Gemini-Powered Bug Hunting Lets Google Patch 1,072 Chrome Flaws in a Single Month

Skynet Chance (+0.03%): Demonstrated industrial-scale autonomous exploit discovery is inherently dual-use — the same capability that hardens Chrome lowers the barrier to automated offensive cyber operations, a classic pathway for AI systems to acquire real-world leverage. Defensive deployment partially offsets this, so the net increase in loss-of-control risk is modest.

Skynet Date (-1 days): Empirical confirmation that LLMs already outperform prior human-driven vulnerability discovery by an order of magnitude pulls forward the timeline for AI-driven cyber capability escalation and the offense-defense arms race it implies. The effect on pace is real but confined to one capability domain.

AGI Progress (+0.02%): This is verified real-world evidence that current models can perform sustained expert-level reasoning over large, messy codebases at a volume no human team could match, a meaningful data point on general capability transferring to economically valuable work. It is deployment scale rather than a new architecture, so it confirms rather than extends the trajectory.

AGI Date (+0 days): AI that finds and fixes its own software substrate's flaws feeds a self-improvement loop in tooling and code quality that modestly accelerates the pace of AI development itself. The gain is incremental and domain-specific, not a step change in the AGI timeline.