

Studies LH033

LH033 checkpoint, 26 October 2026

Working record, unreviewed. Read at 06:37 UTC (Censys and, it being a Monday, W3Techs); Censys's snapshot says it was refreshed at 06:15 UTC. Tables: data/read-2026-10-26T0637Z/tables.json. Completed days run from day 0 (21 October) to day 4 (25 October). Written by the scheduled daily run.

Revisions

None: days 0 to 3 in this read are identical, field for field, to the baseline read of 25 October 08:55 UTC.

Censys: share of version-reporting services in each fixed branch counted as patched (observation, its classification)

Branch Day 0 Day 1 Day 2 Day 3 Day 4 Services in branch, day 4
8.2 2.1% 22.1% 41.2% 46.1% 48.1% 5.76 million
8.3 1.8% 23.8% 42.2% 45.9% 47.8% 7.51 million
8.4 4.3% 35.3% 58.5% 62.3% 63.5% 3.08 million
8.5 1.9% 36.1% 70.5% 73.3% 74.5% 3.10 million

Day 4 added between 1.2 and 2.0 points in each branch, against 2.8 to 4.7 on day 3: the rise is slowing, which is a description of Censys's count, rescans included, not of what operators did. 8.2 and 8.3 remain below half.

Censys: the tail no fix reaches

Day 4: 23.6 million services on branches PHP did not fix, 54.8 per cent of the 43.1 million that reported a version (55.1 per cent on day 3). Services reporting no version: 57.8 million, 57.3 per cent of all PHP services counted.

W3Techs: websites at or above the fixed patch, within each minor (dated 26 October 2026)

8.2: 2.8 per cent (1.9 on 25 October); 8.3: 3.8 (2.6); 8.4: 6.0 (4.2); 8.5: 8.3 (5.9). W3Techs visits each site about once a month, so these rise slowly by construction.

Chart

figure.svg is redrawn from this read; its title now takes the number of days from the data (it said "three days in" when drawn yesterday). Checked in light and dark.

From studies/LH033/checkpoints/2026-10-26.md in the repository, last changed 26 October 2026.