The 10 biggest things in tech, Wed, Oct 7
The ten biggest stories in tech today, scored 1–10 and ranked from 150+ engineering sources. Free, updated every morning.
OpenAI solves 90 open math problems
OpenAI published 722 mathematics manuscripts solving 90 of the top 500 open problems — including the Quasi-Riemann Hypothesis and sub-n-log-n integer multiplication — generated by an unreleased internal model, with independent experts calling it the most significant moment in mathematical history in over a century.
#1 · ai · scored 9/10
FortiBleed campaign still active at scale
FBI and Secret Service confirmed the FortiBleed campaign has harvested 86,644 Fortinet device credentials across 194 countries using FortigateSniffer and GPU-accelerated cracking, with ShinyHunters member 'Rey' detained in Amman and linked to the Oracle PeopleSoft exploit underpinning the operation.
#2 · security · scored 8/10
curl 8.23.0 patches 22 CVEs
curl is accelerating its release cycle to ship version 8.23.0 on October 14, 2026, with 22 security fixes including one HIGH-severity heap buffer overflow (CVE-2026-92392), affecting the widely-deployed infrastructure library.
#3 · security · scored 8/10
SonicWall CVSS 10.0 pre-auth SSRF
SonicWall patched CVE-2026-102255, a CVSS 10.0 pre-authentication SSRF in SMA1000 WorkPlace remote access appliances — the third perfect-score SSRF fix in WorkPlace this year — alongside three other flaws.
#4 · security · scored 8/10
Anthropic Claude security expansion
Anthropic expanded its Cyber Verification Program into three tiers for vetted security teams while simultaneously releasing Claude Haiku 5.5 at $0.10/$0.50 per million tokens with new reasoning capabilities and halved cache-read pricing for Sonnet 5.5, as Meta and Microsoft simultaneously reduce internal Claude usage in favor of proprietary tools.
#5 · security · scored 8/10
AI agent security failures multiply
Multiple reports converge on AI agent security failures: 13,000 screenshots leaked from 343 tech companies via agents bypassing approval controls, PoeLLM malware infected 3,000+ servers using adversarial poetry to hide C2 addresses, and Cisco Talos warned autonomous agents are now executing coordinated cyber attacks faster than human red teams.
#6 · ai · scored 8/10
Developer supply-chain npm attacks
Eight malicious npm packages downloaded 40,767 times delivered the Overlord RAT and movinlike stealer via postinstall hooks, in a sustained campaign by a single threat actor who has published 12 packages since August 2023.
#7 · security · scored 8/10
GitHub AI-agent security and outage
GitHub reported that one in three pull requests now involve AI agents (up from one in ten a year ago) and deployed a sub-2ms classifier to detect secrets in unstructured agent output, while also suffering a 10-minute multi-system outage affecting Git Operations, Pull Requests, Actions, Issues, and Webhooks.
#8 · security · scored 8/10
Country-code registry hijacks hit Google
Attackers compromised the .gh, .sl, and .as country-code domain registries between September 22–27 to obtain 12 unauthorized HTTPS certificates for Google and YouTube domains, which Google revoked and blocked via Chrome CRLSets.
#9 · security · scored 8/10
Atlassian critical flaw exploited quickly
A CVSS 9.3 unauthenticated arbitrary file-access flaw (CVE-2026-21589) in Atlassian Data Center products — spanning Bitbucket, Confluence, Jira, Bamboo, Crowd, Crucible, and Fisheye — drew active exploitation attempts within two hours of public disclosure details being published.
#10 · security · scored 7/10