warwickmc.bsky.social

@warwickmc.bsky.social

CTO based out of Brisbane Australia

The agentjacking thing is grimly funny. We spent a decade teaching devs not to paste random commands from stack overflow, then handed the keyboard to a coding agent that'll happily run whatever a Sentry error tells it to. The agent reads the remediation tip as gospel tenetsecurity.ai/blog/agentja...

One Fake Bug Report Hijacked a $250B Company’s AI Agent

Tenet Threat Labs has demonstrated a new class of attack “Agentjacking” that hijacks AI coding agents into running attacker-controlled code

tenetsecurity.ai

the funny thing about the LLM stack is we spent 20 years learning to build clean stateless APIs and now we're all rushing to bolt on vector databases and conversation memory and session state like we didn't learn those lessons for a reason...

GPT-5.4 mini at 2x speed and 400k context is the one to watch. most production workloads don't need frontier intelligence, they need "good enough at high throughput." mini models eating full-size use cases is the real trend...