Security for the AI era
Today, we’re joined by Sherwood Callaway, founder of Sazabi, an AI-native observability and security platform for fast-moving engineering teams, and creator of cve.now, a free vulnerability feed built for agents.
Sherwood watched October 2026 become one of the worst months in software security history. He calls it Black October. Around 17,000 new CVEs were published in a single month, four times the year before, as AI made finding and exploiting vulnerabilities cheap, fast and within reach of almost anyone. Most security tools still depend on someone predicting each attack in advance and writing a rule to catch it. Sazabi drops the rules: send every log to AI and let it detect, investigate and explain what’s wrong in real time.
Sherwood and Kaz Tamai cover what happened in Black October, how cve.now narrows 100,000+ CVEs a year to the ones that touch your stack, whether exploit code should be public when AI can write it anyway, how a team of five engineers ships hundreds of PRs a day, and what a 10-person startup should do about security this month.
Subscribe to Frameless wherever you listen to podcasts for more conversations about the raw ideas, early convictions, and bold decisions behind standout companies.
00:36 Black October: what happened
02:15 What is a CVE?
02:55 How much of this is AI?
05:06 cve.now
07:27 Which CVEs should a startup worry about?
08:53 Can the CVE system keep up?
10:10 Should exploit code be public?
12:54 What Sazabi does
14:18 Monitoring without monitors
17:37 Observability and security
22:35 What existing security tools miss
23:58 Bring your own cloud
25:57 Five engineers, hundreds of PRs a day
28:07 A security checklist for a 10-person startup
33:11 Beating Datadog


More Episodes
.webp)
Podcast
Latest


