hi bud buds,
last week, google and open ai launched several ai <> education features:
openai launched chatgpt study mode, which asks students questions to test understanding and refuses to provide direct answers unless they engage with the material
google launched video overviews, which turn dense study materials (like textbook pdfs and class notes) into digestible videos, building on their earlier feature that converts study materials into podcasts
google added canvas to its ai mode for search, allowing students to generate personalized study plans by uploading class notes and syllabi
these releases follow anthropic’s launch of learning mode for claude in april, alongside a claude for education plan giving schools / universities access to learning mode for their students.
for some, the future of ai <> education sounds dystopian: kids staring at screens all day with ai doing the teaching, and teachers nowhere in sight. however, i personally believe improving education could become ai’s greatest gift to humanity.
today, american classrooms are designed for scale: 30 students in a classroom, all moving at the same pace. this works for the top ~20% of kids who are self-motivated and well-supported outside school. but it fails the rest: neurodivergent kids, esl students, or just kids who don’t click with a given teacher’s approach.
ai can revolutionize our classrooms:
personalized ai tutors handle core instruction on a 1:1 basis
teachers freed up to answer questions and help kids who get stuck
motivated students can move faster with ai coaching
group + solo projects give students opportunities to apply what they’ve learned
standardized testing ensures baseline standards are still met
in this world: fast kids accelerate, struggling kids get more support, and teachers are freed up to engage more meaningfully with their students.
beyond the u.s., the potential global impact is even bigger. any kid with a smartphone, even in poverty conditions, can access a world-class tutor 24/7. the upside for humanity is hard to fathom. i’m awestruck by how many future physicists, mathematicians, and engineers can be realized, who otherwise wouldn’t have had the education or support to reach their full potential.
so shout out to the product teams working on ai <> education features; i think your work will save our country, and maybe also the world.
to everyone else: have a great week. i love you too.
- dj
last week’s biggest product releases
hell yeah
youtube is loosening its profanity rules, now allowing swearing in the first 7 seconds of monetized videos. this is absolutely insane / feels like a tim robinson itysl bit: cuss as much as you want, but only in the first seven seconds of the video
microsoft launched ai copilot mode for edge (their internet browser). a few weeks back, i wrote about ai browsers being the next big thing; this further validates that thesis
substack launched a tiktok-style video feed to help writers get discovered by new followers. this feels like a fantastic discovery mechanism, and honestly has me wondering if i should switch from beehiiv to substack myself
paypal launched pay with crypto, letting businesses accept 100+ tokens at checkout, including btc and eth. i’m not even a crypto guy, but this feels like a huge step forward for crypto
nice!
apple rolling out priority notifications, using ai to elevate time-sensitive alerts on your home screen
instagram added a 1k follower requirement to go live; hopefully this means fewer isis beheading videos
tiktok launched footnotes, a community-driven fact-checking system like community notes from x. meta also launched a similar feature in april
cash app launched pools, allowing friends / family to divide shared expenses
uber eats added live order chat, letting merchants message customers to confirm subsitutions, clarify special requests, and check dietary needs
tiktok integrated with youtube music, letting users save songs they hear; previously tiktok already supported spotify, apple music, and amazon music
adobe launched harmonize, an ai tool in photoshop that blends objects into scenes by matching light, color, and shadow
quora launched poe developer api, letting devs access 100+ models for voice, text, image, and video. seems like they want to help devs abstract away the decision of which model to use?
foundation models
google released gemini 2.5 deep think, a $250/month ultra-tier model that can explore multiple ideas in parallel
anthropic set new rate limits on claude code to stop 24/7 background usage. shouldn’t they allow this and just bill for it?
for my tin foil hat amigos
jack dorsey launched his new app bitchat on ios, providing bluetooth-only messaging without cellular or wifi
#protectouryouth
youtube deployed age estimation tech to automatically detect teens and proactively apply protections like screen time warnings and removing personalized ads
tiktok launched new parental controls; parents can now link accounts, get notified when teens post, and block specific accounts from their teen’s feed
idk tbh
tiktok launched tiktok pro in europe, a clone version of their main app that provides increased support for charities, but disables ads, livestreams, and shopping. assuming this is eu regulation-driven?
manus added wide research, an “alternative to deep research” that allows you to deploy 100+ agents simultaneously across the web. assume this is primarily marketing, as the end user result seems the same as deep research: more robust answers than instantaneous chat completions
open ledger launched pokétax, a gamified tax prep tool where you battle “tax trainers” to earn deductions. feels like we've lost the plot a bit here. no matter how much dev effort you invest to make taxes fun, they still won’t be. imho, the main job of a tax prep provider is to help me finish my taxes quickly and confidently
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 a cabin in the woods.
love, dj