Tyler Cowen is one of those great minds that whenever an interview is released, I’ll take a listen. An Economics professor at George Mason University, Tyler could be better described as a polymath, and he is most prolific as a writer on his blog Marginal Revolution.

The world I’ve lived in has not changed that much since I was born but that’s about to change. You could say its changed already but it’s not fully instantiated in most of the things we do

David Perell interviewed Tyler about “Learn How to Write with AI in 68 Minutes – Tyler Cowen”. But the interview is really about how Tyler is using AI to research, satiate his curiosities, and make himself smarter - and there were some good ideas that I thought worth taking a note of.

Takeaways from the interview:

1. Using AI as a secondary source while you’re reading

Tyler uses AI when he’s reading. Tyler is the type of reader who uses a book as a jumping off point, to fractal out to other topics. In the past he would buy 20-30 supporting books (Youtube timestamp:&t=79), but now he can simply ask the AI. The idea here is that your AI is essentially a digital colleague who has read a lot of books - perhaps every book?. He describes AI has his main secondary source.

Tyler is using several models (Youtube link) at the moment, favouring OpenAI’s o1-Pro, and also using Claude and DeepSeek. He is a proponent of using the best model available. Later in the interview he talks about using Perplexity (Youtube link), which has replaced his use of Google as it is real-time.

2. You’ve always double-checked sources, haven’t you?

Regardless of whether you’re getting your information from an AI, or from a traditional book or other source, you should always double-check your sources (Youtube link). This has always been the case, and it continues to be the case with AI responses.

3. Tyler doesn’t use AI to write - he prefers to keep his own voice

Tyler doesn’t write using an AI (Youtube link), as he prefers to keep his own writing style. They discussed whether some people in certain contexts, for example when writing memos at work or writing in a certain legal style, may allow the AI to write some of the content or to improve the tone of communications.

4. Learn Prompt Engineering - People are being too general in what they ask AI

In response to a question on how people may be using AI wrong, Tyler thinks most people are being too general in their questions. (Youtube link). They talked about taking time to generate context, and the interviewer talked about recording a few minutes of context (Youtube link) up-front to get better results.

Tyler warned that if you don’t learn how to ask AI the right questions (Youtube link), you will end up unimpressed, be left behind, and you could be left with the false impression that AI is only good for generating basic tasks (mid-tasks).

5. How has AI changed Tyler’s writing

Predictive books (Youtube link), books that forecast the near future, have less or no value now. Things are changing so fast. So you need to be writing and publishing on a much higher time frequency to stay relevant. Tyler suspects he may currently be writing his last book (Youtube link). So what writing will be of value in the future? Perhaps anecdotes about real people - truly human books will stand out, books based on personal experience, or those written through interviews with real people.

Tyler also suggested that the models answer most questions on economics (his specialism) better than him now - and wonder if in that regard he may be becoming obsolete (Youtube link).

6. AI in the classroom

Tyler encourages his students to use AI in the classroom (Youtube link). He is conscious that AI will soon be so integrated into research and work life that students should learn how to do it sooner rather than later.

7. What AIs don’t know

AIs don’t know secrets: Humans know secrets & everyone has secrets. The AIs don’t know those secrets (Youtube link) (yet). So those secrets have value. There may be an incentive to hoard information, and to keep secrets.

Social network value has gone up: Social networks have become more important. Travel and meeting people is essential.

Some prompt ideas from this video

Some AI Prompt ideas I picked up from the interview:

Ask Prompts such as

  • What would author X say about this?
  • Fact check this piece of text. Act like you’re out to destroy me.(Youtube link) (&t=357s). I think Tyler was referring to his research assistant, not an AI prompt, but one worth trying with AI anyway.

Ask questions about things you are curious about. Tyler avoids simple questions that can lead to normie answers (Youtube link). Ask for details of historical examples, so for example: What was so special about XX and what did people think about it at the time? Keep on asking specific, practical questions.

After you’ve written a memo, article, or essay, the AI can advise on how it might be received by a reader:

Creating the Perfect AI Prompt (Youtube link), asking basic questions, snapping photos, creating itineraries