After months of lobbying, Nvidia just got the go-ahead to start selling its H20 AI chips to China again. On the surface, it's a policy reversal. Underneath, it's a strategic move in the race to shape the future of AI and who controls it.

Chips power modern technology, and GPUs — a type of chip designed for high-throughput parallel processing — are especially well-suited for training large-scale AI models. The more powerful the chip, the faster and more capable the AI model — and Nvidia has the best.

As a result, it’s estimated that Nvidia’s GPUs command a staggering 80% of the AI chip market.

The U.S. government has now assured Nvidia that licenses to sell its H20 chips — designed to meet earlier export restrictions — will be granted. Sales had been abruptly halted in April, costing Nvidia $2.5 billion in lost revenue and triggering a $4.5 billion inventory writedown.

Now, Wall Street sees Nvidia recouping up to $15 billion from China in the second half of the year, with full-year China revenue potentially reaching $20 billion. Not a bad rebound for a chip that almost didn’t leave the warehouse.

The U.S. government often operates like a black-box model with limited interpretability, but here are some educated guesses:

Nvidia CEO Jensen Huang is balancing a delicate equation: keeping Washington happy while protecting Nvidia’s stake in China. In an interview with CNBC Huang explained:

Despite facing export restrictions, Nvidia just became the first company to hit a $4 trillion valuation. Its stock rose 4.5% on the news of reentering China—a $50B AI market in the making, and one that’s too big to ignore.

Not a fan of making phone calls? Or just too busy to sit on hold? You’re not alone—and now, in some cases, you no longer have to make those calls at all.

📞Google’s new AI calling feature, now live in Search for everyone in the U.S., lets Gemini do the talking for you. Whether you're checking prices at a dry cleaner or booking a pet grooming appointment, it can call businesses on your behalf to check prices, availability, and more.

Here’s how it works:

✅ It’s especially handy if you want to reach out to multiple businesses at once to compare prices or availability—like calling three auto shops for quotes. Or when your to-call list starts piling up and you want backup.

And great news - the feature is free for everyone, but Pro subscribers do get higher usage limits. It’s a perfect example of offloading some of those more tedious tasks to AI.

If you’re into finding new ways to stay organized, ClickUp is worth a look. The company has grown fast (now valued at $4B), and whether you’re working solo or with a team, it can help make managing your workload a little smoother.

Most AI tools start from scratch every time. ClickUp Brain already knows the answers.

It has full context of all your work—docs, tasks, chats, files, and more. No uploading. No explaining. No repetitive prompting.

ClickUp Brain creates tasks for your projects, writes updates in your voice, and answers questions with your team's institutional knowledge built in.

It's not just another AI tool. It's the first AI that actually understands your workflow because it lives where your work happens.

Join 150,000+ teams and save 1 day per week.

Try ClickUp Brain Free.

Delta wants AI to handle your airfare. By year’s end, 20% of its ticket prices will be algorithmically set (up from 3% now) thanks to a partnership with Israeli startup Fetcherr. The goal? Fully personalized, real-time pricing tailored to each traveler. It’s a revenue win for Delta, but privacy advocates and lawmakers are raising red flags. Critics say it’s likely to have a negative impact on consumers, with each person being charged how much the system thinks they’d be willing to spend.

YouTube updated its monetization policies on July 15 to crack down on “inauthentic” content, such as mass-produced, low-effort videos often generated by AI. While the platform frames it as a minor clarification of existing rules, it’s clearly aimed at curbing the rise of AI slop—voiceover slideshows, fake news clips, and auto-generated true crime stories. YouTube says reaction and clip-based videos are still safe (can’t get rid of all the leeches), but the message is clear: creators churning out spammy AI content shouldn’t expect to cash in.

After stage 4 skin cancer treatment left her with severe joint pain, one woman’s doctors recommended a treatment to ease the side effects—but her insurer, Primera Blue Cross, denied the request, calling it “experimental.” Two failed appeals later, she teamed up with a doctor using an AI tool to draft a 20-page appeal letter in record time, which they also sent to public officials. The result? Approval within 48 hours. Primera later claimed the denial was due to a “processing error,” not an intentional refusal of care. It’s how you win: an internet’s worth of facts, delivered instantly.

Do AI tools actually speed up task completion for seasoned software developers? The team at METR set out to find out. They conducted a randomized controlled trial with 16 veteran open-source devs, each working on tasks from projects they knew inside and out.

For some tasks, they used AI tools like Cursor Pro and Claude 3.5/3.7 Sonnet; for others, they went old school, coding without any help. Before starting, the developers predicted AI would cut their time by 24%. After finishing? They still thought it saved them 20%.

🤔 But the data told a different, and surprising, story. Tasks done with AI actually took 19% longer to complete. That’s not just off—it’s the opposite of what developers, economists (who guessed 39% faster), and ML experts (38%) predicted. The researchers investigated 20 possible explanations for the slowdown, from project complexity to tool familiarity, but the effect held strong no matter what adjustments were made.

The takeaway: It’s not all that surprising that highly experienced developers working on familiar projects were faster without AI. While future advances could change this, today’s real AI advantage lies in enabling a broader range of people to code—namely those who previously couldn’t.