hi bud buds,
in june this year, google launched doppl, an ai-generated video app that lets you try on outfits virtually. you upload a few photos of yourself, shop for clothes, and if something catches your eye, you can generate a video of yourself wearing it.
screengrabs of doppl
a couple months later, openai launched sora, a tiktok-like app comprised entirely of ai-generated video. instagram also launched a feed inside reels made entirely of ai-generated video content.
for a minute, it felt like ai video was about to be the next big thing. however, since then, the results have been underwhelming.
doppl’s estimated downloads are less than 5k/month (per sensortower), a rounding error for google. google search interest for sora is down 75% from peak, per google trends. anecdotally, none of my friends who were early sora users still use it today.
i’ve written before that i’m skeptical text-to-video will ever see mass adoption. there’s definitely solid product-market fit for advertisers and content creators. but most people don’t want to make videos; they just want to watch them.
with all this in mind, i was surprised to see google this past week launch a new shoppable feed for doppl. now you can virtually try on an outfit, then click to buy the pieces directly from a merchant.
google is notoriously quick to kill experiments that don’t show explosive growth. so why keep building here? my guess: they’re betting that while ai video may not work for entertainment, it might work for ecommerce.
the “see yourself in this outfit” use case actually feels useful. online shopping has always struggled with this core friction: it’s hard to know how something will look on you. if doppl can help people find more clothes they love (or increase their order sizes by completing outfits with a few extra pieces), that’s a massive boost for clothing brands.
but i still think this is a feature, not a platform… like “see in your room” for furniture shopping sites, or size calculators on fashion sites. i wouldn’t be surprised if a yc b2b startup pops up to help clothing brands add “try this on with ai” features to their own websites. if someone reads this, builds this, and gets rich, please just give me a tiny kickback so i can take my wife out for a nice sushi dinner.
have a great week, y’all. i love you.
— dj
submit a release for the newpixels newsletter here.
big news! huge!
google released gemini 3 pro, a multimodal model that excels at document, screen, and video understanding, with major gains in visual reasoning for use cases like education, medical imaging, and legal analysis
google released managed mcp servers, enabling ai agents to easily integrate with google services (e.g. maps, bigquery) via managed endpoints
openai launched an apps for instacart and adobe within chatgpt, allowing users to order groceries and use photoshop directly inside chatgpt without switching apps
google launched disco, a gemini-powered tool that turns browser tabs into interactive apps, like transforming an academic article on a meteor system into a navigable 3d experience
uber released uber intelligence, letting advertisers combine their own data with uber’s to improve ad targeting across uber apps
anthropic released claude code in slack, allowing developers to delegate coding tasks directly from slack
personal faves
spotify released prompted playlists, letting users generate personalized playlists using natural-language prompts like “deep cuts from drake i haven’t heard yet”
spotify released spotify music videos, allowing seamless switching between audio and music videos while streaming songs. this is an awesome feature that i think will boost spotify's engagement kpis significantly
amazon released familiar faces, an ai-powered facial recognition feature for ring doorbells that enables personalized alerts like “mom visited 30 minutes ago”
hinge released convo starters, an ai feature that suggests personalized conversation prompts under profiles to help kickstart convos. i’m personally very bullish on ai for matchmaking / dating :)
nice job guys
rivian announced ai assistant, launching across all rivian evs by 2026 with voice control for climate, music, and integrations like google calendar
youtube released stablecoin payouts, allowing creators to receive earnings in paypal’s stablecoin
cursor released visual editor, a vibe-coding product for designers to shape web app interfaces using ai, with fine-grained controls, manual edits, and natural-language requests inside cursor
tools for humanity released a super app, with end-to-end encrypted chat, verified user identities, crypto payments, and direct paycheck deposits. this is sam altman’s crypto company… something to keep an eye on
apple and android announced iphone-android switching, enabling easier data transfer (contacts, messages, and photos) between devices, in response to eureopean regulatory changes
tiktok announced shared feed and shared collections, letting users share the same feeds with friends, and organize saved videos collaboratively
shopify released product network, allowing merchants to recommend products across other storefronts within shopify’s two-million-merchant ecosystem
instagram released your algorithm tool, a tool that lets users see and customize the topics influencing their reels recommendations
at&t released connected life, a smart home security service built upon google home devices
google released emergency live video, allowing android users to share live video with 911 responders during emergencies
meta released a facebook redesign, refocusing the app on friend connections while elevating marketplace and reels
google released real-time speech translation on any pair of headphones
apple released liquid glass rollback on lock screen, allowing users to adjust clock transparency via a slider to improve readability and customization
other llm news
meta announced avocado, a proprietary ai model expected in early 2026. this is a notable shift away from their historically open-source approach!
openai launched gpt-5.2, a three-mode model (instant, thinking, pro) designed to handle everything from quick writing to complex coding and long-context professional workflows
mistral ai released devstral 2, a new coding-focused model with a cli for automation, available in two sizes under different open-source licenses
runway released gwm-1 world model, a world model that simulates real-world behavior to train agents in robotics and life sciences
idk tbh
openai announced chatgpt’s ‘adult mode’ scheduled for q1 2026, offering adult content with age-prediction safeguards for users under 18. i feel this opens the company up to meaningful legal and reputational risk that may not be worth the tradeoff
gemini launched gemini titan, a cftc-approved prediction marketplace for yes/no questions on future events. i’m bullish on prediction markets as an industry, but uneasy as a citizen of the united states about how much they are lowering the friction to everyday gambling