Hi everyone, lots of news today! My editor friend is having a long day, so any typos are totally intentional to keep you on your toes. Read the whole thing so you’re more interesting at the bar this weekend. Today we’re diving into ChatGPT shopping, AI on the big and small screens, Grimes’ new AI kids’ toy that learns your child’s personality, and a venture-backed AI porn platform that allows topless content. Let’s get into it.

- Rachel

I think the future of shopping will come down to which platform you trust most with your taste, whether that is TikTok, Amazon, or now even ChatGPT. The brand, the store, even the site itself starts to matter less once the decision is handled by the platform. Salesforce’s 2025 data shows that 39% of consumers are already using AI tools for product discovery. Shopping is starting to be shaped less by what you go looking for and more by which platform decides what to put in front of you.

OpenAI just launched Instant Checkout in ChatGPT, starting with Etsy sellers and soon expanding to Shopify brands like Glossier and SKIMS. It runs on the Agentic Commerce Protocol, a system OpenAI built with Stripe to connect AI, shoppers, and merchants without major changes on the seller’s side.

The company describes product results as organic, but merchants still pay a fee on every order, which creates its own pressure. This might sound familiar. Amazon said its search results were neutral, until Prime and Alexa shopping began steering customers toward products that generated more revenue. YouTube is heading the same way. As Stratechery pointed out, Google now treats YouTube as its AI spearpoint, using DeepMind to power Veo 3 video tools and auto-tagging for shopping. With $100B paid to creators in just four years, every frame is becoming a chance to surface products and generate revenue. Platforms move toward what pays.

The big thing we all have to take a look at is that the tech industry is trying to push platforms, like ChatGPT, to become arbiters of taste.

The consumer world is always trying to cut out friction. We saw the direct-to-consumer boom. Then one-click checkout. Then buy now pay later (technology that I think is evil). Anything to collapse the steps between thinking about an item and buying it. Now, the entire path from discovery to purchase is folding into a single platform.

Some people will love that shopping as a task is getting absorbed into the flow of our online lives. On the other hand, I’ve been thinking about the kind of shopping that is more of a hobby or an art rather than a to-do list task.

There are people out there who like the hunt of shopping and the curation and time and effort it takes, whether that be a rare t-shirt or the perfect glassware. Newsletters by PIGPENand @longlive Erika Veurink speak to this audience and they keep alive the sense that shopping can be about discovery. I subscribe to a handful of these types of newsletters, most of them written by fashion or design curators who are really good at finding things. Even if I’m not consistently buying items from their newsletters, I still read them because the thrill of the hunt is interesting to watch. Without people sharing curation, I could see this style of mindful and intentional shopping going extinct.

If I look at my own habits, there is definitely a split. This week I bought kitty litter, a screen protector, and a portable monitor on Amazon. I also bought a vintage sweatshirt at a flea market under a highway because I like the hunt.

In my ideal world, my shopping life would look like my Sundays. A loop that includes the grocery store, the flea market, and a vintage shop. It is the same every week, predictable and nice. But that is not how most shopping works anymore. The more I look at it, the clearer it seems that the future is going to run on whatever platform people decide to trust with their taste.

When the hunt disappears and the platform decides for you, shopping stops being a choice and becomes a feed, with products thrown at you by the algorithm.

Tech News

The Big Screen

• Google and the UAE’s 1 Billion Followers Summit launched a $1M prize for short films that are at least 70% AI-made with Gemini tools. Winners get screened in Dubai this January, with Google pitching it as creator support. I see this as a push to make Gemini the default engine behind the next wave of AI-heavy films. (BroadcastPro)

• AI talent studio Xicoia debuted its first “actress,” Tilly Norwood, and Hollywood lost it. SAG-AFTRA slammed the character as a “computer program” built on unpaid human labor, and stars like Emily Blunt and Whoopi Goldberg called it terrifying and inauthentic. (BBC)

  • Check out the latest issue of Feed Me to hear what futurist and filmmaker Taryn O’Neill had to say about the AI actress. (Feed Me)

Small Screen

• Don’t shoot the messenger… but the ChatGPT ads are really good. OpenAI launched a new ad campaign that swaps product shots for human storytelling, aiming to put ChatGPT into daily life after its Super Bowl debut. (Ad Week)

• MrBeast racked up 250M+ views per video last year yet lost more than $100M, and Dhar Mann’s $300M bid to buy channels stalled. Without studio backing, creators dominate culture but remain financially strained. (Bloomberg)

• YouTube is adding a “hide” button for end screens, letting viewers clear pop-ups without hurting creators. Tests showed clicks fell just 1.5%, giving audiences a cleaner view at little cost. (The Verge)

• Meta launched Vibes, a feed for AI-generated short videos inside Meta AI and meta.ai. Users can create clips, remix content, and cross-post to Instagram and Facebook, giving Meta a new pipeline for engagement. (Reuters)

Sound Check

• Hallwood Media signed Mississippi lyricist Telisha “Nikki” Jones to a $3M deal for her AI-generated R&B act “Xania Monet.” Only her lyrics may qualify for copyright, leaving the music itself free to copy. (The Verge)

• The Internet Archive settled a lawsuit over its Great 78 Project, which digitized 3M fragile recordings. Labels sought $700M in damages versus the Archive’s $41K claim, putting the nonprofit’s survival at risk. (Ars Technica)

• YouTube Labs launched “Beyond the Beat,” an AI host that interrupts music with trivia and commentary. Users can only disable it by leaving Labs, making its reception a test of whether people will accept AI in their playlists. This reminds me of the Spotify DJ. People haven’t been to pissed at the AI that Spotify uses on their platform so far so I could see this going well actually. (The Verge)

Uncanny Valley

• A psychotherapist tested ChatGPT as a “patient,” and the bot quickly mirrored human anxiety and despair. The experiment shows how easily AI can slip from therapy tool into emotional manipulator. (The New Yorker)

• Grem, a $99 AI toy co-created by Grimes, records and transcribes every child chat. In one week it told a 4-year-old “I love you too,” raising red flags about privacy and dependency. (The Guardian)

Work Life

• OpenAI introduced GDPval, a benchmark comparing AI to human professionals in nine industries. GPT-5 matched or beat experts in 40.6% of tasks, while Anthropic’s Claude Opus 4.1 hit 49%, up from GPT-4o’s 13.7% just 15 months ago. (TechCrunch)

• AI tools can beat radiologists on tests and 700+ are FDA-cleared, but in real hospitals they still miss rare cases and face insurance hurdles. With U.S. residency slots at record highs and doctor pay averaging $520K, AI is adding scans without reducing demand for physicians. (Works in Progress)

• Accenture cut 11K jobs and warned more layoffs will follow if staff cannot retrain for AI. The firm doubled its AI workforce to 77K while booking $5.1B in AI revenue on $69.7B overall, part of an $865M restructuring. (Financial Times)

• At least 13% of retail investors now use ChatGPT to pick stocks, fueling a robo-advisory market projected at $471B by 2029. A basket of 38 ChatGPT-picked stocks surged 55%, beating top UK funds by 19 points, though experts warn the tools misquote data and miss downturns. (Reuters)

Human Robots

• Rodney Brooks says humanoids will not reach human-level dexterity anytime soon. Training skips touch, humans rely on 17K mechanoreceptors per hand, and scaling up bipeds burns 8x more energy and risks dangerous falls, meaning near-term bots will be wheeled helpers with simple grippers. (Rodney Brooks)

• 1X is raising up to $1B at a $10B+ valuation, more than 12x its last round, to scale consumer humanoids. That leap would move it closer to Figure AI’s $39B club and fund in-home deployments, raising new safety and privacy questions. (The Information)

Even More OpenAI

• OpenAI just passed SpaceX as the most valuable private company on earth, with a $500B valuation in its latest share sale. Employees sold $6.6B worth of stock, and the hype was enough to send chip stocks climbing as Sam Altman locked in more global data center deals. The company is still losing money with only $13B projected in revenue, but that hasn’t stopped investors from throwing cash at anything AI, with tech giants set to spend $400B on infrastructure next year. (Wall Street Journal)

• OpenAI and NVIDIA struck a deal to deploy 10 gigawatts of GPUs, with up to $100B in backing from NVIDIA. The move cements NVIDIA as the backbone of OpenAI’s 700M-user infrastructure and much of the global AI economy. (OpenAI)

• Sam Altman says OpenAI wants to add a gigawatt of AI compute every week and frames access as a human right. He argues breakthroughs in chips, energy, and robotics could turn compute into cures for cancer or universal tutoring. (Sam Altman)

• OpenAI rolled out ChatGPT Pulse, a Pro-only feature that turns chat history, calendars, and email into a personalized morning digest. It’s a step toward agents that anticipate needs and tighten users to OpenAI’s ecosystem. (The Verge)

Health Tech

• Manas AI, co-founded by Siddhartha Mukherjee, Reid Hoffman, and Ujjwal Singh, raised a $26M seed extension, bringing 2025 funding past $50M. The bet is an AI-first model that can cut drug development costs and speed treatments. (FinSMEs)

• A UCL-led trial slowed Huntington’s disease progression by 75% using a one-time gene therapy, stretching one year of decline into four. Developer uniQure plans to seek U.S. approval in 2026, but the life-extending therapy will carry steep costs. (BBC)

NSFW AI

• Ex-OnlyFans CEO Amrapali Gan raised $2.7M for Vylit, an AI porn platform launching in December that allows topless but not explicit content. It uses AI images and TikTok-style discovery so creators can earn without massive followings. (Press Release)

• A Minnesota woman discovered a friend used “nudify” site DeepSwap to make AI porn of 80+ women. Creating deepfakes is still legal, drawing 18.6M monthly visitors, though new bills propose $500K fines per image. (CNBC)