hi bud buds,
this week, tiktok began testing footnotes, their version of x’s community notes. it allows users to attach context or corrections to videos, and then uses social consensus logic to determine whether people with different viewpoints broadly agree. if they do, the note gets surfaced.
the ultimate goal: crowdsource fact-checking.
meta also started testing its own community notes feature a few weeks ago, after years of paying legions of fact-checkers to do the same job.
the rapid adoption of community notes as big tech’s new default feels unsurprising from a cost and liability pov. if you’re a social platform, it’s infinitely cheaper to outsource fact-checking to users than to employ full-time staff. and if something harmful slips through? not your fault — it’s the community’s fault.
i’ve noticed that community notes fit into a broader trend of tech platforms offloading more labor onto users. at grocery stores, you scan your own items. at amazon lockers, you’re the delivery person. on airbnb, you strip the beds and take out the trash. tech platforms boost profitability by finding clever ways to get end users to do more.
as far as i can tell, this trend doesn’t have a clear solution or better alternative. maybe we’ll see a rebound — a new wave of startups offering high-touch, concierge-style service (like superhuman’s mandatory human onboardings).
either way, i don’t think i’m alone in noticing the shift. on a recent group trip, my friends pushed for a hotel over an airbnb. they wanted a more comfortable stay — one where they didn’t have to take out the trash.
until next week…
xxo,
dj
---
last week’s product launches
hell yeah, cool stuff
instagram introduced blend, a shared reels feed you build with friends. love this idea; really fun execution of microsocial (which i’m very bullish on), which builds elegantly on instagram’s existing infra
notion launched notion mail, an ai-powered gmail client integrated with notion. as a happy user of notion calendar, i’m stoked for this
tiktok is surfacing reviews in their comments tab for videos with tagged locations, allowing users to easily add and browse reviews. super interesting to see tiktok vying to becomes the google maps killer for gen z / alpha
other big brain moves
openai launched flex processing, a cheaper, slower option for o3 and o4-mini models, cutting costs for non-production tasks by up to 50%. super smart flex pricing move to optimize supply <> demand
waymo and uber are launching robotaxis in atlanta this summer
uber will handle ride-hailing and fleet operations (charging, maintenance, and cleaning of autonomous vehicles)
waymo monitors the tech and rider assistance
love this partnership model for waymo; allows them to focus fully on the (astronomically) lucrative software layer, and let uber focus on the annoying day-to-day operations (dudes yarfing in the backseat when their waymos careen down the road running red lights)
patreon is testing a native live streaming feature, including 24/7 streams, chat, scheduled broadcasts, and paid-only access. i am
insanelybullish on the future of live streaming <> influencers, so i’m a fan of this movenetflix is building search with ai, aiming for better recommendations via an interactive ai interface. feels like a great discovery mechanism for netflix to better match users with content they’ll love, which they might not have otherwise discovered
major ai product news
openai announced o3 and o4-mini, new ai reasoning models designed to pause and work through questions before responding
google added video generation to gemini, allowing users to generate short 720p clips using images as prompts (google’s answer to openai’s sora model)
xai launched grok studio, a canvas-like feature for editing and creating documents and basic apps
anthropic integrated claude with google workspace—it can now access your gmail, calendar, and google documents
openai launched a series of models gpt-4.1, focused on coding and instruction following
hmmm
openai released a new safeguard that blocks prompts about building biological / chemical weapons. i feel like a more effective solution would be partnering with law enforcement to flag folks with sus queries, but i suppose the two are not mutually exclusive
openai’s latest models supports reverse location search: upload a photo, and the model tries to identify the location based on visual clues. pretty scary; think we should also just “@ fbi” these fools, same as the chem weapon fools in the bullet above ^
google added an ai quiz question generator to google classroom, allowing teachers to generate quiz questions by uploading files or inputted text. not sure the roi on this, as teachers could also easily drop these docs into chat gpt… but honestly i’m all for anything that helps our teachers. shout out teachers; you’re doing god’s work
bye bye, love you,
dj