WordPress.com has been sprinkling AI features into its platform for a couple years now. These started with proofing tools to catch spelling and grammar mistakes, suggest refinements, and generate images. While I don’t rely on them, I appreciate that they’re included in my subscription and are available to experiment with, often providing a mix of productivity and entertainment. Last year Google introduced podcast generation to NotebookLM and it was… interesting. WordPress.com has since launched a similar feature so I decided to try it out.
Generating the podcast is easy – select a date range or a set of specific posts, wait a few minutes, and enjoy a 3 minute audio podcast with notes. I selected the four posts I wrote about Simple Beep because I thought it would provide a reasonable length subject matter with enough content to dig into and discuss.
The result was very similar to what I experienced last year when I asked NotebookLM (now Gemini Notebook) to podcast about a collection of old Macworld magazines. Google’s podcast chose focus areas randomly, fixating on printer advertisements. It overstated the impact of nearly everything, implying that font management was a game-changing capability, and completely missed the important events from the timeframe covered. The WordPress version about my articles wasn’t much different.
Wow, umm… thanks?
The podcast on my own content started much worse actually, by making fun of my posts! The opening statement says it all (emphasis mine). Give it a listen to hear the sarcasm.
Relatively Ambitious — where looking back at a podcast about looking back at Apple is apparently a perfectly reasonable use of everyone’s time.
Thanks asshole for your sarcastic attitude and astute observation.
It doesn’t get any better from there, spending the first third of its three-minute runtime talking about the structure of the podcast and the structure of the articles I wrote about them. From there it launches into random highlights – iTunes versioning, fan noise, and then an awkward wrap up.
Not My Style
The thing I dislike the most about this style of podcast is that every included topic is editorialized. There ‘s always a follow-on statement or “witty” quip about what the “host” “thinks” about it. Phrases like “before anyone pulled the plug”, “not a subtle failure”, “catching up”, “a version number no-one asked for”, and “sounds about right” are littered throughout the script. This is very similar to Gemini Notebook; I think these models are trained on a popular two-person style of podcast where editorializing is common. I personally dislike the style. I want to hear facts and I don’t want over-dramatization.
When you only have 3 minutes to cover a topic, there’s no time to editorialize. I dislike meta-conversation, so don’t tell me about the structure of the content you are about to tell me about. Drop the fluff, choose relevant examples, and get on with it. And please, don’t make fun of the content. Yikes! I wouldn’t recommend that anyone listen to this as a way to consume my content; if you don’t have time to read it, just use text-to-speech please.
Future Promis
While I’m unimpressed with the “hosts'” capabilities, I am impressed with their voices. They sound very natural, complete with intonation, emphasis, and even a bit of emotion. These are the voices I want to read written articles to me. This is what I want Apple to spend my Neural Engine cycles on.
That’s not currently the reality and AI-generated speech is computationally intensive to process, usually handled in the cloud instead of on-device. As a result it’s usually priced based on token usage and has a time limit. During the “experimental” release, WordPress provides each user with 10 free credits every 30 days, and each podcast uses 1 credit. There are additional customization options coming, which may take additional credits, but I’m not sure. At the moment, I can create 9 more terrible podcasts to experiment with 🙂
As you can probably guess, I won’t be turning my articles into podcasts on the regular. As with everything AI, these things will improve, but at the moment they’re not where I want them to be. Hopefully AI text-to-speech will become easier to do locally and without enormous token-gobbling subscriptions.
For reference, you can view the entire podcast script below
Pip: Relatively Ambitious — where looking back at a podcast about looking back at Apple is apparently a perfectly reasonable use of everyone’s time.
Mara: jasontaylorwordpress has been working through Simple Beep, a podcast about Apple history that ran from 2014 to 2021, and writing up follow-up of his own — covering predictions that aged badly, ones that aged well, and a few that are still aging.
Pip: Let’s start with the series from the beginning.
Simple Beep Gets Its Own Follow Up
Mara: Simple Beep was a podcast about Apple and Mac history, and the posts here are structured the same way the show itself opened — with Follow Up. The question running through all four installments is: what did the hosts get right, what did they miss, and what has quietly become absurd in the years since?
Pip: The first post sets the frame directly: “Now that I’ve listened to the entire thing, I thought it would be fun to reflect and do some Follow Up of my own.”
Mara: That’s the whole project in one sentence. And the material rewards it. Take the iTunes prediction from Part I — the hosts in 2015 expected iTunes 13 soon, assumed music would eventually move into a standalone app, and were essentially right about the destination but off by several years and several versions.
Pip: iTunes made it all the way to version 12.13.10.3 — a version number that reads like a ransom note — before anyone pulled the plug on Windows.
Mara: That detail is in Part I, and it connects to a broader pattern the series tracks: Apple’s transitions taking longer, or going differently, than anyone predicted. Part II picks this up with the Intel era. The hosts in 2016 predicted Macs wouldn’t be on Intel chips a decade out. They were right — the Apple Silicon transition happened four years later, not ten.
Pip: The fan noise section in Part II earns its keep. A work machine whose fans started spinning after two minutes of use and didn’t stop until the lid closed is not a subtle failure.
Mara: Part III gets into Apple design books — Iconic and Designed by Apple in California — and corrects the record on flat design. The post is direct: Microsoft introduced flat UI with Windows Phone 7 in 2011, tracing back to the Zune and even Encarta. Apple, by that account, was catching up.
Pip: Part IV closes the series with odds and ends — a Myst miniseries that was canceled, the M1 iMac arriving as the big iPad on a stand the hosts once joked about, and a 2019 quote about ordering a part from the UK that ends with “unless you’re listening to this in a weird dystopian future, but let’s not worry about that.”
Mara: The series wraps with a genuine note of affection for Simple Beep — and a wish that the show had returned for a few more anniversary episodes beyond the two it did produce in 2021.
Pip: Predictions, corrections, fan noise, and a version number no one asked for. Sounds about right for Apple history.
Mara: What holds the series together is that the original podcast was already doing this — revisiting old claims, updating the record. The follow-up just extends the timeline.
Pip: More installments are promised. The past, apparently, is not finished yet.
Relatively Ambitious 