Episode 243: Nuts & Logs

Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:00):
Podcasting 2.0 for December 5th, 2025, episode
243, Nuts and Logs.
Well, ho, ho, ho, in advance.
Hello, everybody.
Welcome to Podcasting 2.0. This is the
board meeting.
That's where we all get together and jam,
jam, baby.
We talk about everything going on in podcasting,

(00:22):
podcast index, podcast index.social, and whatever else
is happening in podcasting.
We are the only boardroom that does not
award Christmas bonuses.
I'm Adam Curry here in the heart of
the Texas Hill Country and in Alabama.
The man who isn't Alt-Left or Alt
-Right, he's Alt-Enclosure.
Say hello to my friend on the other
end, the one, the only, the pod stage,

(00:43):
Mr. Dave Jones.
I like it.
Alt-Enclosure.
Clever.
Hey, man, I put production into this show.
I think about things.
I think about things, man.
I think about, like, what am I going
to say?
What am I going to do?
Well, you're thinking about all the things you're

(01:04):
going to say and do.
Yes.
I'm lining up pieces of chocolate next to
my mouse for a strategic time.
I always have four pieces of chocolate for
the show.
Oh, one for every 15 minutes or every
20 minutes?
One for each segment of the show.
Oh, so have you had your A-block
chocolate yet?

(01:24):
No, not yet.
It's ready.
I'm about to let her rip in a
minute.
But hey, you know what's good though?
I just, I barely made it home in
time for the show.
So I've, my lunch consisted of, well, it's,
of course it's the beef milkshake, but the
the solid material for lunch was a cup

(01:47):
of Chex Mix.
You know, it's that time of year.
Oh wait, Chex Mix with chocolate?
Is that what you're going to tell me?
No, no, no, no.
This is like, everybody has their own permutation
of Chex Mix.
Yes.
And ours is objectively the best.
But the one thing that I'm not used
to in Chex Mix, but we have interloper

(02:08):
Chex Mix.
So we have our own Chex Mix that
we made and then somebody else brought some
Chex Mix and we had a little bit
of that.
So we dumped it in with our Chex
Mix.
Ooh, combo mix.
Yes.
It's like, and there's like a, there's like
a sutterfuge of Chex Mix mixed in where
you, sometimes you'll get foreign Chex Mix.

(02:29):
And, and it's actually, and so sometimes I
don't like, but the one thing that I've
decided that I do like is, um, roasted
pecans.
Oh yeah.
Well, you're talking to someone who lives in
pecan country, so they are very good.
When you roast them, they give you a,
they, they, they like change their molecular structure

(02:51):
or something.
It's they, they're, they're like soft, but they
have a slight crunch to them.
And like, look, I mean, I, I'm a
huge fan of nuts.
He loves his nuts.
I love my nuts and what like all,

(03:13):
all different kinds of nuts, big ones, little
ones, lumpy ones, uh, with flavoring on them.
Like I, I have, I, it's one of
the staples of my diet.
And, um, but I've always like pecans and
walnuts.
This just one classification of nuts that I've

(03:34):
stayed away from because they, they're always kind
of like chewy.
Do you know what I'm talking about?
Like rubbery.
Yeah, I know.
Exactly.
I think maybe we should just stop for
a second.
Okay.
Because I am confident there are foreigners listening
to this podcast, checking into the boardroom, thinking
Chex Mix.

(03:55):
What is Chex Mix?
Furners?
Chex Mix is basically roasted salted Chex cereal.
Now, before you continue, my family tradition is
called nuts and bolts.

(04:15):
Oh, I don't know this one.
And so that's Chex Mix added in pretzel
sticks and Cheerios.
Oh, oh, oh yes.
Okay.
I've, I've seen this as a, it's sort
of like the, the twig and berries of
Chex Mix.
Yes.
But we call them nuts and bolts.
Yeah.
Twig and berries.
Nuts and bolts probably a more family friendly.
Just a little bit.
Just a little bit.

(04:36):
Yes.
Yes.
And for some reason it's a Christmas thing,
isn't it?
Always.
It's always like we, you're not allowed to
make Chex Mix or nuts and bolts or
whatever permutation you have.
You're not allowed to make it before Thanksgiving.
It is a post Thanksgiving only snack.

(05:07):
Yes.
Yeah.
You're excuse me.
I've had requests from the boardroom.
I need to play it.
I just go for it.
John, tell us your, um, peeve about the
fisting method of eating snacks on an airplane.
I see this on the airplane and it's
very annoying and I think it will result
in, in fights breaking out.
Cause it's just so annoying to watch.

(05:27):
Guy takes his bag of peanuts and throws
a pile of them into his palm of
his hand.
And then he makes a fist around the
nuts.
And then he shakes his fist to try
to bring a nut to the little hole.
And then he throws a nut in his

(05:49):
mouth from his fist.
Then he does it again.
He shakes and throws and shakes and throws.
It is annoying as hell to watch.
There you go.
JCD in rare form.
Yes.
He actually does not like that a little
bit there.
Now I can imagine that he wouldn't, he

(06:11):
would find it annoying.
Oh please.
That's just whatever side of the bed he
got up on.
It has nothing to do with it.
It's not consistent at all.
Hey, how was New York city for Thanksgiving?
Uh, New York was great.
Uh, it was, uh, Times Square was nuts.

(06:33):
Yeah.
Thanksgiving.
Uh, it was just walled.
And did it smell of weed everywhere?
Oh God.
I mean, it, I think I texted you.
It literally smelled like weed exiting the plane.
LaGuardia or JFK?
Yes.
LaGuardia.
LaGuardia.
Yeah.
I mean, we, we got off the, the,

(06:53):
the ramp coming out of the plane and
immediately in the airport was weed.
But I will say that flying on Thanksgiving
day.
Oh, is the way to go.
That's pretty good.
That's the way to go.
Cause no one flies on Thanksgiving day.
And why did you go to New York
for Thanksgiving and then fly home on Thanksgiving
day?
Because it was, uh, cause I, if I

(07:14):
waited the day after it would be horrible.
Right.
So we've, did you celebrate Thanksgiving early?
Uh, no, we just, we celebrated, uh, with
family after like the, the, the next weekend,
like this past weekend.
So like on this, on the Sunday after.
Oh, okay.
You were just in New York for a
trip.

(07:35):
Yeah.
We, it was just me and Melissa and
Marigold went up there for, uh, uh, we
were there for like three and a half
days.
Did you go to Broadway?
Did you catch a show at Broadway?
We saw a great show.
We saw, so we went and I think
it was Wednesday.
Yeah, yeah, yeah.
It was Wednesday.
So Wednesday we went to, uh, the, to

(07:56):
MoMA, the museum of modern art, which is
a fantastic.
It is.
And, um, so we spent most of the
day at MoMA and then that night we
saw a Broadway show, a Broadway show called
art.
Oh, with Neil Patrick Harris.
Oh, he's good.
He's good.
James Corden was in that too.
James Corden.

(08:16):
Interesting.
James Corden stole the show.
That guy is a fantastic stage actor.
Oh wow.
I didn't know that about him.
That's cool.
Oh, he's top notch.
He was, he was by far the best
actor on the stage.
And, um, Bobby Caravagli.
Yeah.
I don't know who that is.
Caravagli.
He's one of those guys that's been in
like tons of stuff and you know his

(08:37):
face.
Probably in all the Christmas movies we're watching.
Yeah.
And like, uh, uh, NCIS and CSI, he's
like, he's in all these things, but you
just never know his name.
Um, Caravagli.
Caravagli.
Caravagli.
Bobby Caravagli.
Oh yeah, yeah.
I recognize him.
Sure.
Of course.
Yeah.
He is in all the Christmas movies.

(08:58):
He's a mall cop.
Mall cop.
Yeah, he's a mall cop.
I recognize him right away as a mall
cop guy.
Yeah, of course.
Running around on the Segway.
Yeah.
Yeah.
Um, but it was great because you spend
all the day at the Museum of Modern
Art and then you go see the show
and the premise of the show is, uh,
uh, Neil Patrick Harris's character buys a piece

(09:22):
of modern art.
It's this giant canvas that's just white.
There's nothing.
It's my favorite.
$300,000.
That's great.
And so then, yeah, the whole rest of
the show is the, is like his, his
friends trying to make sense out of, out
of this idiotic purchase that he made.
Yeah.
It's good.
Oh, that's funny.
Oh, I didn't, I didn't toot or tweet.

(09:44):
That's right.
I told you you would forget.
I did.
Cause I was too busy with my chocolate.
Oh, he played Joe DiMaggio.
That's right.
In what?
In Blonde.
Blonde.
Yeah.
Blonde.
Blonde, uh, was the biopic, uh, of, uh,
Marilyn Monroe, but he's been in gangster movies.

(10:04):
He's been on this guy.
This is what you'd call a working actor,
man.
Holy crap.
He's been in a million things.
Yeah.
Every year he's got at least, at least
three projects going, man.
That's fantastic.
God bless him.
Yeah.
It was, uh, it was a great trip.
Yeah.
It was really fun.
Good.
I'm happy for you.
And we stayed, uh, we stayed real close

(10:25):
to central park, but the hotel wasn't super
expensive.
It was, it was nice.
Yeah.
What hotel was that?
The park central hotel.
Oh, that's a nice hotel.
Have you stayed there before?
Uh, I know all the hotels in New
York.
Yeah.
It was like, uh, affordable.
Me, me and the president Trump, we know
everything about real estate in New York.

(10:46):
Oh, that's right.
Yeah.
Bros.
Yeah.
Yeah.
The central park west, I guess.
Right.
That was, well, I think it was south
of central park.
Then it's central park south.
Okay.
As we, I don't know the layout very
well.
I mean, all I know is like we're
right below, what's the little circle that's on

(11:07):
the Columbus circle, Columbus circle.
It was like three blocks south of Columbus.
Okay.
Yeah.
You're, you're central park west then.
Yeah.
I used to live right in that very
area, also known as hell's kitchen.
Oh yeah.
So we were, I think hell's kitchen is
a little bit to the west of that.
Is that right?
Yep.
Over there where Chelsea market and all that
stuff is.

(11:29):
It's a little bit further, but yeah, the
general area.
Okay.
Yeah.
I don't know the regions there.
They say they all kind of blend together.
And did you listen, did you listen to
any podcasts while you were on your trip
or you just completely shunned the entire industry?
I don't think I did listen to any
podcasts.

(11:49):
Busted.
Busted, man.
Busted.
No, I don't think I listened to anything.
No, I did listen to, I listened to
audio book, but I did not listen to
any podcasts.
I was, I was decompressing.
Yeah.
That's good.
You, you deserved it.
Thanks.
Welcome back.
Welcome back.
The API is crashing.

(12:09):
What's going on?
Fixed stuff.
500 errors.
Yes.
That's great.
We'll have that fixed as quickly as possible.
So I don't know if that's it, but
so a lot of a couple of things
that are just really notable that's happening right
now.
We'll talk about alt, alt enclosures and video.
Cause that was pretty hot, pretty hot on

(12:30):
the, on the master.
But also what was I going to say?
Oh yeah.
Everybody, I think this is a good thing.
I'm not quite sure yet, but everybody seems
to be coming a hosting company these days.
Oh, it's rampant.
Yeah.

(12:50):
And, and I think in general, I think
that's a really good thing because it, it,
it kind of solidifies the importance of RSS.
I know some are, you know, uploading to
YouTube with an API or whatever, but, but
pretty much everybody is giving you an RSS

(13:10):
feed, whether they're good or not.
I mean, well, I mean, if it, if
it, if it passes validation, it's technically good,
but you know, I think everyone is declaring
the namespace technically fantastic, not putting in, not
putting in the features yet, but the, just
the whole general idea that you start hosting
and you have to have an RSS feed

(13:32):
that kind of solidifies the whole concept as
you know, a decentralized media network.
Yeah.
It kind of, yeah, it kind of sets
it in stone or whatever, like entrenches it
a little bit.
Yeah.
No, I agree.
I agree.
And it also, it also gives the opportunity

(13:52):
for hosting companies to show, you know, competition
like, Hey, we, we got this.
And the big thing we all know is
stats.
You got to have the right stats.
If you don't have stats, there's no good
stats, but just on that for a second
for you, for the hosting guys, man.
So we have stats with Godcaster and I

(14:13):
did our first Godcaster you for a way.
Great.
By the way, thank you for some of
our clients.
And so what we, what we give them
is a full download of, of the current
and the past month of all of the
data that we've collected, the entire log for
their particular stuff.

(14:37):
And you know, I've been involved.
I mean, I have met back in the,
even back in the on-ramp slash think
new ideas days, we would charge for our
statistics to our customers who are like Reebok
and Continental airlines and Anheuser-Busch, but we

(14:58):
wouldn't just give them some charts, you know,
pie charts and bar charts and, Ooh, look,
number go up.
We didn't give them that.
Yes, we gave them that, but we sold
it as an service, which he called ASAP,
which is an acronym for advanced statistical analysis
program.
Oh, I love it.

(15:19):
That's like a, what's the thing from war
games Whopper.
Well, yes.
So if you can, if you can name
it, you can sell it.
And, and so what we do is we
get on the phone or in certain cases
we fly to the client cause we didn't
have zoom back in the day and we'd
sit down and we'd say, well, here's what,
here's what we're seeing.
Here's the analysis.

(15:39):
Here's what we believe is going on.
And so kind of following on that idea,
because we, you know, people look at, we
have a stats page and it's, it's actually
quite comprehensive because we have plays, we have
shares, we have engagement, we have a funding
link stats.
You know, when someone hit the funding link,

(16:00):
all these wonderful things, but it's basically a
one page and I know exactly how people
use it.
Oh, number go up.
So what I, what I did is that
for this, for this university thing, for this,
you know, kind of educational moment, I said,
watch this.
And I dragged the, yeah, I dragged the
CSV file into a chat.

(16:22):
GPT is what I used in this case.
And I just started asking it questions, stuff
like, well, where was it played most?
Oh, it was played on the mobile app.
Okay.
How many times did someone listen to the
live stream?
Because we have a little lit tag in
all of our stuff and then switch over
to something on demand to listen to a

(16:43):
podcast.
And it gave statistics on that and which
shows were switched to the most.
And then it's, I said, give me an
executive analysis.
And it interestingly said, Hey, you know, it
looks like this, these shows were switched to
the most, these get the most engagement.
So you probably want to highlight this category

(17:04):
of programming, which was Bible teaching and something
else.
That's something that I think hosting companies should
explore, whether you're using your own model or,
you know, you, you tie in with an
API.
When you, when you give your, you know,
the whole start podcasting, keep podcasting.
When you, when you give someone kind of

(17:26):
free reign to ask questions.
And we, we sent all of my prompts
off to everybody so they can try it
for themselves.
It's a whole new way of looking at
your statistics, something to think about because it
really opened people's eyes.
They're like, wow.
Okay.
I can get now, of course they think

(17:47):
that, you know, the way I position it
is, yeah, we, our data set is so
rich and we're so fine tuned for the
LLM and knows exactly what to do with
it.
Well, I mean, it's, that's kind of that,
that's kind of the case though, because you
are feeding it raw, well-structured, well, like

(18:08):
named data, you know, I mean, that's it's
as good as it gets, I guess.
Like, cause you're not just saying you're not
just giving sort of like vague numbers or,
or charts or something like that.

(18:29):
You're, I mean, so what, what that data
is that you drug in there, and this
is, this is kind of an interesting thing
because, you know, like cotton gin said, how
can you trust it though?
It'll hallucinate whatever you want it to.
I am interested in how, so here's, here's
the thing I've been wondering.

(18:52):
I understand how LLMs function in, in sort
of the, I mean, I can't do the
math on them, obviously, but, but it's, it's
an understandable concept that it's a prediction engine

(19:13):
that predicts the next, you know, piece of
text.
And you can even understand like a diffusion
generator that does images where it starts with
sort of like a solid, a solid matrix
and then removes every, you know, removes everything

(19:34):
that's not a guess at what this image
might be.
So these, these, these sort of like tensor
based things are, they're conceptually, you can understand
them.
The one part of the flow I don't
understand, and this is going to, I want
to expound on this in a little bit
because this ties into the new open aggregator

(19:55):
project that I'm, that I just published yesterday.
Um, the one thing I don't understand is
how that gets translated into, into some sort
of action.
So if I say, um, if I, if

(20:16):
I say, if I, if I put a
stats file into a chat bot, LLM chat
bot and say, here's a file, analyze this
file and show, and tell me blah, blah,
blah.
Well, the, the analysis part where it's supposed
to look at the file you gave it

(20:38):
and understand what the columns are and that
it is this sort of, you know, data
representation and that kind of thing.
That's the part that I don't fully understand.
Oh, I can tell you that's not just
text prediction.
I can tell you exactly what it's doing.
Oh, cool.
And you can, and you can see this
very easily with Grok for sure.

(21:01):
I think Chad GPT because I, and I
looked at this specifically, but I've seen it,
you know, I'll give a simpler example.
Um, tell me if during the shutdown in
the United States, the government shut down, if
healthcare insurance stocks went up or down.
So there's an agentic part, which is the

(21:24):
data collection.
So it goes out, it collects stock prices
at first, it determines what is a health
insurance stock.
These things, it's pretty good at that.
Even with the predictive, you know, just understanding
language.
And it does that very quickly, quicker than
I could do it.
So it brings all that in.
And then literally, if you click that little,
you know, the little thinking thing where it

(21:45):
says thinking, you keep seeing the lines of
what it's doing.
If you click that box deliberation.
Yeah.
Well, it'll show you building code and it
literally writes Python scripts to do the calculations
that you asked it to do.
Right.
That's why I love doing my vibe coding
in Python because these things all use Python.

(22:06):
And so that's the, that's the math part
that it understands how to corroborate that to
the question, or at least it tries, then
you can easily verify it.
If that's the question that you, that you
ask.
So as long as in the dataset, you're
talking about plays and shares and support clicks,

(22:27):
that's interpretable from the data we give it
from Godcaster.
So you're saying, okay, let me sort of
talk this back to you and see if
I understand.
So you're saying that when you drop a
CSV file in there, this is a, you
know, the stats log, the, the big, the
big log that we give them.
Yeah.
It goes along with your nuts.

(22:50):
You give them a huge fat log, they
drop it into a chat GPT or whatever.
Chat GPT is, is determining that this is
a CSV file.
And so then it, in the background writes

(23:11):
code that will import the CSV file.
I think that only happens when you ask
the question.
So it sees the file.
It just sits there.
Then you ask a question and then it
creates code to calculate the question.
So it's already, and I believe that simple
questions like, give me a total, give me
an average.

(23:32):
These are all in the corpus.
It, it, you know, the, it, it can
map those things.
And of course, I'm sure I'm not the
only one who's thrown stats in there.
So it, it can take, it already has
more things that it can work with, but
ultimately the it's building code.
Yeah.
Okay.
Well that, okay.
That makes, that makes more sense.

(23:53):
Which is great.
Cause you can copy it and then run
it yourself.
Yeah.
So like, I guess, I mean, I guess
you shortcut that for common things.
Like you, if, if you're, if you know,
people are dropping CSVs into your chat bot
all day long, you probably have a, something
in place that it can just call out
and quit and not have to create code

(24:15):
every single time.
Yeah.
This is that, that I don't think.
You don't?
I've been very disappointed with iterating with an
LLM sucks.
I mean, whatever you're doing, you ask it
like, so I was building, I was writing
the outline for this Godcaster U presentation.
I love outlines.
Outlines work for my brain and you know,

(24:35):
PowerPoint literally has an outline mode.
So you can, you know, stick your outline
in there and then create your slides from
it.
So I throw it into Grok, Chad, GPT
and Gemini.
And I say, build a slide presentation.
Grok says, Oh, I can't do that.
The Chad GPT.
Oh no, I can't do that.
Gemini said, sure, here's a slide deck.

(24:57):
And it created slides for every single node
in my outline with pictures.
I'm like, wow, this is actually, you could
download it, but the slideshow worked right there
in the bra.
I have no idea how to call it.
Cause I, cause I asked it several times
in different context windows.
Hey, create a slideshow.
Didn't do that.

(25:18):
So I just stuck with that one.
Like, okay, I want this very surprised looking
person on slide seven.
I want that person to look like a
middle-aged white lady looking surprised.
And then it deleted, it deleted all the
pictures, gave me no pictures, changed text.
I mean, you cannot iterate with this stuff.

(25:40):
You can't make incremental changes.
It's, it just is not capable.
I haven't found to be capable of that.
Well, that was, that was always also my
experience with mid journey is you would get
the first picture set of pictures you got.
That's as good as you're going to get.
Yeah.
Your attempt to change it, like modify it.
It only gets worse.

(26:00):
It can't do that.
Yeah.
You know, so do you mind if we
foray into, well, actually, I think it was
just, no, I don't want you to foray.
I would like you to sashay.
I would like to momentarily sashay into the
feed parser, but yeah, before you, before we

(26:21):
do that though, you were talking about hosting
and Sam had a question.
Ah, yes.
You heretic.
Like, so Sam's question was if the feed
has an HLS video in an episode as

(26:46):
an alternate enclosure, should you play that first
instead of the audio?
Before, before I give my sort of, before
I give my opinion, what, what do you,
what do you think?
I think I already know it, but what
do you think?
Well, I liked Oscar Mary's choice in this,

(27:07):
which came after my, my vote.
I was the only person that voted when
I vote.
I'm like, oh, a hundred percent.
Oh, you were the one.
It's just me.
It's just me.
The decision that Fountain has made is if
it's mobile on the app, it's audio.
If it's on the web player, it's video.
I like that a lot.

(27:29):
Yeah, that's a good, yeah, I can see
that.
Now what you said, you heretic, you said,
oh, we should probably do it, do video.
Let everybody know we're doing videos.
I did not say it in that voice.
Yes, it was that voice.
I heard that voice.
I did not say it in that voice.
We should probably get, you know, a podcast
industrial complex, everything's YouTube video.

(27:51):
We should do video, make a video.
My keyboard won't even do that voice.
That's not what I did.
That's how I heard it.
What I said was, at the risk of
being a heretic, I'd say play the video.
I said, here's my thought.
Right now, if somebody goes to the trouble

(28:14):
of putting that in their feed, they probably
want people to see the video because it's
such a, it's so rare.
And so I'm like, for now, it's probably
a good, a good signal that somebody wants,
like they want their video out there in
the world, but like long term.

(28:36):
Hmm.
Well, I mean, I'm going to agree with
you, which should sound like this to you.
I'm going to agree with you, baby.
That's the way I heard it.
I'm all in with you.
That's the way I heard it.
I think it is it because I don't
do video, but if I were doing video,

(28:57):
I'm 95% confident I would want video
to surface first.
Yeah.
Yeah.
Because you put a lot of effort into
it.
Yeah.
Yeah.
But then unless you're me and then it's,
but then it never even works.
Half the time it doesn't even have audio.
Are you even doing OBS anymore?

(29:19):
Not right now.
I forget.
This is the problem.
I just, I forget.
Every week I forget to do it.
I've forgotten to hit record, you know, stuff
happens.
It happened not too long ago, actually, when
I forgot to hit record.
And the one, like, I know, I just

(29:39):
can't, there's so much stuff that's broken that
I can't, that I can't get to.
I mean, my umbral node is broken.
Yeah.
You're hurting, you're hurting the business, man.
We're not getting any, we're not getting, you're
hurting the show.
We're not getting any sats.
I know it's, it's, I feel horrible about
it, but the umbral node is that's, that's

(29:59):
a lot of fun.
It's a lot of fun to debug.
It's all Docker containers.
Oh, I know.
And hell, and I've, I've done so many
times, all of a sudden, the whole thing
quits.
I start it back up and it says,
Oh, this Docker already exists.
This Docker already exists.
So then you have to pseudo, uh, stop,

(30:19):
pseudo stop Docker, paste Docker ID, pseudo remove
Docker, paste Docker ID.
You've got to do that for about 10
of them, depending on how many apps you
have installed.
And then it'll start back up.
What I don't get is, is I did
nothing.
I did nothing to this, to this umbral.

(30:41):
That's a feature of umbral.
It's always exciting.
Yeah.
You don't know what you're going to get,
but now when I go to the local,
you know, local host on my umbral, I
get the default Apache two.
It works page.
Oh goodness.
Yeah.
Yeah.
Now it's just an Apache default.
Okay.

(31:01):
What's happening here.
I did nothing.
And well, all I did was install the,
uh, what do you call it?
The, the Albi hub.
Yeah.
Try 127.0.0.1. That made you
better than local host.
I've had that happen.
On the umbral?

(31:22):
Yeah.
Instead of local host.
Yeah.
Same thing.
Apache, Apache default page.
Unplug it and plug it back in.
All right.
See you.
This was terrible.
This is not a good product.
No, no, it's not.
Start nine.
Baby.
That's the way to go.
I was waiting for your ad.

(31:43):
I was waiting for the ad.
Oh yeah.
Okay.
They did give me a box.
That's for sure.
They gave me a box and that's what
helped.
And that's, that's how we got helipad in
there.
No, actually it's not the one terabyte.
I'm not, it's not a dry space issue
because I did fit.
It was getting close and I fixed it.
I cloned it onto a new unit.

(32:03):
It's all, that's been fixed for a while.
Um, yeah, I don't know.
That's, it's, it's, it's a pain in the
butt.
It's a pain in the butt is what
it is.
Yeah.
Um, which, which I seem to be using
that term a whole lot lately.
Well, I can't believe that your day job
is all of a sudden so painful.

(32:24):
Isn't it supposed to end or do we
have another rush towards January 1st or December
31st?
Uh, no, December is my busiest month of
the year.
Oh, what happened?
It's April, October, December.
Yeah.
This April is a lot of after hours
work.
Um, then, uh, you know, then October, October,

(32:46):
same way.
But December is installing all the new year's
tax software and onboarding interns for the new
year so that it's a lot to do
in a small amount of time.
Got any hot interns next year, Dave?
I wouldn't know.
I would not know.
Oh, you're just onboarding them.
Yeah.
We're just all I have is names, which
is funny.
Like the Nick, this new generation, um, that's

(33:10):
coming out of college, the names, very confusing.
Oh, they spell them in weird ways.
Yes.
They spell them in weird ways.
It's, uh, girls with boy names, boys with
girl names.
Like you can't tell who is, is what.
Do you have to, do you have a
pronouns in your corporate profiles?

(33:33):
Uh, no, we do not.
Do you, do you like, it's like the
name, uh, I'm trying to, I'm trying to
think of one I did in the last
few years.
Vivian.
Vivian, that one.
No, like Alex.
Oh yeah.
Oh yeah.
With an I.
Alex with an I.
Yes.
Yeah.
That kind of stuff.
Yeah.
Anyway, young, young people.

(33:55):
Yeah.
Yeah.
They're, they're, uh, they're a waste of time.
When I was a kid, we were just
called Bob and Alice.
Yeah.
Although I did see a resurgence of old
names though.
Yeah.
Like Mohammed.
Older.
No, no, no.
Like older names.
Like, uh, like Jim and Bob and stuff

(34:17):
like that.
That's they're making a comeback.
Hmm.
That's cool.
So let's talk about your, um, your Aggie.
Yeah.
So, um, what have you done?
So I guess one of the, firstly, I've

(34:39):
realized that I can't build everything myself to
spread out.
So I would like to, but I do
have a vision for, for the way that
I think it was things should work.
And I want to try, I would like

(35:01):
to try something.
Um, and I hope that this gets a
lot of, um, support traction.
Yeah.
I hope, I hope people PRs will spread
the word.
Yes.
I hope people will spread the word to
get other people involved.

(35:22):
Um, one thing I'm really terrible at is
getting help with things.
So like sysadmin stuff, developer stuff, and, and
mostly it's because my time is so limited.
It's so crunched that it's, it's really hard

(35:44):
to find like an hour for a zoom
call to even explain to somebody what I
need help with, because then you have to
go through, this is very similar to, okay,
this is very similar.
Thank you, Eric.
I appreciate that.
Like Eric has offered to help so many
people have offered to help over the years.

(36:06):
And, and like, and I feel bad for
not engaging more with them.
Like Archie has asked to help multiple times.
So many people have it.
The, the problem is this is very similar
to the way, like if you have a
piece of software that you've been writing in
private for years and years and years, and

(36:27):
then you want to open source it, it's
scary.
It takes hours and hours of work to
go through and make sure that you are,
that you don't have other proprietary code that
you can't, that would be illegal to release
to make sure you don't have hidden secrets
in there that, that shouldn't be exposed.
Um, like, uh, hard coded passwords.

(36:51):
Some, some things are just, uh, you want
to make sure that you don't have code
that's also used in another place that exposing
this code would also expose the backend of
a running software that doesn't need to be
exposed.
So that now, because you used the same,
uh, low level, uh, function in two different
places, you, you, you make it public here.

(37:12):
Now they know that that's the same function
that you used in this other code that
is not public.
There's so many things that you have to
do.
It takes hours.
This is very, it's in a very, in
a similar way, getting help with a project
like podcast index means spending a lot of

(37:33):
time with a person explaining to them what,
how the system works, giving them, um, credentials
to get into the system, but just enough
credentials because you're not, you don't be handing
out root to the whole world.

(37:53):
So you have to, it takes a lot
of time to bring someone on board to
help in a meaningful way with these things.
So tell me you set this project up
correctly from the beginning.
So here's the idea.

(38:17):
The podcast index, it works, it works fine,
but, Oh, the dreaded F word.
Whenever my wife says, yeah, I'm fine.
It's no good.
You got to step it up.
It worked.
It works fine.
But the, um, how should I say this?

(38:42):
It, there is the design of it goes
way, goes back to the very beginning of
2020 when we built this entire thing in,
uh, about six weeks from scratch, literally six
weeks.
And it was, I mean, it was stupid

(39:04):
fast and we did, it was the, the
goal was get something to work.
So, and then once you have something that,
that works, now you've introduced technical debt for
the rest of your life.
Technical debt.
Exactly.
So we, the technical debt in the podcast

(39:26):
index is a feature, uh, that is in,
and I'm, I've been okay with our technical
debt because it's not the type of debt
that comes back and sort of like bites
you in the rear in the future in
a way where like you, you get stuck

(39:47):
and you can't move forward.
It's not that type of technical debt.
Um, the, this time, this type of technical
debt just means that when things that things
are just not as efficient as they could
be.
And so that touches on the second leg

(40:09):
of, of podcast index, which is always the
financial issue.
We got to run things as cheaply as
possible.
So, um, our type of technical debt just
means that in order to kind of scale
up, we have to throw more money at
it.
And that's the thing we can't do.
So rather than trying to, I've been, I've

(40:32):
been thinking about this for weeks now, rather
than trying to onboard a bunch of people
into sort of a, a part-time sys
admin role.
And again, I'm not opposed to that.
That's not, I'm not opposed to it at
all, but that is very time consuming.

(40:52):
I would rather, and I, in addition to
doing a little bit of that, I would
also like to sort of allow the community
to build new versions of the podcast index
to build the next generation podcast index themselves

(41:15):
in public.
And that way they already know how the
system works just because they built it.
And so as we build it together, all
together, we can document the way the pieces
fit.

(41:35):
And I can bring in somebody like Archie
to begin to put these new pieces together
from the beginning.
So you, we sort of build, we build
the community sys admin effort into the new

(41:56):
product.
I like it.
Yeah, no, that makes a lot of sense.
Uh, so the first sort of attempt at
this is the feed, the feed parser repository
that I just put up.
And, um, it is, so there's, there's a

(42:22):
few things that we need to go over.
I will stop now to ask for, to
pause for any questions you have.
Uh, nope.
Okay.
The, so the first thing is we're going
to build it.
Clearly we're going to build it in the
open.
So this, this will be the first piece

(42:46):
of software that will live in.
We'll do this in some, we'll do this
in phases.
Okay.
And I need to, I need to document
this whole, there's going to be, there needs
to be a repo for this whole build
process and we'll do that.
I'm going to do that too.
That's the next thing I'm going to do.
The, so the way this will sort of

(43:08):
this phased approach, I think I envision it
going like this.
First thing we need is an aggregation stack
and the, the, the aggregation stack
will look like this.

(43:30):
It will, it will be a feed polling
agent and we already have one that's, that's
pretty good.
I think it can be improved, but aggravator,
which is our current feed polling agent, is
a sync, is an asynchronous, uh, feed poller
written in rust.
It's very fast.
And I think it, I think it works

(43:51):
very, I think it works good.
Um, that thing outputs it, it, it download,
it downloads the files that the XML feed
files and saves them as files on the
file system with some metadata attached.
I think we're good there, but we will
bring, we will bring that along as well.

(44:11):
The, this repo I just posted, uh, yesterday.
Did you post it on the podcast index
GitHub or?
Yeah, it's on, uh, there's a link to
it in the IRC.
Eric PP posted it in the chat room,
the boardroom.
Uh, so then, so the feed parser, what

(44:35):
it's going to do, and here's the goal
for this repo, uh, it needs to be
a rust, uh, it's a rust built process,
a single process that will, that will be
launched.
It will pull in all of the XML
files that were downloaded by aggravator.
It will parse them using an XML stream

(44:56):
parser and then output a, uh, a text
file that contains a Jason representation of a
SQL insert or update statement.
And, and I say like a representation in

(45:22):
quote in quotes, it's not, it's going to
be some, it'll be an object that can
be easily assembled into a text SQL statement.
So that's, it's meant to be run, it's
meant to run and then exit, not be
a continuously running process.

(45:43):
Well, that, that you just lock me out
of developing for it.
Okay.
Sorry.
So it will, it will execute, you know,
on a timer or whatever.
The inputs folder is where the XML feed
files from aggregator aggravator are going to live.

(46:03):
It will pull those files in process them
using the stream parser and then output a
Jason file.
And if we can make it have some
sort of plugin architecture, that'd be great.
But that, that's, that's not, so this is
the, uh, no e-tag URL and then
index ID.

(46:25):
Yeah.
So that's the, the file names will have
a pattern.
Yeah, exactly.
Okay.
Um, and then, so that is the feed
parser.
So what you're going to be left with
once you're going to be left with after
the feed parser runs is an outputs folder
that has a bunch of Jason files in
it that will then be picked up by

(46:48):
the SQL statement builder process.
And that will run on each aggregator node.
So there, this is the next piece of
the puzzle.
You'll have a SQL statement builder that will
read the Jason, uh, object, build a SQL
statement from it and stick it into the

(47:12):
queue server to stick it in the queue.
And that'll be the next piece of the
puzzle.
There'll be a queue server.
Now this can be something as simple as
like Redis or something like that.
I mean, this, this doesn't necessarily have to
be something we build, but there needs to
be some sort of queue server, um, that
is, that lives in this, in the infrastructure

(47:33):
where all of these SQL statements are going
to be placed in a first in first
out fashion.
Then you'll have a, the next piece of
the stack will be a SQL execution agent,
which picks objects off of the queue in
FIFO order and just executes them to put

(47:55):
them into the database.
Um, now this SQL execution agent that is
plucking off the queue and executing this thing
needs to, it needs to monitor a few
things.
It needs to monitor the process, uh, excuse
me, the, it needs to monitor the, the

(48:19):
database server to look for how much, uh,
load it currently is experiencing and adjust its
insert, update, delete pressure accordingly.
So you don't, excuse me, you don't want
to have it just slamming the server.

(48:40):
You already have an endpoint for that, for
that, for that server load?
No, I think, I think we're going to
have to tap into whatever the underlying database,
uh, metrics are in that the database server
provides.
Now, um, I have used MySQL forever.

(49:05):
It's the only database server that I've, well,
I mean, I've also used Microsoft SQL server,
but I've, I've never used Postgres and I
know that Postgres is what everybody loves and
I'm fine with switching.
I will just have to learn this new
thing.
So if, um, you know, if we want,

(49:28):
so the reason I say that is initially,
I think we should start by just replacing
the aggregation stack and have it tap into
our current database.
Can I ask a question?
Okay.
What are the intended downstream benefits of this
new aggregator?
So the way the current aggregators work is

(49:51):
they, uh, the, the parser and the SQL
insertion functions are happening in the same process.
So the node, the, these are node JS
processes that run, they pick up, they pick
up the feed files, parse them into a,

(50:14):
into a JavaScript object in memory, and then
convert those objects into SQL statements that are
then executed.
So there's, uh, there was 10, there's nine
aggregators.
Currently there was 10.
Now we dropped a nine.
There's nine aggregators and they, all nine of

(50:37):
them are constantly connecting to the database to
write to it.
So none of them know at any given
time how much load there is.
They're all just competing to put their stuff
into the database as fast as possible.

(50:58):
And this means that it's the load on
the server is, is very spiky.
It just goes through surges, um, all the
time and I'm constantly having to sort of
tweak it and massage things to make sure,
uh, you don't, you don't get log jammed.

(51:21):
Yes, exactly.
Right.
Yeah.
Um, yeah, Eric, it could do your exactly.
It could do that too.
Yeah.
We could just cue the JSON rather than
SQL.
It's fine.
Can I have another question inserted here?
Yes, sir.
With a new aggregator aggravator stack, will we
be able to detect changes made to older

(51:42):
episodes or feed items like image tag, et
cetera, and reparse those?
Yes.
That's what I call a downstream effect.
Yes.
So the, and, and the reason, the reason
is because if you have all the SQL
adjusting, um, actions, if you have all the

(52:03):
SQL actions being queued, then you're not where
you'll, you can adjust the load accordingly.
Now you, you know, queues also have problems
too.
I mean, none of the, all, everything has
trade-offs.
And the one thing you can say about
the current model is it's fast.
It's very fast.

(52:25):
Yeah.
And it is, um, it, it's fast at
the expense of sometimes bringing the server like
to, to its knees.
I mean, it can just get, if we
spun up part of the reason the database
doesn't actually just crash is because, or lockout

(52:48):
is because we have underpowered aggregators.
If I, if I boosted the horsepower on
all the aggregators to let's, you know, to
beefier VMs, it could pump so much data
at that database.
It would just lock it.
Right.
So the queue thing is really good for

(53:09):
being able to just feed the database with
a, at a predictable rate, but predictable is
also a synonym for slow.
Hmm.
So we need, um, so it is going
to come at the expense of a little
bit of slowness, but that doesn't mean like

(53:30):
the data has been aggregated.
It just, it's going to take a little
longer to get updated in the database.
Yeah.
Right.
Right.
Right.
Right.
Right.
So, so that's the downside, but I, I
think the trade-off is worth it.
Um, and I, and when I say slower,
I don't mean like significantly slower.
I mean, I'm talking about like the difference

(53:51):
of a few minutes, right.
You know, versus a few seconds.
So if something shows up five minutes later,
then it would have under the current system.
I think that's a, that's okay.
That's acceptable.
There goes my pod ping pitch.
90 seconds, everybody.

(54:12):
90 seconds.
Your podcast is updated, but you know, we
don't have to treat everything as equal also.
Ah, okay.
No, no agenda show plus five plus five.
Yeah.
Yeah.
Um, you know, so if certain things like
live, live items, or we could have two

(54:34):
cues, you know, one, one for, uh, one
for live, one for paying, one for paying
customers and one for the rest or more.
I mean, you can have as many cues
as you need.
Like you can have, you know, a live
queue, a standard polling queue, and then a
pod ping queue.

(54:55):
So we can configure this however we need
to, but this overall, this is a better
architecture.
Yeah.
And so I think phase one is build
the feed parser, uh, in such a way
that it, um, will run and push things

(55:17):
into our existing database.
And then, um, and then once we have
that, we can look at potentially modifying, changing
the database.
If that's something that, that we all feel
like is a good thing to do.

(55:37):
And, uh, you know, I'm fine.
I'm, I will eat it.
I will eat my discomfort.
It's fine.
So we can do, we could do that.
And then I think phase three would be,
uh, build a new version of the API.
And everybody has, everybody can, can take a
hand in that and we can all build

(55:58):
a version two of the API together.
What, uh, what improvements or what things are
you thinking about the API?
Just more functionality, more different, different types of
endpoints.
Yeah.
Faster.
Um, the having some things that the index

(56:18):
doesn't have now, like paging.
Oh yeah.
Paging is a big one.
Yeah.
So just, this is mostly, this is not
so much because things don't work.
This is more of, um, attempt to bring

(56:40):
everybody on board, but in a way that
they under, that they built it themselves.
So they already know how to administer it.
Ooh, I like that.
And please put in hooks everywhere for my
potty bot.
Yes.
Version two of the API will be potty
friendly.
Yes.

(57:00):
So the potty can go in and help
you.
One moment.
Fully.
One moment.
Let me check.
Let me check that for you.
Oh, it looks like you're in q5 deposit
$100 in Bitcoin to move to the top
of the stack.
Thank you for using potty bot.
It was an RBF that replaced by fee.
Yeah, yeah, yeah, exactly.

(57:22):
Yeah.
R I Q replacing Q.
Yes.
Uh, yeah.
So everything, it'll be, it'll be fully potty
trained.
And, um, that's the second part of this
is that everything in that repo right now
was built with AI.

(57:43):
I have not handwritten a single piece of
code in that repo.
Really?
You using, uh, um, uh, what's that guy?
His system.
Who, uh, uh, Jason, Jason.
Yeah.
Jason from podcast guru using his, uh, his,
um, infrastructure setup.

(58:04):
I'm using his, um, his advice, but I'm
using, um, jet brains.
They have a new thing called Juni.
It's their LLM Juni meet potty.
There's a perfect match.
And so, uh, G Juni is integrated into

(58:26):
the ID.
And then you can just, uh, you can
just tell it, it sees your whole project.
You can, you don't have to put everything
in one file.
So what has your experience been with Juni?
Uh, I, I love it and hate it
at the same time.
Hmm.
I love the fact that I can, uh,

(58:48):
have it do a bunch of the boilerplate
sort of drudgery code for me, but I,
I also despise the feeling that I don't
understand what, what I just did.
Like, I feel so mentally disconnected from the

(59:10):
code.
Like you can, for instance, you can go,
so, so somebody could clone this repo right
now and you could get into, um, I'm
just going to tell you from the, from
the jet brains perspective, you could get into,
you can open up your jet brains.
Idae a rust rover is what I use.
You could open up rust rover, go into

(59:31):
the Juni agent and say, um, and say,
add support to this, um, add support to
this program for the podcast colon transcript tag.
And you'd be 95%.

(59:53):
Probably it's going to work the first try.
Wow.
It's really, really good.
But then you have no idea what it
did.
And you know, it works because it compiles
and it gives you output, but then you
have to go through there and figure out
what it's doing.
Yeah.
And because otherwise it's just this opaque.

(01:00:16):
I do find that when I'm vibe coding,
uh, it documents things very meticulously, which for
me is like, Oh, okay.
I see what this block is doing.
I get what you're doing here.
I can, I can handle it when it's
documented properly.
And that's probably the having worked with developers
for most of my professional life.
That's probably been the biggest bone of contention

(01:00:38):
amongst developers as well.
It's like, what do you do?
Can't you document your code, document your code,
man.
Oh, the worst experience with that I ever
had was trying to modify, um, frontier, the
compiled version of frontier.
There was nothing, nothing, nothing documented.

(01:01:00):
No, not a single comment in the entire
code base.
And I was like, every now and then
you'd run into a slash slash DW and
a date.
Yeah.
You're like, thanks Dave Weiner.
That doesn't help me.
Um, so yes, it's doing a pretty good
job.
And I've also had to be explicit about,
I will say add support for the iTunes

(01:01:23):
colon block tag and document your changes in
the code.
I'll say something like that.
And it's pretty good about now.
And I want this thing to be LLM
friendly.
So that means people are welcome to, for

(01:01:44):
their, to, when they, when they submit PRs
on this thing, they're welcome to just go
nuts, use every AI tool, LLM tool, whatever
you want to use, because this is all,
this all needs to fit into our lives.
This needs to be easy, easy on our
time.
And we can, this is another goal with

(01:02:04):
this project is to, is for me to
force myself to become comfortable with these tools.
And the only way to do it is
to do it.
Uh, so, um, people are welcome to, to,
to change the, to make this code.
It's compilable.
You can download it, run it right now.

(01:02:25):
And, and it will work.
There's sample, there are sample files in the
repo.
So there's two folders.
There's a sample inputs folder and a sample
outputs folder.
And you can see, you can take all
the files from the sample inputs folder.
And those are just XML feeds.
You can, you can copy those over to
the inputs folder, cargo run to build and

(01:02:45):
run the, the, the process.
And it should give you an outputs folder
full of JSON files.
Now you, what IDE are you using just
out of curiosity?
Uh, it's called Rust Rover.
Okay.
Yeah.
It's from JetBrains.
And, uh, it's, I will have, I have

(01:03:07):
to say, I have not used a whole
ton of them, but as far as like
the way I like to work, having this
LLM agent, like baked right into the IDE
is pretty rad.
Uh, it, it feels really nice to just
say, um, you know, go, um, like build

(01:03:29):
a new module for blah, blah, blah.
And it will create the file for you.
It'll take, you know, put your class in
and all this kind of stuff.
Like it re having, having the agent have
control of your IDE feels like finally the
way that it makes sense to me to
work.
Um, now the, the one thing I wish

(01:03:58):
it did, and it may do this, but
I just haven't seen it yet.
And, and I'm still new at this is
I w I do wish that it would
give that there are some way to have
the get commits also include the prompts that

(01:04:22):
were used to create the code.
So like, because like I said, I've not
written, I have not handwritten a single part
of this yet.
I just know what I want and I
know how Rust works.
And so it's, it's easy for me to
tell it what to do.
Um, but, and I can tell if it's

(01:04:43):
doing it in, in, in the right way,
because I already know what it's supposed to
kind of look like.
So I can say something like, um, you
know, the, the initial version of this was,
I was just, I just started, you know,
saying what I wanted, you know, create a,
uh, create a process that opens each file

(01:05:04):
in the inputs folder.
Uh, these files have this format.
I want it to be parsed with this
part, with this, uh, crate and the outputs
stream to this other file.
I just started describing and it was slowly
building this thing.
And at one point I said, okay, now,
now modify this, uh, the current code to
be a, to, uh, process the files asynchronously

(01:05:27):
and it rearranged all the code, made it
all, made it all async await and, and
all that went fine.
But it really would be a, a nice
thing to have is a track record of
all the prompts that I used.
And as we go forward, if somebody adds
support for something, it would be great to
have those prompts in a, in a place

(01:05:47):
where we can find them.
Sure.
Sure.
And so I just haven't explored that part
yet.
So I don't know if that's something that's
easy to do or not.
Um, cause I would like everybody else to,
to build this with LLMs as well.
Let's, let's just try to, let's try to
do something crazy here.
Um, I love it.

(01:06:08):
I love it.
Sounds good.
This could be the, this could be the
first crack at it.
And then if this goes well, uh, and
people are contributing code, um, and it's building
and it's, and it's, it's working, then, um,
yeah, that we just go into the next,

(01:06:28):
go into the next step and just start
building these pieces one by one.
And, and you know, the, the thing that
gets you with all RSS aggregators is the
edge cases.
Of course.
And those are well documented in your brain.
And in party time.
So we can, we can go through and

(01:06:51):
sort of pre, uh, pre pre, but we
can kind of like save ourself from the
broken stuff ahead of time, you know?
So we don't have to build it and
wait for everything to break again and then
fix 25 different things we've already fixed before.
Right.
Um, but you have to go through that,
you know, we just, we just have to

(01:07:11):
go through and that'll be something to do
for me as I go through the party
time code and look and say, okay, here's
where I made a, uh, special workaround for
this edge case.
Here's where I did this edge case.
And we can document that.
And then yeah, exactly.
Chat pre pre break it and then, uh,

(01:07:32):
and then fix it before it bites us.
So I don't know.
We'll see how this goes.
I love it.
No, I think it's a great idea.
I'm a, I'm installing a jet brains pie
charm as we speak for myself.
Pie charm is a great tool.
I think that's what Alex uses.
Well, it's time for me to get into
an IDE.

(01:07:53):
You, you need to graduate.
Yeah.
You need to have an Yeah.
I've been resisting it, but I don't like
the feeling of being mentally disconnected from the
code.
I don't like, I compare that to Godcaster
and podcast index code where I'm, I may
not remember exactly what the code does, but
I have a mental picture all the time

(01:08:15):
of where every single thing is.
Of course, of course.
I know exactly what it's doing and how
it works.
And so I can usually go to that
code within, you know, 10 seconds.
Right.
Have an idea of what the problem is.
Yeah.
And good logs, good logs, good logs is
a good thing.
Yes.
Lots of logs.
Yeah.
And so I don't like that feeling.

(01:08:37):
It's, it's very, uh, it, it feels like
wearing glasses and somebody's kind of smeared vaseline
all over them.
Uh, looking through a glass darkly.
Yeah.
Yes, yes, exactly.
And that's not a great feeling, but I,

(01:08:57):
I mean, I don't know, maybe, maybe at
some point you can, you can go through
and, and have the LLM comment things out
in such a way where it's clear.
I don't know.
I'm going to have to, I'm going to
have to tackle this problem at some point
for my own sanity, but it is working.

(01:09:20):
Love it.
I think it's a really good idea, Dave.
And isn't that, isn't that really the world
we live in now with LLMs?
It's like, well, it's working.
What's it doing?
I don't know, but it works.
Yes, it's true.
That's true.
But I, I have a pretty good idea
what I've built.

(01:09:41):
Yeah.
Um, but when it, cause you know, I,
I'm not yet, I haven't graduated yet.
So I'm just, here's one big file, 4
,000 lines.
Enjoy.
And then it'll fix the problem, but then
something else happens.
Like what did you do?
Then it won't tell me until I say,
Hey, this is broken.
Oh yes.

(01:10:02):
You're right.
Good catch.
Oh, that's see, that's interesting.
Okay.
It's interesting to hear you say this because
like, if I wanted to know what it
did, I would just do a diff and
then I could see, but, but you're not
thinking of it in the code terms.
I'm lame.
Could you, could you do, could you say,

(01:10:24):
make a, uh, make a diff and then
explain it?
Well, I can, I can do a diff.
I mean, I never, I never even thought
of doing that.
Honestly.
I was like, but if you had it
explained the diff, like, Oh yeah, but yes,
but the point is, is like, I'm not
expecting it to do anything.
It runs.
And then I'd find later, Oh, this thing

(01:10:46):
broke.
And then, then now you're into iteration mode.
Well, go back and fix that, but leave
this.
And then something else breaks.
Now that you might as well just start
a new session, paste the code in, start
over and it'll fix it within five seconds.
If it's, if it's in its context window
to get so confused after a while, all

(01:11:07):
of them, all of them have the same
problem, but you can take something that's broken
in grok, put into Chad GPT and Chad
GPT says, Oh, you were almost there.
Here.
I fixed it.
This is really, it's really messed up.
I know people love cursor Daniel and Eric.

(01:11:29):
I know people love cursor a lot.
I just, I've been in jet brains world
for, for, for a decade.
And I just, it's how I think I
don't want to change IDEs.
I can't, I'm just too stuck, man.
I'm too stuck having to learn two things.
This is always the problem with, I think
anything in it, developers developing or, or says,

(01:11:51):
I mean, if you have to learn two
things at the same time is too much.
Yeah.
You really need to do one thing at
a time and, and learning this, learning how
this stuff is working.
That is enough for me.
You know, I'll, I may tackle some, you
know, something like that later, but yeah, we'll,

(01:12:13):
we'll, we'll see how this goes.
Um, let me see, where are we at
time-wise?
We got a few more minutes.
Um, did I see that, um, our boy,
uh, Spurlock, he is, uh, um, basically filtering
out TTS podcasts on OP3.

(01:12:37):
TTS podcast.
Yeah.
It took me a second.
I'm pretty sure that stands for text to
speech.
As in AI.
Yeah.
He says there's been a deluge of TTS
podcasts.
And unless that's some kind of network I'm
unaware of, I'm thinking it's a text to
speech.

(01:12:58):
Um, I don't know.
I'm at where, where did he post this
on podcast index social?
Yeah.
Yeah, he did.
It's today.
I'm looking right now.
I'm curious what that means.
Uh, let's see.
Uh, C to C top CDN.

(01:13:24):
Oh yeah.
Mass mass produced TTS podcasts.
Love DAI, especially for gambling ads here in
New Jersey.
And they all seem to use gotta be
it.
That's gotta be it.
New data updated my ranking of the top
third party analysts or share of new episodes
measured by number one pod track down main
do mainly to my better filtering of TTS

(01:13:46):
podcasts.
TTS podcast.
I'm pretty sure it's gotta be text to
speech.
Yeah.
But what in what, well, of what, what
is the text coming from?
Oh, no, it's just a, that's just his,
his version of saying these are AI podcasts.

(01:14:10):
Oh, Oh, is that just, is that just
a euphemism for AI?
I think so.
Yeah.
That, that was kind of my question.
Yeah.
Mass produced TTS podcasts.
So I want an example of what those
are.
Cause I mean like certain, if it's, if
there's certain things that qualify as like legitimate

(01:14:31):
garbage, I will take them out of the
index.
I mean, if it's just, if it's just
like episodes of, of some bot reading the
phone book, like that's here it is under,
under his a live wire update, November, 2025
TTS podcasts, which is not Tourette's syndrome.

(01:14:53):
That's what I thought first.
Like, wow, cool.
I need to be in that group.
Turbo Tourette's there's been a recent market market
increase in the number of podcasts across multiple
hosting companies that are completely generated by text
to speech engines.
Known TTS podcasts are observed, but excluded from
this analysis as they are not a reflection
of the health of the podcasting world in

(01:15:15):
terms of human output.
Filtering these out is a significant amount of
work.
And I thought about stopping this monthly analysis
altogether or making it paywalled or charging and
changing the metric entirely.
But ultimately I want these numbers to be
available to everyone to be as consistent and
as regular as possible over time.
You'll notice a one-time drop in pod
track share in November, since most of these

(01:15:36):
generated TTS podcasts tried to monetize by ad
insertion and use pod track to measure performance.
How about that?
Okay.
So this is interesting.
So would that be inception point?
Yeah.
I don't know.
Is that, is it that type of stuff?

(01:15:58):
Yes, I believe so.
I think that's exactly what he's talking about.
Here's some inception.
Here's bass fishing daily.
Yes.
I think that's definitely one of them.
Yeah.
With 12 year old Wyoming angler lands world
record large mouth on tungsten jig.
Okay.
So let me go into the feed and
see if it's pod track.

(01:16:21):
Where's the enclosure?
Oh yeah.
Yeah.
That's pod track.
Yeah.
Okay.
So I wonder how he's filtering these.
I wonder what he said.
It's a significant amount of work.
I wonder if he's having to use some
sort of heuristics or something.

(01:16:41):
Yeah.
I want to talk to him about this.
Cause if, I mean, if, if we're going
to be, if we're, if we're collectively saying
that these things need to be filtered out
of things like stats.
Well, I think he's saying it very clearly
for himself.
Cause he's looking at the health of human

(01:17:03):
-based, uh, podcasting.
Yeah.
Do we need to filter these for real?
I don't think so.
I don't think so at all.
So you think this is something specifically to
him?
Yeah.
He just, what he's wanting.
Yeah.
He says, I just want to show the
health of human-based podcasting and that's why

(01:17:25):
he's excluding them.
But I don't think that makes any of
these podcasts invalid.
No, no, I don't think so.
I mean, we're literally talking about writing code
with AI, like, Oh, I'm sorry.
We can't, that's, you know what I mean?
It's like, it is what it is.
I'm, I'm, I'm a lot less, I find
that a lot less problematic than other things

(01:17:47):
taking place in the world.
For sure.
For sure.
I'm, I've been going around and around about
this kind of thing in my mind for,
for a long time now.
And, and I don't, so it comes down,

(01:18:08):
so many people are troubled by, by AI
and it's, it's hard to determine like the
actual arguments for and against the for AI
are, are, are easy.

(01:18:29):
They're sort of like the Fisher price version
of, of rhetoric.
It's easy.
It's like, Oh, it makes, you know, helps
me do something I couldn't do before this
way.
It's, it's cheaper because blah, blah, blah.
I mean, they're just, they're just, they're effortless
arguments.
I'm not interested in those arguments.

(01:18:49):
I'm more interested in arguments against AI generated
content, not because I'm trying to find a
reason not to like it or, or it's
not a, there's not a bias there.
I'm just interested in good arguments.

(01:19:13):
I'm interested whether, let me see how to
explain, how to sort of phrase this.
It's always interesting to me to explore ideas
where you can't really put your finger on

(01:19:35):
a good argument against it.
And what I mean is like, uh, every
argument you come up with, you sort of
talk yourself out of it.
You're like, you're like, well, but I could
see it going the other way.

(01:19:56):
And you're like, you're like, Oh, you're like,
okay, well, AI generated kind of, you know,
AI takes away.
It's a threat to people's jobs.
And then you're like, well, it is it,
it's not really.
And you're like, okay, well, AI content is,
um, it's filling up the web with, with
crap.
And you're like, well, well, but is it

(01:20:16):
really?
Yeah.
The web is the web is crap.
I mean, yeah.
Listen, half the news reports, if you, I
watch Euro news and I watch Africa news
and I watch, um, Deutsche Welle and France
24 and I hear AI voices all the
time, all the time.

(01:20:37):
Like where do we draw the line?
I mean, I've built entire radio stations for
customers with AI voices.
Doesn't invalidate it, you know, now the scammy
part is, I mean, just, I hate dynamically
inserted ads in general.
I just think it's crap.
Half of them I get over here in

(01:20:58):
Spanish, like, Oh, Oh, he's in Texas.
Must be, must be a Spanish guy.
You know, I was like that, this it's,
there's all kinds of, and it's not relevant.
It's always gambling, you know, or tequila talk
about racist.
Yeah.
And that's, that's kind of where I'm, that's

(01:21:23):
kind of where I'm getting at is you,
for all these, for all these arguments, you
kind of already see that, that there was
something there that pre-existed the AI that
was already a problem and the AI is
just accelerating it.

(01:21:43):
It's, it's an accelerant.
Yeah.
It's a, it's a, it's a garbage accelerant.
It doesn't, you know, and, and I, you,
you sort of, I'm just sort of groping
in the dark for a, for a sort
of like, understand, like a, a, a theory

(01:22:06):
of the way this, what this stuff means.
And I think, I think, I mean, I'm
not alone.
We all are in a way we're all
trying to, trying to understand this.
And like you said, you've, I mean, you've,
you've built stuff.
That's amazing.
And, and I don't, yeah, it's just, it's
like, it's a curious thing.

(01:22:27):
I really have been kind of like, I
walk a lot and, you know, during the
day and get up from my desk and
walk around and almost every time that I
walk around for a while, I think about
this topic for a while.
I think, I'm not a huge, what's that
guy's name?

(01:22:48):
The guy with the big, the big fat
rim glasses.
Wow.
Okay.
Prognosticator or whatever.
You know, you know, the insidification guy.
Oh, oh.
I'm totally drawing a blank here.
Yeah.
Cory.

(01:23:09):
Cory Doctorow.
Cory Doctorow.
Yes.
Cory Doctorow.
Yes.
Yeah.
I mean, I'm not a huge Cory Doctorow
fan.
I like some of his stuff.
I think he's a little bit nuts on
some other stuff, but you know, but he
does have, but he's, you know, we can
learn something from everybody.
And I think he does have good ideas
in some, in, in some really important areas.

(01:23:30):
And this is, this is one of them.
And he wrote a good article today, actually.
It's, well, it's actually not an article.
It's a transcript of a speech he gave
yesterday at the University of Washington, where he,
he calls AI the, let's see, what's it

(01:23:54):
called?
What's the, what's the thing that's like a
horse with a human torso?
A unicorn?
No.
Man, bear, pig.
Yeah, yeah, no.
I know exactly what you're talking about.
I know exactly what you're talking about.

(01:24:15):
Yeah, I can't remember the name of it.
Plural, here it is, pluralistic.
I'll try, let me link real quick and
I'll drop it in the boardroom.
The reverse is centaur.
Centaur.
Centaur.
He called AI the reverse centaur.
Yes.
And what he's saying is a centaur is
something that is like a, a centaur in

(01:24:36):
technology terms would be like a human that's
driving some sort of machine.
The human is in control, so it'd be
a little sort of like a machine with
a human head.
Yeah.
And AI is like a human with a
machine head.
Yeah.
Which I thought was a very good analogy.
So it's, it's more like you have flipped

(01:24:56):
the script.
It's like the machine is, is, is in
charge and you're just sort of like walking
around with this thing on your head.
And that's what it, that, I think that
kind of goes back to my discomfort with
this code that I'm generating that I don't
understand is, is I feel like I've had,

(01:25:18):
I felt like I have like a machine
on my head.
I feel you brother.
I feel you.
And I think, so there is a, this
is a long-winded way.
And if, if I'm nothing else, I'm long
-winded.
There is, this is a long-winded way
of saying, I think there is a, there

(01:25:39):
is a legitimate core here that is causing
a discomfort.
And it's something that we should not dismiss
and we should take seriously.
I just think that we don't exactly fully
know how to explain it yet.
And it will become evident over time.

(01:26:01):
You know, as we go forward into whatever
this thing is, I think that it's just
not clear yet.
And this is, this is the way of
the world with technology.
Technology does this to us all the time.
It makes us uncomfortable.

(01:26:21):
And sometimes it takes years to understand it,
to understand why and to see what happened.
And, and we now like, you know, we
were, we now know that, that this, you
know, square, this rectangle in our pocket is
a horrible device.
It's caused so much, you know, so much

(01:26:48):
emotional harm to people.
And it's, it's, it's a, it's putting your,
putting the internet in your pocket was a
terrible idea, you know, in so many ways.
Worst idea ever.
Dumpster fire.
Yeah.
But then, you know, but, but it takes
a long, it took us 10 years to

(01:27:08):
understand that.
And this is, it'll be the same way
with this.
It's going to take, it's going to take
a while, but I did, but there is,
just because you can't explain why right this
second doesn't mean that it's not there.
Yeah.
Let's thank some people, Dave, so you can
get back to your busy job.

(01:27:30):
All right.
Didn't mean to bring you down.
I got to find my paper.
Yeah.
You find your paper.
Not a lot of a boost coming in
during, uh, during this board meeting, but that's
okay.
I think everyone was enthralled with, uh, the
content of it.
Salty crayon, triple five.
That's a row of swans.
He says, welcome back, Dave.

(01:27:51):
Howdy devs of fountain.
Since we were never asked, since we never
asked for video, can you please flip fix
the flipping chapters that continue to show the
same artwork across the entire music episodes?
Sincerely with a cherry on top, v4vmusic.com
for a better, better experience.
Sounds like a bug.

(01:28:13):
And, uh, my experience is, don't start off
your bug fix request with since we never
asked for this, that's just me.
That's just me.
Helpful hint.
That's a helpful hint.
There's Martin Linda's code with a freedom boost,
1776.
He says, Adam, uh, you are the tour
guide of the big apple.
Dave, glad to hear you had a good

(01:28:34):
time.
How did the climbing competition go?
Oh, it's good.
She's, she's currently qualified for regionals.
Very nice.
I joined the live now at 7 45
PM central European time.
By the way, I agree.
It's a good thing with the new podcast
hosting companies entering the market.
I am using alley to captivate and pod

(01:28:54):
home at the moment.
I'll move tea party, tea party, media podcast
to true fans hosting.
And that baller was a Dreb Scott, 10
,000 sats.
Thank you very much, Dreb.
Yo Dreb.
And, uh, with that, I hit the delimiter
already.
So up to you, my friend, I have
not heard, but did you receive your Christmas
present?

(01:29:16):
Um, I have not yet.
Have you received your Christmas present?
I have not.
Okay.
Then you should be getting, but you should
be getting an email.
You sent me an email said be on
the lookout for this.
Like, I don't want to see what it
is.
Don't send me though.
I like packages.
I like surprises.
Yeah, but I can't control it.

(01:29:37):
It's a, it's a, they're going to, they're
going to send it an email.
And I'm, I was afraid the email might
get to you before it's a combo.
It's, it's like an email and a physical
product.
And I, I was like, I was like,
I don't know when they're going to send
the physical.
Oh, finally an AI, a real doll.
Thank you, Dave.
So excited.

(01:29:57):
I went, I hope I went with blonde.
I hope that's fine.
Thank you.
Can I, can I, can I cut her
hair?
However I want to?
Yes.
And it grows back.
I did not.
I got you one physical product, but there
is a software component.
Oh, yeah.

(01:30:18):
They, they, this is a, I wanted the
physical thing, but they didn't sell the physical
without the digital.
And I was like, okay, all right.
Okay.
Okay.
Had to got it.
Well, but if you haven't gotten the email
yet, it got rejected by your spam and
it's quite, it's quite possibly if it has
the word unsubscribe in it, then I know

(01:30:39):
where it is, but I'd have to go
looking for it.
Okay.
Okay.
Well, that's all right.
Ignore it.
And if it gets close to Christmas, I'm
going to check back in with you and
see what the issue is.
All right.
Well, let me know when my gift arrives
for you.
We'll do.
Okay.
Thank you.
Sure.
Okay.
Uh, Oscar, Mary 200 smackers.

(01:30:59):
Nice.
Thank you, Oscar.
And, and friends, Oscar and friends.
Thank you very much.
Diagnosing currently a problem that Oscar is having
with the API.
He is getting rate limited and we're trying
to figure that out.
Uh, and so I think it's, it's periodic.

(01:31:21):
So we'll get it.
We will get it, uh, fixed.
Okay.
Um, Oh, look, we got Barry from pod
home.
Look at Barry Barry, a hundred dollars.
Barry has figured out the art of, uh,
getting your products known on podcasts.

(01:31:42):
He also donated to the no agenda show
for black Friday.
Yeah.
Yeah.
Pod home.com.
I think he has a special going on.
Did he have a note?
He does.
He says, Hey guys, Hey, here, here's an
early Christmas present for the index from pod
home, uh, pod home.fm. I really appreciate
what you do.
And I hope you realize that you are

(01:32:03):
really making a big impact in the world.
Oh, thank you, Barry.
I think Barry has a special prices for
the season.
So if you're looking for, for modern up
to date, uh, all tags included podcasting, pod
home.fm. So my cats love, I got,
we had two cats and they love it.

(01:32:24):
When I wad up a paper wad and
throw it in the house, they'll run and
chase it and then bring it back to
you.
They bring it back to you.
Yeah.
Oh yeah.
Oh yeah.
Can they love it?
Can they fetch your slippers?
Oh, if only, but they are now the
door to this room.
I podcast in his glass.
And every time I wad up these notes
and throw them, they're both sitting there at

(01:32:46):
the window staring.
They're dying for me to throw this, throw
some paper, daddy.
Let's see.
Uh, $50 from, uh, Sergio management group incorporated.
No note.
Oh, and this is a subscription $50 a
month.
Wow.
I'm going to big ball or that that's

(01:33:08):
a 20 inch blaze on the Impala.
Let me see if I have an email
that might go along with this.
Um, just want to make sure I didn't
miss a note here.
Let's see.
Uh, Oh wait, I do have a note.

(01:33:29):
Uh, just set up a $50 monthly donation
to pockets index.
That's my, that's appreciated.
Perhaps you could mention my name, Ralph S
step jr.
Uh, and I am the content creators accountant.
I have four active shows and my network
network is okay.
Yes.
This is a note.
Okay.

(01:33:50):
Hey guys, I just set up a $50
monthly donation to your podcast index.
Perhaps you could mention my name, Ralph S
step jr.
And I'm the content creators accountant.
I have four active shows and my network
site is, uh, ask ralph.com.
I just launched a new show.
The content creators accountant, where I'm working to
help podcasters and other content creators with the

(01:34:11):
financial side of the business.
Best wishes Ralph.
Oh, I see him right here.
Welcome to ask Ralph media.
Oh, interesting.
Ralph looks like Ralph, but Ralph looks like
a fun fellow.
Does he look like a fun fellow?
Yeah, he does.
He's smiling.
Yeah.
Very nice.
Thanks Ralph.
Yeah.

(01:34:31):
Appreciate that.
Uh, that is, let's see, that is it
for those.
We do have some boost grams get in
here and find the boosts.
Okay.
I'll sort that.
Here we go.
So we got, uh, Seth that's says with

(01:34:52):
a three instead of an E uh, 1183,
an authoritative way to point to the actual
authoritative and main feed is so needed.
Yes.
And so this is good.
This goes back to the move, your podcast
.com.
Yeah.
Yeah.
Which we'll need, which we will need to
build.
Yes.
Yeah.
In addition to everything else we're doing.

(01:35:13):
Yeah.
We'll build building every billet.
Let's build it all.
Uh, Oh, 33, three 33 from a Cameron.
Wow.
This is a through fountain and the message
is just IPFS Cameron.
Ah, yes.
IPFS podcasting.net.
Yep.
Thank you, Cameron.
Appreciate you brother.
Uh, Neil Vaglio 2338 through, through, uh, through

(01:35:37):
true true fans.
Okay.
He says great episode explaining the importance of
the index, which we all know, but the
reminder is always good.
Yes.
Thank you for supporting it.
Mm hmm.
Uh, Bruce, the ugly quacking duck 2222 through
podcast guru.
He says, I was trying to add a
card on strike at the same time.
Everything was going down.
Funny how I had to create a support

(01:35:59):
ticket and they got back and said, all
was fixed.
I felt loud.
Thank you for the episode.
73s 73s.
Oops.
See, uh, let's see.
You got, I already did that one.
Did that one.
Is this the delimiter?
Uh, yeah, we get two delimiters.

(01:36:20):
Yes.
We get, because we didn't have a show
last week.
Oh, that's right.
Yes.
Oh, that's a good point.
Thank you.
Come as your blogger for it.
We get double delimiters, two commas in a
row.
Uh, comma strip bloggers, what he should be
called.
Uh, 17 Oh one five sass through fountain.
He says, how do you Dave and Adam,
my fellow Pomeranian fan, a lady Vox from

(01:36:42):
Alaska wolf and her cohost Dame Loka from
Texas.
I've just dropped yesterday new episode of their
podcast.
Find it at a website, www.grumpyolddames.com
or search for grumpy old dames in your
podcast.
Appy app.
It's about quote.
We also grew up about how our oversensitive

(01:37:06):
society has turned everything, including Thanksgiving into a
culture war end quote.
Yo, CSB AI arch wizard.
Yeah.
I'm expecting the AI arch wizard to contribute
to our party time project.
That's why he hasn't yet because we weren't
doing AI.
So now he, he has no excuse, no
excuse whatsoever.

(01:37:28):
Oh, excellent.
Um, well we got a double though.
We got a double.
There's another one.
Comma strip blogger.
17 Oh one five through fountain.
He says, how do you Dave and Adam?
And it's the exact same exact same message.
Okay.
Yeah.
Thank you.
See, it's me.
We got some monthly.
Yeah.

(01:37:48):
Way to go.
Uh, Joseph Morocco, $5.
Kevin Bay, $5.
Cameron Rose, $25.
Dreb Scott, $15.
Chris Bernardik, $5.
Michael Kimmerer, $5 and 33 cents.
Randall Black, $5.
Christopher Reamer, $10.
Cone Glossbox, $5.
James Sullivan, $10.

(01:38:10):
Podpage and Brendan over there.
$25.
Mark Graham, $1 and new media.
That's Martin $1.
Wonderful.
Throw that at your cats.
Thank you all so much.
Uh, board board meeting adjourned.
Uh, are you working this weekend?
Do you get to hang out?
What are you doing?
Uh, we are going to the firm Christmas
party tonight.

(01:38:32):
Ah, nice.
This is where you get your Christmas package
from the firm.
You get your little Christmas package.
That's a, that's a big thing in Holland.
It's called the cash pocket.
Uh, so Christmas package and people will be
talking to us.
Hey, uh, what'd your company give you in
your cash pocket?
And that's like a, like a bonus or
something.

(01:38:52):
No, it's literally like a package and it'll
have maybe a bottle of wine and some
cheese and you know, and sometimes, you know,
like a cool gadget.
Um, so it's the cash pocket.
Everyone's talking about it.
And right now, what is it?
Yes.
What is the Dutch cheese?
Like what did the Dutch, are the Dutch
famous for a particular type of cheese?
Why I'm glad you asked.

(01:39:12):
Yes, they are.
Gouda or Gouda.
Oh yeah, yeah, yeah.
The Americanized Gouda.
Yes.
And uh, and you know what they also
make in Gouda is candles.
Holland is one of the, uh, one of
the largest candle producing nations, I believe.
And they make them out of Gouda.

(01:39:32):
Well, no, if you get your Gouda, the
wax on the outside is of course from
the, from the wax people.
Then they pivoted that into a candle making.
I'm not sure about that.
I'm not sure.
I'm sure they pivoted into some kind of
food that we're eating that we don't know
about.
We're always eating candle wax.
So isn't that what Crisco was?

(01:39:53):
I think Crisco was, uh, uh, was candle
wax and they turned out, oh, well we
can cook with this.
You should never cook with Crisco.
No.
Because it is candle wax.
Crisco is good for one thing and one
thing only, and it shall not be mentioned
on this board.
All right, everybody have a great weekend.
We'll see you next Friday, right here for
more of Podcasting 2.0. Podcasting

(01:40:34):
2.0. Visit podcastindex.org for more information.
Go podcasting!
I'm a huge fan of nuts.

All Episodes

Episode Transcript

Popular Podcasts

Stuff You Should Know

Dateline NBC

The Burden

.css-15opob5{left:0;position:absolute;top:0.8rem;} All Episodes

.css-14f5ked{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:2;overflow:hidden;}Episode 243: Nuts & Logs

Episode Transcript

Popular Podcasts

.css-r6mb8g{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:1;overflow:hidden;}Stuff You Should Know

Dateline NBC

The Burden

All Episodes

Episode 243: Nuts & Logs

Stuff You Should Know