Loneliness, 3D, and my fascination with Micro Apps
ShippedWith.AI is a weekly roundup of interesting things built atop AI, paired with ideas for entrepreneurs and investors in the space. The AI market moves fast. Our goal is to help you understand and find inspiration in it.
Hi folks! It’s Round Two, Halloween Edition — I hope you’ve got your costumes ready. As for me — I’ve reached “I’m ordering a Top Gun jacket from Amazon” years old.
Thought of the week
Two separate AI founders recently mentioned people using their products to combat loneliness. If today’s stateless LLMs are already serving that need, expect a flurry of “companionship apps” once LLM APIs support per-user state & prompt subroutines (“IF (USER mentions a stressful event) THEN (ask them the next day how it went).”).
Coincidentally, Scott Galloway’s book Adrift: America in 100 Charts was making the rounds last week for a figure showing that nearly 20% of American men and women report only having a single close friend.
This week’s links
Holovolo TV by Forrest Briggs - Generate Volumetric 3D scenes with Stable Diffusion.
Reaction: As a parent, I promise you this: everyone with a child will buy the app that lets us relive the memories we’ve saved. Just give me a commercial of someone stepping inside a video of their child taking her first steps, followed by a QR code labeled “take my money”.
More seriously, I’m convinced some of the slow growth of the VR market is supply constraint: it’s just too much work to build good 3D worlds. The headsets keep improving but the App Stores are depressingly empty. If that’s true, then prompt generated 3D assets sounds like a pretty strong sell to would-be VR app builders.
Explain Paper by Aman Jha and Jade 🧃 - An AI Postdoc that explains academic papers as you read them.
Reaction: “Explain it to me” is the new “define it for me.” Until OS vendors figure out how to ship this as the new system dictionary — and that will take years — LLM-powered explanations like this provide a great usability wedge for new tools to break into a complicated domain and expand from there. As a former academic, a tool like this, plus a citation summarizer, would have been a godsend. Caveat: I still consider open-ended question answering “too dangerous to fully trust” — LLMs are quite happy to make things up.
gpt3-email by Daniel Melchor - GPT-3 Chrome Plugin in a few hundred lines of code.
Reaction: GPT-3 in your GMail with a few hundred lines of Typescript? Put it next to a company like Compose AI ($2M Seed 2021) and you’ve got the perfect AI Rorschach Test.
I think there’s plenty of room for both — this is a sign of technology-market fit. LLM-assisted writing will transition to table stakes, not moat. For a dorm room developer like Daniel, the cost to prototype is near zero. For a company like Compose, the value-add is what it always has been: delivering a product that better addresses needs of its users. LLMs, like databases, will be along for the ride in either case.
What will your next video be about? by Yonatan Bendahan - Suggest the next YouTube video topic for creators.
Reaction: I am fascinated at LLM micro-apps because I keep wondering where their eventual home will be. The iPhone created a natural place for micro-entrepreneurship that the web never had: payment, distribution, and shelf-advertising all bundled in. A small, focused, app selling for 99c can bring in meaningful revenue for a weekend warrior. That’s a great motivator.
It’s unclear to me what the analogue for prompt engineering is. Do these AI micro-apps go the way of the web, in which the overhead to charge 99c makes them largely free? Or the way of the app store, in which easy non-zero price points are built in? It may be that there’s nothing new here: some will deploy to the web; some will deploy to the app store. But the fact that, for now, inference drives the cost basis tells me AI companies have a shot at influencing this distribution question.
Do you have something that should be on this list? DM me @edwardbenson on Twitter.