hi bud buds,
last week was microsoft ignite, their big annual conference for developers and partners.
imo, the three most important announcements:
1) ai mode for excel + powerpoint: microsoft is adding an ai drawer directly inside excel and powerpoint. you can now clean data, build models, generate slides, and rewrite content without ever leaving the documents themselves
2) a context layer for copilot: work iq is a new “intelligence layer” that aggregates context about your role, writing style, team data, and company data. copilot will utilize this context to generate responses that are more accurate, personalized, and aware of your end goals
3) ai at the os layer: microsoft continues to bake ai directly into windows. file explorer now supports natural-language search and instant ai summaries; just select a file and ask copilot to analyze or explain it for you
imo, (3) is the most important news, because it shows where the puck is heading: ai is moving upstream, from the application layer to the operating system itself.
the same upstream trend is happening in hardware, as big tech companies rapidly integrate ai into the devices we use every day: dedicated copilot keyboard buttons on pcs, airpods with live language translation, meta smart glasses.
put together, the os + device layer is quickly becoming the dominant entry point for most ai workflows. this shift has profound implications for anyone building software. if you assume that 80% of user workflows in the future will start with ai assistants from apple or windows… where does your product fit into that world?
have a great week, y’all. i love you.
— dj
submit a release for the newpixels newsletter here.
big news! huge!
google launched ads in ai answers for the first time, labeled “sponsored” and located near the the bottom of responses
google released gemini 3, its latest foundation model with stronger reasoning and record benchmark scores. it’s getting rave reviews so far, which is a big deal after many felt gpt-5 underwhelmed
openai launched chatgpt group chats globally, supporting up to 20 participants at a time
zoox launched robotaxi rides in san francisco, available to select users in specific neighborhoods. zoox (amazon-backed) will begin competing head-to-head with waymo, just as tesla begins testing robotaxi service in arizona. there’s a robotaxi war brewing!! the future is here!!
google released antigravity, an “agent-first” coding tool with browser control, artifact-based proof of work, ide views, and agent manager views
personal faves
perplexity released comet (their ai browser) on android, with ai search and voice mode. i’ve been using comet on mac for the past couple months and i’m loving it. it feels like a ~15% productivity boost in my day-to-day work
letterboxd announced letterboxd video store, enabling in-app film rentals. absolutely love this. feels like both a seamless user experience, and a great new revenue play for letterboxd. win/win.
meta released content protection, helping creators automatically detect when their content is used by another account, and either add attribution or takedown the post
nice job guys
target launched a chatgpt integration, letting you shop target products, build baskets, and check out, all inside chatgpt. this feels like the future: every retailer will need to integrate natively with chatbots
openai released a version of chatgpt for teachers, with secure student data handling and school district workflows. to me, this foreshadows that industry-specific versions of chatgpt are coming for other major verticals (e.g. finance, healthcare, legal)
google released ai 'flight deals' tool, helping users find (and plan) cheaper trips via ai-powered travel search
google maps released gemini features including ai-generated “know before you go” tips (e.g. secret menu items, reservation tricks, parking intel)
coinbase announced a prediction market powered by kalshi, letting users trade event outcomes using kalshi’s regulated exchange rails
amazon released ai season recaps, generating theatrical-quality video recaps (including narration, dialogue, and music) for past tv seasons
tiktok released ai-generated content controls, letting users adjust how much ai content appears in their feed
whatsapp released an instagram notes-like feature, allowing users to post short text updates that disappear after 24 hours
snapchat released topic chats, enabling public discussions on trending topics while keeping profiles private from non-friends
apple released new app review guidelines, requiring apps to disclose and request permission before sharing personal data with third-party ai
youtube is testing an in-app private messaging feature, allowing users to share videos and chat directly inside youtube
google released ai image detection in gemini, letting users ask “is this ai-generated?” to identify synthetic images
uber announced sidewalk robot deliveries in europe
android released quick share compatibility with ios airdrop, enabling file transfers between android and iphone devices
spotify released playlist transfer, allowing users to import playlists from other services
roblox announced age checks for chat access, requiring facial verification for chat features
other llm news
google released generative ui, enabling ai chatbots to generate custom ui / visual layouts based on any query
meta released sam 3 / 3d computer vision models which can identify objects and people and reconstruct them in 3d
google released weathernext 2, an ai weather model with two-week forecasts and tropical storm predictions, aiding energy traders and other weather-dependent industries
google released nano banana pro, an image-generation model producing more detailed images, accurate/multi-language text, and stylized lettering
xai released grok 4.1 with lower hallucination rates, and stronger creative, emotional, and collaborative performance
idk tbh
meta released ai-powered morning briefs, a daily personalized recap of your facebook activity. ai-generated daily briefs are definitely the future, but not sure social networks will win this category, versus your day-to-day chatbot, or your email provider 👀
meta released multiplayer hyperscapes, allowing friends to hang out inside photorealistic digital replicas of real places you’ve built… big “visit my igloo in club penguin” type energy
x released about this account, showing location, username history, join date, and install method… supposedly to help users spot bots? instead of offloading this cognitive load onto users, shouldn’t x instead focus on stronger signup checks or better automated bot detection?