Claude Code leaked: 512,000 lines expose AI’s security illusionAnthropic accidentally leaked Claude Code’s entire source code this week. Nearly 2,000 TypeScript files, over 512,000 lines of code, dumped…Apr 4Apr 4
The skill half-life: why understanding it will keep AI from taking our jobsThe AI debate is loud. But perhaps, we are asking the wrong question?Mar 7Mar 7
When AI tools start replacing security teams, who wins?Claude’s Scanner Crashes Security Stocks: 24 Automation ThreatsFeb 25Feb 25
Published inGoPenAIWhen security theater becomes a business modelThe question is not whether AI agents are secure. They are not. The question is whether we care enough to stop deploying them until they…Feb 14Feb 14
Published inGoPenAIThe code editor you trust just became a e-trojan horseWe’ve spent years worrying about whether AI will write secure code. Turns out, the real question is whether the AI-powered editor itself is…Feb 13Feb 13
Published inGoPenAIAI writes the code. AI hacks it. And human watches.Software development is entering a new normal: AI writes the code, another AI tries to break it, and humans are increasingly just watching…Feb 9Feb 9
The hidden cost of vibe coding: My 2-week experiment w/Claude CodeFor many developers, having an AI agent actually hold executive command over our computer is pretty awesome! We are seeing a flood of…Jan 31Jan 31
When our children are raised by bots…When our children are raised by bots, they will practice empathy with entities that cannot bleed.Dec 6, 2025Dec 6, 2025
Published inGoPenAIAI’s dangerous confusion between belief and factToday, I came across an interesting paper from Suzgun et al. (2025) published at Nature Machine Intelligence that suggests LLMs have a…Dec 5, 2025Dec 5, 2025
“I think, therefore I am” and the skinnier of the “I”When René Descartes wrote “Cogito, ergo sum” (I think, therefore I am) in his 1637 work Discourse on Method, he was trying to find an…Nov 14, 2025A response icon1Nov 14, 2025A response icon1
