hi bud buds,
last week was openai’s dev day for 2025. here’s what you missed:
1) apps inside chatgpt
companies can now create apps that work directly inside chatgpt, using the new ‘apps sdk’. for example, you can link your spotify account to let chatgpt access and edit your playlists with a custom spotify interface built into chatgpt. chatgpt understands what's happening inside these apps, so you can type commands like, “now remove all the songs over 5 minutes,” and chatgpt will update your playlist directly.
to access apps, users can mention them by name when giving instructions, such as “make me a halloween playlist” followed by “now save it in spotify.”
notably, chatgpt can also recommend relevant apps based on your conversation. for instance, if you ask about buying a home in dallas, chatgpt might suggest running a zillow search and show the results in a custom zillow window—all within chatgpt.
2) easier agent building
agentkit is a full toolkit for building, deploying, and evaluating agents. it brings together a canvas-based builder for visually creating agents, custom user interfaces, and new enterprise controls to help manage data and connected tools.
there are also built-in evals for measuring agent reliability and performance over time. everything is designed to make building and managing agents as straightforward as possible.
my key takeaways
the most interesting paradigm shift to me: openai is now recommending which apps users should use next within chatgpt. right now, this is limited to launch partners like zillow and spotify, but it’s easy to see how this could evolve into paid placements. someday soon, apple music and spotify may have to bid to be the default “music app” when users ask chatgpt for a playlist.
we’re seeing more and more signals of openai plans to monetize, and the potential seems astronomical.
have a great week, y’all. i love you.
- dj
last week’s biggest product releases
big news! huge!
google launched gemini enterprise, a toolkit for companies to build ai agents that connect with apps like salesforce and microsoft 365. hard to see how ai agent startups will compete with the distribution advantages of google (and microsoft)
amazon is piloting pharmacy kiosks at one medical clinics in los angeles. if these can miraculously avoid getting immediately broken into and plundered, i love this idea
openai is expanding chatgpt go (its < $5 plan) to 16 more asian countries. another signal that llms are commoditizing, with prices racing to zero
personal faves
zendesk launched a new ai agent aiming to resolve ~80% of support issues autonomously. similar to gemini enterprise, i predict startups will struggle to compete with zendesk’s data and distribution edge
facebook shipped updates to the reels algorithm, adding ai search and more feed customization. more cute stuff!! less goggins!!! (i can’t stop watching)
nice job guys
instagram shipped clearer location sharing for instagram maps, after users got confused about what they were actually sharing
google is expanding virtual try-on shopping, signaling positive traction thus far? very interesting…
netflix launched tv games, letting subscribers play group party games on their tv using their phone as a controller
apple teased glasses running full visionos, a pivot from headsets to lighter, meta-style glasses for folks who want smart displays without the bulk
square launched several new features for merchants, including:
ai-powered voice tools for restaurants to automate taking incoming calls for reservations and takeout
accepting bitcoin payments
an ai assistant with local insights like weather and nearby events
youtube launched a second chance program, letting banned creators apply for a new channel after a year. techcrunch speculates this is from political pressure related to reinstating accounts banned under covid or election policies
x is splitting its verified orgs product into two tiers:
premium business: for companies; providing a gold checkmark, hiring tools, and brand monitoring
premium organizations: for governments; providing a grey checkmark
google launched chrome notification mute, disabling browser notifications from sites you haven’t engaged with recently
other llm news
openai and amd announced a partnership allowing openai to take a 10% stake in amd, in exchange for deploying about 6gw of amd gpus over the next few years
google launched gemini cli extensions, allowing outside companies to build directly into the gemini cli
google launched expanded access to opal, its vibe-coding app, to 15 more countries
figma launched a gemini ai integration, for faster ai-powered image editing and generation
idk tbh
doordash is testing sidewalk robots in los angeles. as someone who lived in la, these things are gonna get kicked over so fast
otter.ai released an enterprise suite with better api integrations, mcp server support, and ai-powered meeting search. i still don’t see how they compete long-term with zoom and google meet, who have already built these features...?
google launched nano banana in ai mode and google lens, letting users edit and generate images inside search… unclear what the user story is here.
“as a user searching for flights, i want to generate an image of a plane”? lol.
that’s it for this week; thanks for reading. if you enjoyed this issue, please consider sharing it with a friend so i can become famous and acquire hard to get dinner reservations in nyc.
love, dj