🚀 Chatbots all the way down

Plus, how to know if your consumer social app might hit it big!

Mar 05, 2023

trending 📈

👻 Snap launches AI chatbot. Snapchat announced “My AI,” a chatbot powered by OpenAI. You can message the bot within Snapchat - it’s essentially an in-app version of ChatGPT (with some restrictions). This is an experimental feature that is currently only available for Snapchat Plus subscribers who pay $3.99 per month, but the team plans to eventually roll it out to all 750M+ Snap MAUs.

🔗 OpenAI releases APIs. More big news in the AI world: OpenAI announced APIs for ChatGPT and Whisper (its speech-to-text model). Some apps have already been using these APIs - for example, language learning app Speak leverages Whisper. We’re now officially in the chatbot era, with a number of new ChatGPT-powered products announced this week from Shopify, Instacart, Quizlet, and Snap (see above).

⏳ TikTok controls screen time. TikTok is making strides to limit the amount of time that minors spend on the app. The company announced this week that every account belonging to a user under 18 will now have a default daily screen time of 60 minutes. If they want to extend their screen time, the user will have to enter a passcode - and if they’re under 13, a parent passcode will be required to unlock another 30 minutes.

In other TikTok-related news, Canada has banned the app on all government devices due to privacy and security concerns. This follows a similar decision from the European Commission (the executive branch of the EU) last month.

deiniolb 🐻👉 aisuite.io @danberridge

I'm speechless. Not peer-reviewed yet but a submitted paper. The 'presented images' were shown to a group of humans. The 'reconstructed images' were the result of an fMRI output to Stable Diffusion. In other words, #stablediffusion literally read people's minds. Source 👇

File this under “crazy if true.” A paper submitted by a team of researchers in Japan posits that Stable Diffusion can read minds (via fMRI outputs).

Subjects were shown an image and their brain activity was recorded with an fMRI. The researchers then used Stable Diffusion to re-generate images from the activity data - and the results are pretty impressive.

what we’re following 👀

What happens to all the office furniture when tech companies scale down?

Commerce guru Nik Sharma digs into what’s behind all the seasonal Oreos.

Meta has dropped their VR headset price as Bytedance edges into the space.

Are you a builder, investor, or fan of consumer social apps? Check out a new piece from our a16z colleague Bryan Kim on the metrics that matter.

After years of leading growth at Snapchat and then investing at a16z, Bryan has identified a few factors that are early predictors of whether an app will go mainstream or sputter out. For the first time, he’s shared these + internal a16z benchmarks of okay vs. good vs. great metrics for early stage social apps.

His takeaways?

“Great” seed stage social apps have 50%+ MoM growth - ideally, this should be almost all organic. If your consumer social app doesn’t have an element of virality, you may not have product-market fit quite yet! Unfortunately, no amount of marketing dollars can fix a product that isn’t capturing user attention.
Even more important than growth is engagement. The most straightforward way to measure this is DAU / MAU - the % of monthly active users who are on your product daily. Social giants like Facebook, Twitter, Instagram, and TikTok all have DAU / MAUs ratios around or even above 50%, which is our “Great” benchmark.

The most important metric of all is…retention! Bryan measures retention in the most strict definition, called n-day retention. For example - if you have 100 users sign up for your app today, and 25 of those users use the app 30 days from now, your day 30 retention is 25%. Our “Great” benchmark for n-day retention is 70% at d1, 50% at d7, and 30% at d30.

If you’re working on something in consumer social with “Great” metrics, reach out to Bryan on Twitter. And if you’re not there quite yet, don’t worry - social apps often take a few tries to hit it big!