Through my 4 (!) podcasts I obviously have built up a lot of opinions on podcasting over the years. Here’s some of them. The two outlier podcasts of our time are Dwarkesh and TBPN, and I will explain my mental model of them to the end - the following advice is what I strongly believe for the “average” podcast that wants to do well but not necessarily being full time top tier influencer level. Also, this is primarily for professional/technical interview guest-led podcasts, but I also do have opinions on non-guest-led podcasts as well.

The Job To Be Done

Podcasting is edutainment. If you try to have too highbrow a view of yourself, and only do education, you will lose. Don’t forget to have fun (rather than trying to get them to “drop alpha”), don’t forget to have a real human to human conversation (rather than blowing through your prepared question list point by point without regard to what the guest just said).

Podcasting makes things human. Often when people do a talk, or write a blogpost, or publish a paper, they feel they have to dress their work up in formalisms and make things grander than they really are. At minute 45 in a podcast, they’re using the words they normally use in a conversation, and that can get at the essence or depth and that is useful for both the guest (who doesn’t otherwise have the forum or time to put those things down) and the audience.

Podcasting is church. Cal Newport just sits and rambles by himself for an hour each week and hundreds of thousands of people listen in, not even necessarily listening to every word, but simply to reaffirm their connection to his weekly Sermon at the Church of Deep Work. When you have a Cause you promote - whether relevant to your listeners’ lives or work or sense of self, you have loyalty. (“if you stand for nothing, what will you fall for?“)

Curation

(again this is assuming you have a guest led podcast, and there are many other forms of podcast)

Your guest and topic curation is the single highest order bit of growth of your podcast.

Growth is path dependent; insist on A-List launch. When starting Latent Space I absorbed advice from another podcaster with a recent good start that I can’t remember now, who said he waited and waited and waited until he got a big A-list name (eg Arnold Schwarzenegger) before he launched (hence we insisted on launching with OpenAI). This is counter to other creator advice - he could’ve just launched first, and figured it out later, but then he’d always be the B-list podcast, whereas starting strong meant that when booking other A-list guests, he could answer their first question well, e.g. “who else have you had on”. Getting A-list also solves the audience cold start problem because there are people who will just drop everything to go listen to that conversation not caring who you are, but might stay. As for getting that A-list - there are many that you can’t get as a newbie but there’s always going to be SOME who are kind enough to give you a break.

Post launch, watch your curation and topic weighting. If you ship weekly, you have 50 podcasts a year, and every guest is 2% of your weighting. If you aren’t vigilant, “easy (inbound) guests” who are less prestigious /impactful /impressive /whatever will far outnumber your “hard (outbound) guests” - which isn’t in itself a problem if you truly do not care about prestige. The real serious matter comes in mismatching coverage, where eg if you say you cover LLMs but 80% of your content is about open models, then you are inverted in your coverage when 80% of usage is on closed models. Ideally you are providing an accurate “map” of your chosen coverage area when people scroll back on your archive.

Nailing The Intro

• Rehearse the opening. The intro is the only “stateless” part of your podcast that you can do without the guest there, so you might as well not stumble over it
• Brag a bit. Encourage the guest to brag about their impact - prompt them to prep impressive numbers or accomplishments to remind the guest why they should care, and if they are being too humble, then be thick skinned on their behalf.
• Corrections are golden. “Did I miss anything?” / “How else do you like to be introduced?” gives guest a chance to close the gap between the common public perception vs how they see themselves

Planning The Content

A few ways to do it

• Chronological work history (if it has an obvious through-line to their current work)
• Cover their current thing up front, then go back in time and/or get commentary on other hot topics in the market later
• Start with results up front, then dive into internals/ask questions about alternatives, then future work (kind of like a paper review)

It can help to TELL THE GUEST YOUR HIGH LEVEL PLAN so that they will know to not jump the shark by starting a topic out of sequence when you have it planned for later. But if they do, improvise, don’t get upset.

Nailing the Outtro

• Get the CTA: literally “Any calls to action”? good alternatives from other podcasts:

• “how can people help you / and/or give you money?”

• “what are you hiring for?”

• I almost never ask “where can people find you” because obviously you’ll put it in the show notes

• Credit others: “who are you grateful for”/ “who else did great work on this that we should talk to?”

• sometimes a mixed bag with team leaders because nobody wants to name favorite child

• “Anything else you wish you were asked more about?” (if time permits as this may elicit a long last section

Principles of Podcasting

• Self critique: listen back to your own podcast and generate ideas for what you’d like to improve. Only way to confidently develop your own style.

• Identify their mission. Their message to the world, their source of alpha, the words they live by, what they want on their gravestone. That’s both your episode title and what they’d send to their friends, what you and your audience will retain when all else is forgotten.

• Mine for rants. Treat podcasting as a game where you are seeking the true passion and excitement of the guest. When you get them talking fast, talking loudly, changing pitch, dropping their voice, any vocal variation, you have won.

• Laughs are golden.

• Don’t ramble. Try to say the minimum number of words to let the guest know what topic you’re prompting - drop hints that you’ve done your homework if you can, but not strictly needed. But if you’re taking 1-2 mins to ask a question you’re probably doing it wrong. A confidently stated “why do you think X is important?” can be plenty.

• Definitions are load bearing. point blank, no bullshit, no ramble:

• “What is _?” (or: How do you define ____?)

• “What does __ do?”

• this is very quotable and referenceable and revealing and often gets forgotten to ask.

• Follow up questions. You can ask between 0 to 2 followups, average 1, so that you are reacting and doubleclicking and showing active listening to what the speaker is saying.

• The highest art of followup is disagreement - challenging what they said, pushing back on it. If you are worried about looking bad, you can “acknowledge the elephant in the room” or bounce it off a wall “others might say that ____?”

• Verbal highlights for listeners. Checkpoint where you are in the conversation, where you’re going, where you’ve been. Periodically recap.

• Tell the audience why they should care (who may not have shared context or may not be listening as closely). Repeat words back verbatim with surprise, shock, gravity, elaboration in your own words.

• Leave homework. References to readings, other people’s work, lets people use your podcast as a list of resources (this is not a major point and I also don’t do this often)