Signals, not summaries

This is a fascinating experiment: Tom Tunguz put a prompt box on the top of his web site that will generate a custom blog post to answer your query — based on his body of writing.

But I don’t think that’s why people read his posts (or emails). People don’t read his posts because they want to answer a question about startups, AI, blockchain, etc. They read his posts because they want to know what’s on his mind.

For a while now, I have been thinking about how good AI is at both generating text and summarizing text. The obvious long term implication is that nobody is going to read or write anything raw on the web anymore. It’s the opposite of compression: your thoughts get expanded by AI and then reduced on the other end by the receiver. Actually it’s not just reduced, it’s reinterpreted with your preferences in mind because the GPT you use will know something about your preferences. That’s a scary idea because it’s like playing the game “telephone” with some unreliable algorithm in between each person.

I don’t think that is what will actually happen.

Content marketing is going to have a rough decade. Here I agree with Tom: I am tired of trying to glean something useful for myself from an endless litany of articles published for the branding value. This kind of content is going to be summarized and reinterpreted and it will be improved in the process.

In this scenario, what is blogging for? Why post to BlueSky or Substack? Why read it?

The answer is that humans are social creatures and we want to know what other people are thinking about.

Where are you putting your attention? That’s the signal I get from Tom’s blog and it’s why I subscribe to his and many other newsletters.

What do I think about something? That’s the value I get from writing my own posts and thinking in public.

So I like the idea that I can generate a custom article derived from Tom’s posts, but I am guessing that’s not going to replace posting anytime soon.