/Hive Five #280 · Jun 26, 2026
The Untrainable
/The Bee's Knees 🐝
The headline picks from this issue. The full issue has more.
-
Wiz Red Agent + Claude Opus Found 3,000+ Critical Risks Every Week
Wiz's AI-powered attacker scans over 150,000 production web apps and APIs weekly, using frontier Opus models to chain multi-step exploits that traditional scanners can't see. Zero false positives, every finding comes with proof of exploitability.
Read : Wiz Red Agent + Claude Opus Found 3,000+ Critical Risks Every Week -
The Gift That Keeps Giving: Exploiting Git Integrations in Cloud Services
Four vulnerabilities found in Google Cloud Platform services show how git integration becomes a powerful exploitation primitive. Includes RCE via path traversal in git config, argument injection, and a CVSS 10.0 cross-tenant access chain chaining npm symlinks through JGit.
Read : The Gift That Keeps Giving: Exploiting Git Integrations in Cloud Services -
The Untrainable: What AI Can't Commoditize
Sarah Guo maps the landscape where AI eats the measurable and leaves behind the untrainable corner: work whose correctness exists only in private data, behind security reviews and customer trust. The absorption frontier keeps rising, but value keeps sliding toward what models can't reach.
Read : The Untrainable: What AI Can't Commoditize -
Has AI Already Killed How-To Nonfiction?
Tim Ferriss shares his own catalog sales data showing an 80% decline from 2022 to 2026, with the steepest drops after LLMs went mainstream. Self-help publishing is the canary in the coal mine, and podcasts, courses, and advice blogs are next.
Read : Has AI Already Killed How-To Nonfiction? -
Do Things You'll Love Yourself For
David Cain reframes self-discipline as a loving relationship between Present Self and Future Self. Shine the sink before bed, freeze the leftover pizza before a trip: acts of love for the person who inherits your choices tomorrow.
Read : Do Things You'll Love Yourself For