Spot on. The drop in stock for the big players feels like a massive overreaction. Finding a vulnerability is the easy part; it’s the 400 meetings, compliance hurdles, and 'will this break prod?' anxiety that actually slows security down.
If AI can automate the patch, great—but as you pointed out, the attackers get that same speed boost for free. We aren't looking at the end of cyber; we're looking at a world where 'security by obscurity' is officially dead. Good luck to the SAST-only shops, though... that lunch is definitely getting eaten.
This is a good take IMO. There is still a lot of infra, and other layers anthropic isnt really touching yet. But I do think Claude Code Security will be pretty effective at taking down some of the SaaS cyber companies.
accelerator credentials are doing load-bearing work they weren't designed for.
the trust intangibles you're describing here (reliability, auditability, "partners they can trust when something breaks”) are exactly the layer that failed twice this week with litellm and delve. speed didn't make the code worse. it made the blast radius of misplaced trust worse
Spot on. The drop in stock for the big players feels like a massive overreaction. Finding a vulnerability is the easy part; it’s the 400 meetings, compliance hurdles, and 'will this break prod?' anxiety that actually slows security down.
If AI can automate the patch, great—but as you pointed out, the attackers get that same speed boost for free. We aren't looking at the end of cyber; we're looking at a world where 'security by obscurity' is officially dead. Good luck to the SAST-only shops, though... that lunch is definitely getting eaten.
This is a good take IMO. There is still a lot of infra, and other layers anthropic isnt really touching yet. But I do think Claude Code Security will be pretty effective at taking down some of the SaaS cyber companies.
accelerator credentials are doing load-bearing work they weren't designed for.
the trust intangibles you're describing here (reliability, auditability, "partners they can trust when something breaks”) are exactly the layer that failed twice this week with litellm and delve. speed didn't make the code worse. it made the blast radius of misplaced trust worse