SECOND MEASUREMENTS · RECOUNTS
Some people we read publish a number every month, or every release. We recount it from public data, a second way, and send them the whole result. Their figure and ours, side by side, updated each time they publish.
Microsoft Patches More Than 1,400 Security Flaws (13 Oct)
Their figure
1,400+
October fixes
Our recount
1,415
Microsoft CVEs on Patch Tuesday, from Microsoft's own update data
Holds · every count in the post
CVEs published in October 2026 (monthlyCVEStats, 1 Nov)
Their figure
17,060
October, from NVD; 89,859 for the year
Our recount
17,059
from NVD and from the CVE List alike; 89,857 and 89,858 for the year
Holds · within one; October is the largest month on record
“22% of those (38,995) landed on the registry with a dependency that could not be resolved” (27 Oct)
Their figure
22%
of 175,891 new npm versions in a day
Our recount
20.4%
of 527,331 over 26–30 Oct; 19.0–21.4% each day
Holds · our daily volume is lower, 100–110k
“We have only published two CVEs with severity HIGH since 2021” (7 Oct)
Their figure
2
HIGH by curl’s rating, plus one announced in the same post
Our recount
3
by curl’s records, counting that one; CISA’s scores put 55 at HIGH or CRITICAL
Holds · the CVSS gap in the post holds too
Each dot is one recount; a dashed dot is the next one. Every recount is a public data note with its scripts and a receipt in the Markovian log. These are notes, not papers: no DOI and no review gate. If you published one of these numbers and something here is wrong, write to hello@markovianprotocol.com and we fix the card.