An exposed population register. Critical flaws in PHP. A $6.36 billion crypto theft. Thirty-one days of cybersecurity, seen through Hacker News.
Explore the monthWe trust strangers with our personal data, our servers, our money. This October brought reminders of what happens when that trust breaks. Three stories put the stakes in focus.
Security stories first seen on October 21 account for 40.9% of their daily cohort’s recorded HN points. More than two in five. Was that a sustained drumbeat or a single spike? Trace the full month below.
of recorded HN points in this day’s group went to security stories
8,625 OF 21,103 POINTS45,283 security points out of 484,547 across the full archive. The peak is one day, not the whole month.
Daily groups use the first observed front-page date. Points are the scores saved by the archive, not a count of votes cast that day. The final day is partial. Inspect the complete calculation
Find the patterns. Follow the stories. Make up your own mind.
In April, Project Glasswing put Claude Mythos Preview in defenders’ hands to find and fix weaknesses in critical software. The premise was a head start for defense.
By October 8, Anthropic said it had not yet achieved enough reduction in cyber risk. Finding flaws was easier; verifying, prioritizing, and fixing them remained hard.
A patch announcement and a stolen wallet sit side by side in this archive. They are different outcomes of the same race: can defenders turn discoveries into deployed fixes faster than attackers turn them into exploits? These headlines cannot establish which side is winning, or that AI caused an attack.
ARCHIVE SNAPSHOT
31 OCT 2026 · 20:18 UTC
We read every available daily file in hckr news’s front-page archive, plus its live page for days not archived yet. For October 2026, that is 2,878 unique story records across all 31 days. There is no minimum score. Dates are first observed front-page appearances in UTC, not submission dates. Jobs are excluded. Different stories can cover the same incident.
It captures all matching records returned by the archive, including low-scoring stories and those submitted before October. That is broader than the original daily top-30 sample. It still cannot prove that a story appearing only between polls was captured, or that the archive had no outages. Where available, each story also links to an independent HN Rankings record with actual top-30 observations. Missing rank history remains missing; no time on the front page is inferred.
For each UTC day, we sum the source’s saved HN scores for security-related stories and divide by the saved scores for every recorded non-job story first seen that day. The monthly figure uses the sum across the full month, not an average of daily percentages. HN points are a proxy for attention, not raw upvote counts, unique voters, or votes cast that day. Scores are snapshots.The final day is partial. Inspect every record in the denominator.
Title and URL keywords identify six themes, including defensive work and digital privacy alongside attacks. Explicit, documented corrections exclude unrelated meanings such as job security, hardware hobby projects, or a contractual breach. Each story has one primary category. Security-related here includes research and policy, not only confirmed attacks. Classification can still miss context; the original sources are one click away.
Points and comments show attention, not damage. They are the counts saved by the source; currently featured stories can be updated on refresh. The attention analysis stays fixed to its saved archive snapshot. 24 stories on October 21 is the busiest recorded daily count for 2026. Some headlines concern historical incidents, research, or policy rather than an attack that happened in October.