Loading...
Loading...
Nvidia put money on a safety-first superintelligence lab, new research mapped exactly which work tasks bend to AI first, and unions kept writing the rules into contracts. The money week, decoded.
No models shipped this week that made me rewrite anything. Instead, three stories about money and work quietly redrawed the map.
Nvidia backed Safe Superintelligence, Ilya Sutskever's lab with the famous one-product-one-mission pitch: get superintelligence right before shipping anything. A chip company whose revenue depends on everyone building fast, investing in the most go-slow lab in the industry, tells you something about where credibility now comes from.
When the arms dealer funds the safety inspector, the market has decided trust is a product feature.
Adobe published survey data showing generative AI is now standard equipment for creators rather than an experimental tab. The number that stuck with me was not adoption. It was that heavy AI users reported higher output AND more client revision cycles, because when volume goes up, review becomes the bottleneck.
OpenAI also published research mapping where ChatGPT stretches task boundaries inside real jobs, and unions responded by tightening contract language around workplace AI. The pattern in the data is clean:
AI can do the task. It cannot be responsible for the task. That gap is the entire future of employment.
As a CTO hiring for a student-led startup, this changed how I screen candidates. I care less whether you can write a sorting function from memory and more whether you can spot the bug an agent confidently introduced. That skill is the new junior developer job description, and most candidates have not figured that out yet.
Cheers, Yassen