INT21: Agent swarms + 1,700 tok/s inference to outrun 27-second AI network attacks
bingxu_ · x · 2026-10-27
Security firm INT21 argues ultrafast inference and collaborative agent swarms are needed to counter AI-accelerated attacks: intruders spread across networks in 29 minutes on average, as fast as 27 seconds, while median detection time is 14 days; AI-enabled attacks grew 89% last year (CrowdStrike). INT21's agent swarms built a runtime for DeepSeek v4.1 Flash in one hour, hitting 1,700+ tokens/sec on 8×B200 — up to 63% faster than the officially supported option. The piece doubles as company outreach, but its analysis of agentic ransomware (e.g., Sysdig's documented attack issuing 600+ commands in 31-second recovery loops) and defense architecture is substantive.
More from Infra
- Ada's agent retrieval on Cohere Embed: 77ms median latency, 4x smaller vector storage — aidangomez · 2026-10-28
- Neywa Labs exits stealth with local AI engine and deployment infra built for Apple Silicon — yb2698 · 2026-10-28
- AMD MI355X hits 2,100+ tokens/sec per user on DeepSeek V4.1 Flash inference — JiaZhihao · 2026-10-28
- Strata DualGPU runs two mismatched NVIDIA GPUs for 5.6x faster local decoding — GabGarrett · 2026-10-28
- Lenders price $9 billion of junior debt for the AI buildout above 10% — YvesMulkers · 2026-10-28
- NVIDIA's first Rubin NVL72 Kimi K3 submission hits 850 tok/s/user at 2.8T params — scaling01 · 2026-10-28