You're tuned in to Tech Beat, where we break down what matters in the world of technology.
OpenAI's GPT-six Astra has done something historians and cryptographers couldn't manage in over two centuries. An engineer at SentinelOne fed the model a Napoleonic-era cipher that had sat unsolved for two hundred and seventeen years, and within six hours, the AI had cracked it — without any prior knowledge of the specific code. It's a striking demonstration of how language models are beginning to outpace specialized human expertise in pattern recognition at scale.
That capability cuts both ways, and nowhere is that tension more visible than in the Hacker News community right now, where a thread asking how people are coping with high estimates of AI-driven catastrophe has drawn genuine, unguarded responses. Engineers and product leaders are describing real anxiety about autonomous agents proliferating into dangerous hands. It's a rare public window into the private dread of people who build these systems for a living.
And in the crypto world, a story that reads more like a heist film than a protocol audit. The DeFi platform Near Intents recovered nearly three point eight million dollars after its team publicly identified an attacker and issued a forty-eight-hour ultimatum to return the funds. The money came back. Whether that counts as security or just leverage is a question the industry hasn't fully answered.
That's your briefing for today. Keep surfing. Tech Beat out.
