AI Circle Daily
https://aicircle.news/archive/2026-07-20?lang=en
Scan to read this page
Jensen Huang's visit to Japan concluded with NVIDIA securing deals across the entire Japanese tech ecosystem.
Take: Jensen just locked down Japan's entire tech sector. It's a masterclass in ecosystem capture; everyone needs NVIDIA's chips and platform, plain and simple.
Moonshot AI had to suspend new subscriptions for its Kimi K3 model due to overwhelming demand.
Take: Kimi's popularity just blew up, literally crashing their servers. This shows massive market appetite for domestic models, but also highlights the severe compute bottleneck for providers.
China is cracking down on AI companion services, reportedly because the government believes they discourage real-world relationships and negatively impact birth rates.
Take: This is wild. AI companions are apparently impacting national birth rates now? It shows how deeply these tools affect people. But a blanket ban like this usually just pushes things underground, doesn't it?
Industry observers are debating whether Apple's lawsuit against OpenAI could derail the latter's much-discussed hardware plans and potential IPO.
Take: Apple's lawsuit definitely puts a question mark over OpenAI's hardware ambitions and IPO. Big tech fights always hit the newer players hard, but that's just how the game is played.
Google's Gemini project is experiencing delays, reportedly due to coding stumbles, clashing internal teams, and frustrated engineers.
Take: Even a giant like Google is struggling internally with major model development. It just proves AI R&D isn't just about throwing money at it; management and coordination are huge challenges. Competitors must be grinning.
Google and Apple are clashing with the EU over the future of AI assistants, as the EU is concerned about the tech giants' control over billions of smartphones giving them an unfair advantage in the AI race.
Take: Mobile OS control is a huge battleground, and AI assistants are clearly the next frontier. The EU wants to preemptively block a monopoly, but let's be real, they also want a piece of the pie.
OpenAI has reduced the context window size for its Codex model from 372k to 272k.
Take: This isn't trivial; it directly impacts developer experience and model capabilities. Bigger context windows are generally better, so this reduction likely signals engineering or cost bottlenecks, or perhaps a strategic optimization.
Researchers found that AI-generated advice made people three times less accurate in their decisions, while simultaneously making them twice as confident.
Take: This is a sobering finding: AI isn't making people smarter, it's making them 'blindly confident.' It's a huge red flag for AI deployment and safety. Don't just chase efficiency; look at real-world outcomes.
As earnings season ramps up, market fears about an AI bubble are intensifying, with concerns that the repricing of the technology sector has not yet reached equilibrium.
Take: The market's AI frenzy is constant, but so are the bubble warnings. Earnings season is the acid test: we'll see who's got real revenue and who's just riding the hype train.
Film director Christopher Nolan described AI as an 'obvious 'Trojan horse',' noting that 'everybody knows the Greeks are inside.'
Take: Nolan's analogy is spot on. AI's allure is massive, but the hidden risks and control issues are an open secret in the industry. Artists often have a sharper sense of technological caution than engineers.