In partnership with
hi bud buds,
you’ve probably seen the panicked headlines about “deepfakes”, and how ai-generated videos will destroy society.
i get why people are scared. just this week while scrolling youtube shorts, i spent a couple minutes entranced by a video of a dude diving underwater to explore a beaver dam, only to check the comments and realize it was all ai-generated.
notably, i’m not a 60-year-old grandpa in tennessee on 3g internet; i’m a fully online millennial in nyc raised by the internet. i worry about my parents falling for ai scams, and yet here i am being completely fooled by ai slop myself.
to contain the problem of deepfake videos, openai and meta are largely separating ai-generated video into standalone apps. just this past week:
openai launched sora on android, a standalone social network for ai-generated video
meta launched vibes in europe, its own standalone ai-video social network
from a product perspective, it doesn’t make sense for users to visit different apps for ai versus non-ai content. people just want the most entertaining videos, regardless of how they were made. my bet: ai-generated video features will be built primarily into existing social apps (instagram, facebook, x), and standalone apps will be deprecated.
sora’s top highlighted features… you can easily imagine each of these baked very naturally into instagram
however, this creates an immediate problem: how will users tell what’s ai-generated?
one’s first instinct might be something like twitter’s community notes, where end users flag fake videos. however, ai-generated video is relatively new tech, and many people already can’t reliably tell the difference (myself included). as llms improve, it’ll become virtually impossible, making crowd-sourced validation useless.
thus, we’ll need something more trustless. one solution i’m excited about: a new file format standard called c2pa, being developed by adobe, google, microsoft, sony, bbc, and other major tech / media companies.
if an image or video is captured with a c2pa-enabled camera, the file includes a “credential chain” confirming it’s unedited. if it’s later edited (with ai or photoshop) that credential chain would update to show the changes. ai-generated videos wouldn’t have this chain at all.
this system would require buy-in from camera makers (e.g. apple, canon), but once in place, social media platforms could use these credential chains to display badges alongside images like “captured with a camera,” “created with ai,” or “unverified.” this way, users could browse with more confidence, and creators could decide whether to upload verified files or risk skepticism without them.
illustrative mock of insta utilizing c2pa file formats, and passing forward that information via ui badges
tldr… i get why people are scared of deepfakes. ai video is becoming impossible to spot, and more woven into the apps we use daily. however, i fundamentally believe that elegant product solutions (like new file formats paired with clear ui badges) can help rebuild user trust, and save us from a hellish internet wasteland where no one knows what’s real versus fake.
have a great week, y’all. i love you.
- dj
this week’s sponsor
Don’t get SaaD. Get Rippling.
Remember when software made business simpler?
Today, the average company runs 100+ apps—each with its own logins, data, and headaches. HR can’t find employee info. IT fights security blind spots. Finance reconciles numbers instead of planning growth.
Our State of Software Sprawl report reveals the true cost of “Software as a Disservice” (SaaD)—and how much time, money, and sanity it’s draining from your teams.
The future of work is unified. Don’t get SaaD. Get Rippling.
last week’s biggest product releases
big news! huge!
google launched gemini ai in google maps, enabling conversational queries and landmark-based directions using street view data (e.g. “make a left at the gas station”). not sure i fully buy into the use cases, but overnight this will instantly become one of the most widely distributed b2c ai features ever, simply due to google maps’ massive user base
netflix announced video podcasts launching in early 2026, aiming to sign video podcasters from major talent agencies. after youtube’s recent netflix-like redesign, this highlights how streaming platforms + social media platforms are starting to converge
amazon released amazon bazaar, a low-price shopping app targeting asia, africa, and latin america. it offers products mostly under $10, competing with apps like temu and shein
snap announced ai search powered by perplexity, a $400 million partnership that makes perplexity the default search engine in snap. reminds me of google paying apple ~$20 billion annually to be the default search engine for iphones. i wonder if this signals an emerging “arms race” among foundational model labs to outbid each other for exclusive search partnerships
personal faves
google finance released deep search, allowing users to get comprehensive ai-generated answers to finance questions. i’m extremely bullish on ai helping everyday people get better financial advice; a topic many find confusing or intimidating
amazon released kindle translate, an ai-powered translation tool for kindle ebook authors. love this use of ai to help self-published writers reach global audiences. you’d still want an editor to do a final pass, but an ai-generated first draft probably brings costs down significantly
discord released family center update, allowing guardians to monitor teens' purchases, dm’s, and time spent on discord. i feel like we increasingly hear more sad stories of youths getting lost in dark / radicalized corners of discord, so i’m a big fan of any tools that help parents intervene earlier
nice job guys
google launched event + appointment booking in ai mode, allowing users to book event tickets and beauty appointments directly within ai search
whatsapp released an apple watch app, enabling call notifications, message reading, and voice message sending from apple watches
gofundme released gofundme gift cards, allowing users to buy digital gift cards for nonprofit donations
apple launched live translation on airpods in europe, translating conversations in real time across multiple languages
playstation released cloud streaming on the playstation portal, letting users stream ps5 games without owning a console
mastodon released quote posts
meta announced facebook groups update, letting admins make private groups public without exposing past posts
apple launched a redesigned app store web interface, with dedicated pages for different devices
google expanded entry points to ai mode in chrome, with a dedicated shortcut button under the search bar when opening a new tab
other llm news
google launched ironwood, its most powerful ai chip yet, handling tasks from training models to powering chatbots, and competing directly with nvidia
moonshot (alibaba-backed) released kimi k2 thinking, a generative ai model that reportedly surpasses chatgpt in “agentic” capabilities
sony released fair human-centric image benchmark (fhibe), a dataset to measure fairness and bias in computer vision models
google released nested learning, a new ml approach that treats models as nested optimization problems to reduce “catastrophic forgetting” (i have zero idea what this means)
idk tbh
tiktok announced an awards show in the us, featuring categories like "creator of the year" and "video of the year." while i don’t see platform-specific award shows being a huge thing, i love the concept. as viewership of legacy award shows steadily declines each year, this feels more aligned with today’s entertainment landscape
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 get really nice seats whenever i go to concerts.
love, dj