Read on Twitter
Incredibly thorough and fascinating pod with @a16z 16 min episode with @smc90 & @eugenewei on:
Tiktok's unique qualities, "seeing like an algorithm", and applying the lessons we learn from TikTok to other future products.
Learned so much from this one!
Notes in Thread!
Tiktok's unique qualities, "seeing like an algorithm", and applying the lessons we learn from TikTok to other future products.
Learned so much from this one!
Notes in Thread!
1/ Uniqueness of Tiktok:
⢠Closed-loop ecosystem: users create training data that it then trains its algorithm on
⢠Functionality: High-production film techniques accessible
⢠Creative Network Effects: For You Page makes trending ideas salient â easy remix of ideas
⢠Closed-loop ecosystem: users create training data that it then trains its algorithm on
⢠Functionality: High-production film techniques accessible
⢠Creative Network Effects: For You Page makes trending ideas salient â easy remix of ideas
2/ Structural Components of TikTok
⢠Duets â easily 'react' to people's content without the blank page issue when first coming up with content
⢠Full screen video, captures users attention and data effectively (how long you watch/scroll past etc)
⢠Licensing of Music
⢠Duets â easily 'react' to people's content without the blank page issue when first coming up with content
⢠Full screen video, captures users attention and data effectively (how long you watch/scroll past etc)
⢠Licensing of Music
3/ Free vs Managed?
⢠TikTok is a mix of a free market and a managed economy.
⢠Managed: Trending items at the top of the page, so you know if you do it you have a higher likelihood of virality
⢠Free: Anyone can become famous
⢠TikTok is a mix of a free market and a managed economy.
⢠Managed: Trending items at the top of the page, so you know if you do it you have a higher likelihood of virality
⢠Free: Anyone can become famous
4/ TikTok as an âassisted evolutionâ ecosystem
⢠Uses algorithm rather than a social graph, as the primary axis of distribution.
⢠Algorithm allows you to see content from people that you donât follow that you would enjoy otherwise
⢠Uses algorithm rather than a social graph, as the primary axis of distribution.
⢠Algorithm allows you to see content from people that you donât follow that you would enjoy otherwise
5/ TikTok's "seeing like an algorithmâ
⢠Huge operations team that when videos are made, are tagging videos with features and attributes
⢠Full screen one video mode allows them clues as to your sentiment on individual videos/tags
⢠âAlgorithm friendly designâ
⢠Huge operations team that when videos are made, are tagging videos with features and attributes
⢠Full screen one video mode allows them clues as to your sentiment on individual videos/tags
⢠âAlgorithm friendly designâ
6/ TikTok is a 'sorting hat'!
⢠Efficient at matching content, to people who will enjoy that content
⢠Breaks down geographical barriers
⢠Peopleâs tastes change over time. TikTok algorithm can quickly detect this
⢠Efficient at matching content, to people who will enjoy that content
⢠Breaks down geographical barriers
⢠Peopleâs tastes change over time. TikTok algorithm can quickly detect this
7/ Tiktok within the taxonomy of Social Networks
⢠Not really a 'social network', but an entertainment network
⢠Primary purpose is to match these entertaining videos from creators, to the audience that would enjoy them
⢠Not really a 'social network', but an entertainment network
⢠Primary purpose is to match these entertaining videos from creators, to the audience that would enjoy them
8/ TikTok Acquisition?
⢠People may be overrating the algorithm in terms of how unique the algorithm is itself
⢠BUT, without it, would take a long time to rebuild a new one
⢠May not be as responsive and lead to churn
⢠People may be overrating the algorithm in terms of how unique the algorithm is itself
⢠BUT, without it, would take a long time to rebuild a new one
⢠May not be as responsive and lead to churn
9/ Applying Lessons from TikTok
⢠Cut straight to the interest graph vs standard social network
⢠We underrate how much people respond to video vs text (High Bandwith Medium)
⢠For video to scale, you have to overcome some of the challenges inherent to video
⢠Cut straight to the interest graph vs standard social network
⢠We underrate how much people respond to video vs text (High Bandwith Medium)
⢠For video to scale, you have to overcome some of the challenges inherent to video
10/ Final Thoughts 
⢠Body language is 'universal' and that's what video-centric media does
⢠TikTokâs impact = model for building a new sort of network based on algorithmic recommendations
⢠âTikTok is such a fertile source for origination, mutation, and disseminationâ
⢠Body language is 'universal' and that's what video-centric media does
⢠TikTokâs impact = model for building a new sort of network based on algorithmic recommendations
⢠âTikTok is such a fertile source for origination, mutation, and disseminationâ
