hi bud buds,
a couple weeks ago, google announced a new dynamic view feature for their ai search, powered by gemini.
when you ask a question, instead of receiving a wall of text, gemini instead generates a mini interactive website, custom-built to answer your query.
for example, i have a japan trip planned for early 2026 with my wife, and need help planning activities. when i asked chatgpt for help, i received a long, multi-paragraph response back.
with gemini dynamic view, i received a bespoke webpage with images, maps, and interactive controls. drag a slider and my entire itinerary updates. ask a follow-up question and the entire ui reshapes itself.
when you drag the slider, the entire website / itinerary changes
this feels like the future of the internet: custom ai-generated ui that morphs around your questions, and is personalized based on your past conversations and stated preferences.
for trivial questions (“what time is sunset?”), this will definitely be overkill. but for complex workflows (travel planning, budgeting, home buying, medical or legal research), dynamic view is a paradigm shift. if ai can generate personalized answers within bespoke websites, then we’re heading toward a world where traditional third-party websites and applications become obsolete.
why go to a generic nerdwallet page when ai can spin up a personalized credit card comparison dashboard that knows your credit score, spending patterns, and travel habits?
why visit a generic travel blog when ai can assemble a bespoke travel microsite for your next trip, with tabs auto-generated around your top interests?
why open a recipe website when ai can generate a custom cooking site for you, with sections organized around your favorite cuisines, dietary constraints, and fitness goals?
in this brave new world, founders and product managers have to ask: what does it mean to build software when the ui is no longer yours?
the second-order consequences are still fuzzy to me, but a few possible hypotheses:
structured data becomes the new gold: whoever exposes high-quality, structured, real-time data to llms will be more valuable than whoever builds the prettiest ui
navigating websites is dead; declaring intent is the new normal: instead of users clicking around unfamiliar websites, they’ll simply state an intent (“help me refinance my loan,” “help me remodel my kitchen,” “help me compare crm tools”), and ai will generate custom experiences in real time
products will be split into two layers:
the data + logic layer, provided by third-party companies
the presentation layer, which is ai-generated, completely personalized, and entirely ephemeral
more zero-ui companies are coming: a new generation of tech companies will emerge whose core value is exposing clean + structured data to ai, while letting ai handle all the ux
in summary, in a world where the internet can rebuild itself per user, per query… our jobs may be shifting from building applications → building the underlying truth that applications are built on.
have a great week, y’all. i love you.
— dj
submit a release for the newpixels newsletter here.
big news! huge!
openai and perplexity announced ai shopping assistants, helping users research products and complete purchases directly in chat. assistants can recommend items using details they already know about a user’s location, past conversations, and preferences
personal faves
amazon announced an internal autonomous threat analysis system, which uses ai agents to find vulnerabilities in amazon’s security infrastructure and propose fixes. idk much about cybersecurity tbh, but this seems like a great use case for ai
facebook released nicknames for facebook groups, letting users post within groups under nicknames. i’m pretty bullish on pseudonymous social media to drive higher volume / quality engagement, so i’m a fan of this
character.ai released stories, letting users engage in controlled chats with their favorite characters, weeks after restricting chats for users under 18 due to mental health concerns. this feels like a reasonable, youth-friendly compromise
nice job guys
alibaba re-launched qwen, its chatgpt rival, surpassing 10 million downloads in its first week
alibaba released smart glasses with qwen ai in china
uber launched driverless robotaxis in abu dhabi
openai released voice mode in chat, allowing users to start voice conversations directly in the main chat window (instead of switching to a dedicated voice screen like before)
amazon is testing satellite internet service with a few early-access enterprises, in an effort to compete with spacex’s starlink
other llm news
anthropic released claude opus 4.5, offering state-of-the-art software-engineering performance at roughly 67% lower cost
microsoft released fara-7b, a 7-billion-parameter model that runs locally on pcs to automate complex ui workflows via mouse-and-keyboard control
idk tbh
kraken launched money app, offering a debit card with 1% cashback and automatic paycheck-to-defi routing… an attempt to position kraken as a traditional banking alternative. i’m generally more bullish on tradfi (traditional finance) players (e.g. robinhood) adding crypto features than crypto companies trying to expand into tradfi, since the former already have far greater existing customer distribution
that’s it for this week; thanks for reading. if you enjoyed this issue, please consider sharing it with a friend so i can get rich and buy my wife a new subaru.