The Limits of the Walls
Six sessions handing a live home network to an AI agent. Part 3: the memory that made it work, the division of labour, and what I'd never let it touch.
Tag: security
11 posts
There Is No API
The router I was reconfiguring has no public API. Part 2: driving the private endpoints the web UI calls — and the wall the agent could not script past.
Don't Saw Off the Branch
I let an AI agent rebuild my live home network over Starlink — no console, no documented API. Part 1: building three backout paths before a single write.
The Fight
Eight Minutes #3: a packet capture taken mid-attack, Google's own audit logs, eight abuse reports, and the passkey that ends the story.
The Fall
Eight Minutes #2: inside the attacker's session — the live relay, the Binance pivot, and the safety nets that fired after I'd already been beaten.
The Trap
Eight Minutes #1: the vishing call, the case number, and a lure genuinely signed by Google — why every check I'd been taught to run came back green.
Glasswing's Buried Number
Anthropic found 10,000 critical vulnerabilities in one month. Fewer than 1% are patched. The announcement buried that figure — and what it means.
Robot Dogs Are a Security Nightmare — And We Can Prove It
Eight CVEs. A wormable Bluetooth exploit. An encrypted backdoor to Chinese servers. And police departments buying them anyway.
Multi-Agent Safety Is the New Supply Chain Security
Multi-agent AI systems reproduce software supply-chain failure at the cognitive layer. The security playbook transfers.
Zero-Build Web Development
What happens when you build a three-zone operations platform for a martial arts club with no framework, no build step, and no npm.
120 Models, 18,176 Prompts: What We Found
120 models, 18k prompts, 5 attack families. The raw compliance numbers — and why calling them "attack success" needs a demonstrated refusal floor.