It’s amazing that Meta will continue to promote an app that cannibalizes engagement from Instagram purely to maintain a pipeline of training data for their AI models and preserve their moat against ChatGPT.
More and more this feels like the only longterm advantage in the AGI race. Without a daily pipeline of knowledge, all other LLMs are basically as smart as a coma patient.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
There is a tendency for app designers to create layers & subgroups to deal with complexity:
Mastodon attempts this with usernames—which have 2 parts.
For every part of your app that you fragment, expect to increase your app’s overall probability of failure by 50%.
Users don’t have the patience to learn about the subworlds of your community. They are more motivated to churn than to understand.
Early products already have a limited inventory of content. When you fragment things, average engagement per post takes a hit, which is the key metric to track the health of your app.
Ladies and gentlemen, I’d like to introduce Nikita’s Shitposting Club—a collection of 69 Nikitas. Owners will get access to “Nikita’s app” (there is no app).
After 10 years of building consumer social apps, I've decided to start exploring new areas. Building these products is an unforgiving grind—but I learned a lot along the way.
For those embarking on this path, here's everything you need to know:
TIME FOR A THREAD 👇
A reproducible testing process is more valuable than any one idea. Innovate here first.
All things equal, a team with more shots at bat will win against a team with an audacious vision.
Most product ideas are Dead On Arrival because the conditions to derive value are impossible to orchestrate. Getting 7 adult friends to install an app on a reproducible basis is non-trivial. If you can figure out how to do that, that's a bigger idea than your original concept.