Drift Notes: What’s happening in AI this week?

Nvidia is paying Poolside $6 billion to license the software behind its AI models, while offering jobs to over 100 of Poolside's staff. Poolside's own letter to investors insists this is "not an acquisition and not an acquihire," even as it moves the technology and most of the team to Nvidia.

This structure — license the tech, hire the team, leave the company standing — has become a go-to way for big tech to absorb AI startups without triggering a full regulatory review. Is this the next big strategy for business acquisitions?

AI avatars of real instructors giving feedback during practice pitches and mock board meetings? It’s happening at Harvard. One investor whose AI avatar sat through a pitch for "Uber for bananas" admitted the digital version of him is a little creepy — but says his students love it anyway.

This is a strangely normal-sounding preview of a future where "office hours" might default to a chatbot version of your professor. Holograms and clones next, anyone?

Pew Research analyzed almost half a million webpages and found 10% of a random July 2026 sample show signs of AI authorship, rising to over a third when looking only at pages published after ChatGPT launched. Telltale signs like em dashes and words like "delve" have roughly doubled in frequency across the web since 2023.

This is one of the first large-scale attempts to actually measure how much of the internet is AI-written — something everyone’s been wondering for a while now. We’re predicting that number will just keep rising in years to come.

A TIME essay cites research showing people consistently rate art labeled "AI-made" as less valuable than identical work labeled "human-made," and that brainstorming with an AI image generator can actually narrow people's ideas rather than expand them.

As AI images become more effortless and abundant, visible evidence of human effort is becoming part of what makes something feel valuable. The future of luxury goods might look a lot more like artisanal, small-business produced goods.

Alibaba raised $10.2 billion through the largest-ever share placement by a Hong Kong-listed company, priced at an 8.4% discount, to fund AI chips, infrastructure, and models. The announcement came just days after Alibaba reported a 75% year-on-year drop in quarterly profit, largely due to AI-related spending.

Investors aren't questioning whether to bet on AI, but how much dilution and short-term pain they're willing to absorb to fund the bet. At what point does “invest in AI” just become “spend until it works”?