Steve Yegge says agentic coding is widening the attack surface faster than defenses
AI Engineer · youtube · 2026-07-21
Steve Yegge's talk says the real title is basically: be scared. The core argument is that agentic coding multiplies the software attack surface faster than teams can secure it.
- A bank security architect's rule of thumb frames the problem: if everyone ships 10× faster and defect rates hold, the vulnerable surface grows 10× too.
- Yegge highlights slop squatting: a model hallucinates a package name, someone publishes that exact package with a backdoor, and the build still passes.
- His answer is to split security into its own pass, with real tools like Snyk and Chainguard checking the work before and after the agent acts.
- He also points to an industry warning that open-source models may be able to autonomously hack production systems within months, not years.
More from coding & agent
- Devin adds e2b sandboxes for remote agent execution — badphilosopher · 2026-07-22
- Hermes Agent Refactoring Proposal: Decoupling via Event Bus and Monorepo Slicing — Promptmethus · 2026-07-22
- ty now reads Pydantic config keywords and field metadata — charliermarsh · 2026-07-22
- Pensar Launches AI Security Agent to Autonomously Discover and Patch 0-Days — andriy_mulyar · 2026-07-22
- ty adds first-class Pydantic support, including strict and lax field handling — charliermarsh · 2026-07-22
- Google launches Gemini 3.5 Flash Cyber for CodeMender, with limited access for governments — GoogleAI · 2026-07-22