Updates on Crypto, AI, and social networks
Crazy Week
I have little to add about the crazy crypto week. My one major takeaway from Crypto is summed up in this tweet. Not knocking anyone specific (part of venture investing is swinging big and often missing). Many of the very smartest people have operated in ways that now look mind-boggling. Bullish about my future as I continue to do the work.
Substack social network
Twitter has been newsworthy at a bare minimum for the past 2 weeks. It seems likely that someone will try to swoop in during the chaos and usurp Twitter’s position. This is highly unlikely given the network effects (and Twitter is a default behavior for many at this point). Mastodon is structurally flawed - the federated model cannot scale to the same extent centralized Twitter can.
One more promising take is from Substack. The Substack Chat offering is being likened to Twitter 2.0 in some circles. The key elements
A per-writer function - this is a series of individual communities, not one monolith
Ability to paywall - helps writers monetize
Uses the Substack iOS app - this is a benefit to Substack more than the writer
I like this concept, but it bears no resemblance to Twitter. This is a way for Substack to help publishers monetize and maintain control. In the launch blog note, Substack admits writers are moving communities to other platforms like Discord and Telegram. ‘Community’ is probably the next frontier of social, though no one has truly grokked it quite yet. I would expect Substack to more closely resemble Patreon in the coming years in order to offer creatives the best platform to earn a living.
AI hype train
Most of Tech is quite beaten down, but AI startups are raising money like the glory days haven’t ended. I’m a big proponent of the value of these recent AI advances, yet much of the fundraising strikes me as problematic (see the first block about Crypto). My biggest concern here is the idea of building a company on top of publicly available models from other companies. OpenAI seems like a stronger position to be in than any of the companies leveraging Stable Diffusion or Dall-E. I expect a rush of similar companies using similar data sets - a bad recipe. The other concern here is that the assumption AI will be differentiated is very much an open question (see Ben Thompson and others on the subject). The very existence of Stable Diffusion (widely available and open source) should call into question the assertion that AI models will remain controlled by the few.