🎬 What to know about the Sora discontinuation ↗
Sora’s web and app experiences have officially hit the off switch. Not a pause, not a “we’ll see” - discontinued.
Users can still export their content through the sunset flow, and any remaining Sora credits can be used for Codex. A peculiar little baton pass, to put it mildly.
💸 AI can cost more than human workers now ↗
The AI budget conversation is getting less glossy. Some companies are now spending so much on tokens, subscriptions, and infrastructure that AI can rival - or beat - human salary costs.
The tone has shifted from “look how advanced we are” to “wait, what’s the ROI?” A very corporate panic, but not an unreasonable one.
🧒 Canadian Province of Manitoba Says It Will Ban Social Media, AI For Youth ↗
Manitoba is moving toward banning young people from social media and AI chatbots. That’s a big policy swing, and not exactly a gentle one.
The pitch is child safety. The unanswered bit - and it’s a chunky one - is how enforcement would work in practice without turning the internet into a bouncer with a clipboard.
🇨🇳 Economists Rethink Chinese Forecasts as AI Fires Up Import Surge ↗
AI chip demand is pulling China’s import forecasts higher. Economists now expect imports to grow faster than previously thought, driven by companies hoovering up high-end hardware.
It’s another reminder that AI is not just software magic in the cloud. It’s ports, chips, supply chains, and a lot of spreadsheets sweating quietly in the background.
🛠️ Context decay, orchestration drift, and the rise of silent failures in AI systems ↗
Enterprise AI has a quiet problem: systems can look healthy while behaving wrongly. No red alerts. No dramatic crash. Just confident nonsense moving through workflows like fog in a warehouse.
The argument here is that classic monitoring is too shallow for AI. Teams need to track grounding, context freshness, and agent behavior - not just whether the service is technically “up.”
🧠 AI is frying our brains - here’s what leaders need to do about It ↗
AI was supposed to reduce busywork. Instead, it may be adding a new layer of cognitive load, with workers managing prompts, checking outputs, and doing more high-intensity thinking.
The fix suggested is almost stubbornly human: quiet time, better AI habits, and using the tools to sharpen thinking rather than just multiplying tasks. Simple, but not easy - classic.
FAQ
Why was Sora discontinued?
The article says Sora’s web and app experiences have been officially discontinued, not temporarily paused. Users can still export their content through the sunset flow. Remaining Sora credits can also be used for Codex, which suggests OpenAI is redirecting some user value into another product rather than ending access outright.
Can AI cost more than human workers?
In some companies, yes, AI can now rival or exceed human salary costs when token usage, subscriptions, and infrastructure are counted together. The issue is not just one expensive tool, but the total operating cost of AI across teams. That is why the conversation is shifting toward ROI, budgeting, and whether AI spending is generating enough value.
Why is Manitoba considering a ban on social media and AI for young people?
Manitoba is moving toward restrictions on youth access to social media and AI chatbots, with child safety as the stated reason. The article highlights a major unresolved question: enforcement. In practice, rules like this can be difficult to apply without intrusive checks or systems that tightly control online access.
How is AI affecting China’s import forecasts?
AI chip demand is pushing some economists to revise China’s import forecasts upward. Companies are buying more high-end hardware to support AI development and deployment. The article frames this as a reminder that AI growth depends on physical supply chains, ports, chips, logistics, and planning, not just software running invisibly in the cloud.
What are silent failures in enterprise AI systems?
Silent failures happen when AI systems appear to be working normally but produce faulty or misleading results inside workflows. There may be no obvious outage, crash, or alert. The article suggests that traditional monitoring is not enough because teams also need to watch grounding, context freshness, and agent behavior.