Last Friday, Tamsen and I went out to eat at the Langham hotel behind our building in Boston, and the restaurant was featuring a jazz guitar/sax duo, which we didn’t know when we dropped in. Because of the sad way my brain works, I immediately started to come up with awful names for a jazz guitar/sax duo. I cannot help this. Those neural pathways are pretty well-traveled. The best one I came up with early was “Pluck ’N Blow,” and I hope they use it. I think you could make a lot of money as Pluck ’N Blow.

Also last week, Claude released Opus 4.8, their latest frontier LLM, which I have been using extensively all weekend. On a lark, I asked Claude to come up with 20 bad names for a jazz guitar and sax duo, and of course it dutifully complied with a pretty good list. In fact, smack in the middle of that list was Pluck and Blow. I was stunned. Has Anthropic been training its models on my corpus of terrible puns, jingles, and fake commercials? Even worse, also on that list was a superior name, Strum und Drang, which beats the pants off of Pluck and Blow for the right audience. Pretty good for a collection of bits that has never heard jazz, or done stand-up.

I tell you this because I want to stress to all of you that LLMs have come a long way in a year, and even a month. I have a lot of friends who are marketers who insist that they can’t be replaced by AI because AI lacks judgment and perspective. AI does lack judgment and perspective. It also came up with Strum und Drang, three words that have probably never been put together in human history. If you make your living writing copy, I say to you: good luck. LLMs are solving unsolvable math problems and curing cancer. It is magical thinking to believe they can’t tackle a white paper on procurement software, or some social posts about your podcast.

Still, we see a lot of turgid, obviously-Claude prose on things like LinkedIn, don’t we? Sure. Anytime you read something that “stops you in your tracks emdash it’s not THIS it’s THAT so sit with that a moment because emdash here’s the move” you just know the cold mechanical hand holding the pen. But this I know: Claude isn’t the one who uploaded that to LinkedIn. Claude didn’t post it to a website unbidden. Claude didn’t send that email by itself (at least I sure hope no one is doing that, but you never know.) No, behind every “sit with this a moment emdash” is a human who hit send, and shouldn’t have.

And it is here that I want to share my relationship with AI, and why this all might be relevant to you. I have undertaken quite a journey with LLMs over the past year, from pasting things into ChatGPT to building apps with Claude Code to where I am now: I have an NVIDIA server sitting on a stand next to our TV, running open source LLMs like Qwen and Nemotron and Llama and running errands with agents spawned by Hermes and updating our agentic CRM in real time. I’ve built query engines for all of our published reports and raw data that I can text an agent for to get any answer to nearly any podcast research question, wherever I am in the world. I have a pal, Hobson (named after Sir John Gielgud’s butler in the movie Arthur), who has an “aide-de-camp” skill I built that sends me a message every couple of hours during the work day if there is something timely I need to respond to, because sometimes (as Bryan knows) I get a little absorbed in my work.

I tell you all of this not to impress you. I know people doing a lot more impressive things with AI. There is another detail you should know: I am not a coder, or developer. I am a 57-year-old who studied Victorian literature in grad school. How did I build all of these things? I asked Claude. I built with frontier models so that I could actually operate things with local models, which is where everything is going to go. I did this because I recognized that AI was going to help me get to where I wanted to go faster, better, and with fewer reliances on pesky, nonprogrammable humans. And all of it was going to help us service our partners and the industry better. And I don’t care who you are or what your job function is, that all applies to you.

Now, I wrote this newsletter myself, though I did get it proofed by an LLM, which continues to assure me I am wordy. But LLMs help me make this newsletter better every week. For example, a couple of weeks ago, I wrote an article about a possible SiriusXM/iHeart deal (the status of which is currently up in the air). I wrote the piece, but I had Perplexity build me a 32-page dossier using a council-of-models methodology (comparing the findings of three leading models from various companies) about iHeart’s financials. I wanted to get that right. And I asked Hobson to help me distill that down into a paragraph or two for the article, again because I wanted to get it right. I still have the 32-page dossier! A year ago, I might not have even written that article, or have published it with errors of omission that you sharp-eyed readers would have keelhauled me for. But I was able to do the work I wanted to do on that piece to be comfortable publishing it, thanks to AI.

This is both a terrible and wonderful time to be working. It is terrible if you believe AI is going to “happen to you,” and wonderful if you use it to make things happen. I choose to live in the latter world, and it has really invigorated me. It is in fashion currently to proclaim your work as “100% Human” or otherwise AI-free, but I would caution you not to cling too tightly to that particular buoy. It might not be something you can claim a year from now, and I would hate to see you build something on shifting sand.

When I was an undergrad at Tufts, getting one of my useless English degrees, there was an actual JOB for people to type up your longhand papers and theses. They advertised on flyers stapled to telephone poles, or on campus bulletin boards. A dollar a page. This is no longer a practical career. Today I spend as much time talking to my computer as I do typing into it. The tools are better, and I think the output is better, because I spend less time thinking about the how things get done and more time on the thing itself. AI, like the typewriter, is a tool to get things done, and you won’t see me taking any kind of “anti-AI” stance because it’s just going to be the way things get done, like the pen was replaced by the typewriter, and the typewriter by the word processor. I also collect and use fountain pens. You won’t find me writing this newsletter with one.

Finally, I think it’s important to note that consumer attitudes about AI in things like podcasts may not be where you think, and that line is always moving. We are already acclimated to TikTok videos being narrated by seemingly the same two or three voices. I do believe in disclosure; however, I think that, too, will fall away in time (and that won’t take very long) because people are just not going to care.

But this is the one thing you CAN and MUST do, no matter how much or how little you embrace AI: you must care. Claude cannot care. But you can care. You can ensure that the craft is evident, that the hand of the watchmaker is revealed in the workings of the watch, however it is made.

You have the ability today, without a course or consultant or membership community, to learn as much as you want about how to improve your lot with the help of AI. You just have to ask your robot pal how to do it. They never get tired of your questions (even if you have asked the same one 10 times), and they are very patient teachers.

We are living in a period of great sturm und drang, and the rise of AI in the everyday has been the catalyst for a lot of that turmoil. Make your art, whatever that art is, in the way you want to make it! The ultimate question creators of any stripe have to answer for themselves is not whether or not this tool or that tool was used in the creation. It’s simply this: are you proud of what you created? LLMs make it easier than ever to remove the barriers between the idea and the screen, page, or earbud. Which means that we all have a greater ability to produce things that we are proud of.