Channel: LLM - Large Language Model discussion
https://blog.brianbalfour.com/p/the-next-great-distribution-shift 🔥 my verdict: it stinks!
There are several glaring factual or interpretation errors that cloud his analysis of the space.
But first, his analysis is that "tech platforms" go through a three-stage cycle:
- Identify the moat
- Open the gates
- Close for monetization
This is, on some level, accurate. There are a lot of people trying this on Substack, or elsewhere. Give it away to get viral growth when you are small, charge and profit when you are big. Business 101.
His analysis of how this worked for multiple large tech companies, unfortunately, is fatally flawed.
Facebook - to start with, several of the "added features" like Marketplace and Photos existed years before the ill-fated "Facebook Platform" ever launched. But, the real problem is that Zynga-style gamification cluttering activity feeds was a disaster 💡 although, in hindsight, it was probably better than the political takes that followed it. They could not allow it forever, so it stopped. And, without a limitless reservoir of free advertising, Zynga crumbled.
Also, constantly throughout this post, the author is nostalgic for what can best be described as awful crap. The "quizzes" and "vampire bite" apps were not something to be defended; the people who ostensibly "made millions" from them not to be celebrated.
Apple - first off, the "70-30" revenue distribution is considered as both "open" and "closed". Once again, people being able to make millions off the "iFart" app is not to be celebrated. And the demands that the platform never include features that are in platform apps are unreasonable.
But the more serious concern is the specter of government regulation. Most of the "privacy restrictions" he criticizes are imposed by the government. And the anti-trust concerns govern pricing more than anything else.
Google - the SEO ⚙️ search engine optimization industry is largely the scourge of the earth, and should be destroyed. 💡 there seems to be a trend that it is bottom-feeders, creators of frivolities, and abusers of the public common that suffer from this "contraction". Unlike the author, overall I would prefer those changes
The problem here is that the moat is contrived and unrelated to the changes. The increase in oneboxes ⚙️ in Google parlance, a "onebox" is a structured-content response, other than organic search or an ad. For example, a "weather" onebox for a search for the weather. They were called "oneboxes" because there would be only one per search result page, but that seems to have gone by the wayside. was certainly a change from the 2004 philosophy ... but it's unrelated to market-share growth against Yahoo.
LinkedIn - I'm not even sure what he's talking about. I don't believe there was a "2 year window" where LinkedIn encouraged content creators. I think it was either a COVID effect, or a personal observation of this guy extrapolated to the whole site.
And, once again, the people creating "B2B Marketing" content are people I do not want to succeed.
as a reminder, viruses are bad. these people act like cockroaches, and then are offended that people want to make them go away.
So the conclusion, that platform cycles are accelerating, cannot be supported by this data. And the attempt to project this onto ChatGPT is bad.
The moat has to be chat memory ... because this guy can't think of anything else. And, even if it is ... the platform that can access your email history is more valuable than the chat history.
Also, ChatGPT can't shut down the API. For one, it's profitable. For two, the alternative of "self-hosted models" is too good for this to meaningless harm the (now)-competitors.
The ChatGPT "market share" advantage is driven by higher name recognition among non-technical people, and people who haven't tried competitors. They are in the position of Yahoo! and MySpace. I'm not saying that this means they can't be successful ... but there is not such an advantage that Google and Facebook are irrelevant.
And, finally, the complaints that ChatGPT is competing with enterprise players like Glean are pathetic. Presumably the feature is "connectors" ⚙️ https://techcrunch.com/2025/03/17/openai-to-start-testing-chatgpt-connectors-for-google-drive-and-slack/ : being able to access data in your Google Drive account. This is a feature that Claude has had for a long time. Once again, this is a ludicrous demand that a platform remain incomplete because someone else implemented a necessary feature.