Vercel is the platform for front-end developers, providing the speed and reliability innovators
need to create at the moment of inspiration.
Founded by the creators of Next.js, Vercel has zero configuration support for 35+ front-end
frameworks, including SvelteKit.
We enable the world's largest brands like Under Armour, eBay, and Nintendo to iterate
faster and create quality software.
Try out Vercel today to experience the easiest way to use Svelte.
Welcome to Svelte Radio.
Hey, it's another episode of Svelte Radio.
Today, we're two people here.
Well, actually, two in addition to me.
I mean, sorry.
So I'm joined by Brittany and Sean.
Anthony is missing today.
How's it going?
Good to be here.
Good to catch up as well.
What's new in the world of Svelte?
Yeah, what is new in the world of Svelte?
But maybe first, what's new in the world of Swix and Brittany?
What have you guys been up to?
Brittany, you go first.
I've been crazy busy.
I did JAMstack conf last week, and then this week, I'm recovering from a cold, and I have
to MC Remix conf EU tomorrow.
So a little bit of a traitor, I know.
Remix is special.
Remix is cool.
It's one of the good ones.
What have you learned from MCing so many different things?
I think I'm just getting more comfortable with it and rolling with the punches, but
also that being more prepared is better.
They have slides with actual notes for everybody, and I love that.
So I think I'm going to use that in the future, how I have little notes about every talk and
things going forward.
So no improv anymore?
That's what you're saying?
I mean, there has to be improv.
That has to exist to make it fun.
But it was funny.
I was adding in some of my own notes and personal experiences to these, and the person came
back to me and she's like, "Yeah, we're not really supposed to talk more about the sponsors.
Just put them in here and leave this like that."
I'm like, "Oh, okay."
So maybe I can't improv this one.
I mean, so how many conferences have you MCed now?
I did Svelte Summit, and then I did VEET conf, and then now Remix conf.
I was going to do Nuxt Nation yesterday, but they had a weird situation where they had
me and Jinjoo Na doing a different day each, and then they wanted somebody to do both days,
and I was not available today.
So they just went with her, which is totally fine.
I don't want to do two in one week anyways.
That seems reasonable.
They just released version three, I think, as well.
I have no idea what's in it, but-
I really am not familiar with that ecosystem either.
So it was probably best that I did not do that.
How about you, Sean?
I just did a writing retreat, which I know, Brittany, you wanted to come for.
But that was last week, and basically I led a workshop of 25 folks for five days straight,
and that was very tiring.
Honestly, one of the most fun and challenging things I've done this year, and I don't know
what to conclude from that.
Obviously, most people cannot do writing retreats just because you have to travel, it's one
whole week, it's a long slog, but I think people really appreciated essentially air
cover for a holiday that has some professional development, but then also some local community.
I think writing is a very lonely endeavor.
And if you have a brief period where you're talking to other people who are also interested
in the same things, that's the benefit of having a small little retreat format like
It looked really nice.
I saw some of the pictures.
It looked like you had a good turnout.
Was it good?
Was there a storm as well?
I saw some tweets.
Yeah, there was a hurricane.
Well, there's a tropical storm on the Wednesday, and then the forecast was like, "Okay, Wednesday,
Thursday, Friday is going to be like this."
And so one of our attendees was actually freaked out and moved her flight so that she could
And then right after she left, the storm cleared up.
And it was beautiful.
Oh, that sucks.
That does suck.
So when it comes to a writing retreat, that must be kind of different to what you're used
to doing in general.
Yeah, I've never been to a writing retreat, and here I am organizing one.
So partially, I have modeled it essentially on what I know normal writers do.
I think a lot of developer events you try to invent from first principles, but actually
you can just steal from things that exist in the regular human world and bring it over.
There's another world?
Professional writers do these sorts of things.
They get together and then they share their goals.
They work on their own projects, but then maybe sometimes they'll do some activities
together that stretches them a little bit.
For me, I also served as a mentor for people who need that kind of mentorship, but there
are some people there who definitely don't need anything from me.
They are successful authors of their own.
Philip Kiley wrote Writing for Software Developers, and Monica Lentz also came who runs blogging
So I think the appeal of like, "Come join us and in the morning we'll work on some skills,
and in the afternoon you can work on your own projects and then we'll get together every
I think that's a really good format for most people who are trying to improve on any skill.
It could be writing, it could be dev rel, it could be programming.
Kind of doesn't matter.
Oh, I think a dev rel retreat would be really awesome.
Yeah, we actually talked.
We had two YouTubers there, three actually, and we were talking about what would it take
for a dev YouTube retreat?
Turns out there is a small little...
I mean, YouTube is actually probably more developed than writing, and they've already
done a couple of retreats already just among YouTube creators.
I think sometimes that can be a bit intimidating because a lot of these are very well established
channels with hundreds of thousands of subscribers.
It might work for people who are smaller and just getting started.
Then you might want to do a separate thing.
I don't know.
It is definitely true, the logistics of organizing a retreat, you're no longer just creating
content, you're also telling people where to go and what time to eat.
If people have any other issues, you need to take care of that.
So, yeah, it definitely needs a lot of logistical overhead, and I couldn't have done it without
Michelle who is my operational partner.
Kevin probably knows a little bit about that, right?
Trying to organize all that.
It wasn't a retreat, it was just a conference.
Yeah, but still, organizing and getting all the schedule and the food and all that, you
did a great job.
Do you have anything new, updates?
Yes, yes, but before that, I wanted to ask you one more question, Sean.
You mentioned it's a pretty lonely thing, writing in general, right?
And it is, but have you experimented with writing together with someone?
Because I know you write a lot of blog posts and stuff, right?
No I have not because I don't think that adds anything to me.
Also my writing schedule is very random.
Sometimes it would just be late at night, I just feel like writing and do it.
Sometimes it'll be Saturday in the afternoon after lunch or something.
So I don't really have the time to schedule with someone.
And if we're going to sit on a Zoom call and type in silence, I don't think it's a very
Like pair writing instead of pair coding.
I think occasional getting together is useful and then for most of it, you don't have to
have any accompaniment.
I think it's just nice to know that you're not alone in facing some of the problems that
you might face.
I guess it could work more as like at the concept stage of a post or whatever, right?
Like discussing the merits of something.
One thing we do now at Netlify is have pillars of the content.
And so we fill out the pillars and then we share that for feedback.
And so we'll get feedback on our, it's like your persona and who you're writing for and
what you're writing about and why you're writing it.
And then you'll do your outline and get feedback on that and then you do your post and get
So it's kind of like that, but you're writing all of it.
You just get feedback and comments along the way.
So what have I been doing?
I mentioned last week on last week's episode that I was doing subtitles for all of the
videos and the streams.
That's now done.
So we have OpenAI Whisper subtitles for all of the videos on the channel and even some
translated ones as well so you can go back and watch the Svelte Society Day France 2020
if you want to.
Thank you so much for doing that.
It's actually, I'm very tempted to build like a service that does this.
Yeah I think that there's some automation work that we could do there to make it easier
I mean I already have like an outline of what I want to build around this so it's going
to be a lot of fun.
Oh that sounds fun.
It's going to be a lot of fun.
Unless I want to turn it into like some SaaS or something.
And then keep it secret.
It's not very hard thing to build though I suppose.
That's what I've been doing.
I've been working a bit on the kit base template thing.
Upgrading the pocket base version to 0.8 which is a pretty large update so there's a lot
of things that are broken that need fixing.
So I'm currently working on that and then yeah.
And then I'm going to start.
Is that the pocket base API that upgraded and broke something?
They changed the API a bit and so I need to.
That's always fun.
It's the whole kit.
So let's get into the topics that we have lined up today.
First off we have I don't know how to pronounce it.
I think that's right.
I think it's Threlt.
It's probably like the easiest way to describe it is it's a more mature version of Svelte
Cubed kind of.
That was my impression as well.
So it's built on 3JS.
I think so.
It's a library that's under the hood.
Under the hood.
And then I haven't been able to look at it that closely yet but I'm going to take a look
at the stream that Willow did I think a couple of days ago.
I watched a lot of the video but I had to leave towards the end so I missed some of
it but yeah it looks amazing.
Like I just love all the 3D work that you can do with these libraries.
It's kind of crazy and it's very declarative.
Like Willow was mentioning a lot of the code is like very easy to write and rather than
3JS you have to write more low level stuff.
This is more declarative so you can tell a little bit easier what's going on.
It looks like so they posted in the subreddit as well and I think they only had like a like
a video or it might even have been a game or something like a pong game.
Looked very cool at least.
So I'm going to post that as well.
That would be really cool if we start getting games built in Svelte.
So this is this is actually extra interesting to me for a number of reasons.
So I've been playing around with something called a Remotion lately which is a React
framework for writing I guess videos which is very cool so you can programmatically write
All the transitions and all of that stuff.
I'm wondering if maybe you could combine that with a D3 library like Threalt.
I don't know.
So to do like 3D transitions in video?
Just 3D stuff in general in videos.
That would be cool.
Doesn't Remotion require React?
I'm just like thinking out loud if there's a tool like Threalt but for React I'm sure
Can we make Remotion in Svelte?
I don't know.
Actually I asked this specific question on Twitter and I got a reply.
So there's actually someone who did that but it's very much like a beta thing.
It's called Svelte in Motion and it's basically the same thing.
I'm gonna link that here as well.
I tried out Remotion.
I felt like it was just so hard because you have to can code every single CSS variable
and transition and all the tweening or whatever.
I don't think it's that productive unless you want like programmatic control like you
It's not super productive compared to a dedicated video editing software but I mean for sure
it's good for some use cases but I found it less productive compared to just producing
a regular video.
Even Keynote right?
If you want to do some simple animations in Keynote that would probably be much much faster
than doing it in Remotion.
Yeah or even just using Figma or something.
You could do transitions there as well.
So I think for me the use case would be something like Audiograms.
Like you cut a bit of audio from the podcast and then maybe you put a video in the background
and then you have some text and some like audio waves that you can post on like YouTube
short or Instagram or something or Twitter even.
So that would be what I would use it for.
Like I wouldn't edit like a full on like one hour video in it I think.
But maybe I don't know.
Did you see that Remotion got funding?
Yes, yes I was super happy with that.
And it's Remotion in general just it's such a cool thing.
Even if there aren't that many use cases for like longer form videos it's just like a really
cool technical thing.
All right so maybe at some point we can try out Threalt and give it a review.
Yeah I know we went like from Threalt to Motion to we went through the spectrum there but
that's what we started talking about.
Yeah, throughout the XYZ.
I would say the Threalt team have been marketing every single day.
They put it definitely been putting a lot of work into the docs.
So this is one of the most, I don't know, just serious efforts to come out this whole ecosystem
And yeah it's pretty cool.
Someone's really attacking this.
I like I want to put this I want to put something like this on my own landing page or homepage.
So I just need to figure out what I'm going to show.
Yeah it's like it's awesome when you get to like a website and it has this cool like 3D
Yeah it's just a small thing it's very very enticing and very what's the word.
And that's the thing with 3D right like there's just so much to go over that having more and
more examples is super useful.
So hopefully I can find something to copy and paste.
The life of a programmer.
All right so that's Threalt and next up we have something called Svelte CMS which is
very much maybe not a thing yet.
We only have a tweet and the video and a video demo.
I was going to say I went looking for it and could not find a website or anything.
Yeah I don't think it's launched yet.
So we'll see we'll see what happens but you know they tagged Svelte Society on it and
I looked at it and I was like okay like you know it's a pretty serious CMS project and
if you want to you know if you want to Svelte build CMS like this one this might be it.
But I don't know like I think it was an interesting new project that came on the scene this week
so I figured I'll throw it on the board.
It is it looks really nice too it's a nice UI.
Not much else we can really say about it.
Yeah I wish there was like a GitHub repo or something that we could at least like look
at what's going on but.
Well 17 minutes ago they said it will they will make the main repo public soon.
Oh look at that like right on time.
Maybe we'll put it in the show notes if we get it.
All right the next thing up is Svelte-osis.
If you haven't heard about Mitosis which is what the original one was builder.io released
this product where you could write JSX like code and then transfer that into other languages
But now Svelte-osis which was a side project has been integrated into it so now you can
write Svelte code and also transfer that into other languages.
And it is released and active.
Kevin's got some bugs with it.
Yeah it looks cool like like just if you go on the Mitosis website like the mitosis.builder.io
you click the like you'll first have the mitosis JSX thing selected right then you just switch
to this Svelte-osis one and just half the code disappears which is always nice.
Yeah I know I love seeing like the lines disappear.
Yeah yeah it's it's it's pretty cool.
So now you can write Svelte and export it as react so you can write Svelte at your react
Exactly so now if they won't just switch to Svelte then you can just write it anyways.
But there are there are so I discovered a small bug with reactive statements so if you
have a function in the reactive statements it just doesn't work.
And I think the the developer or someone close to the developer replied and said that there
were a couple of other bugs as well.
They were they were working on them.
They said it was working when it was outside of mitosis and something possibly with the
migration had caused some bugs so hopefully they will get those sorted out soon.
Yep all right.
Vercel is the platform for front-end developers providing the speed and reliability innovators
need to create at the moment of inspiration.
Developed by the creators of Next.js Vercel has zero configuration support for 35 plus
front-end frameworks including SvelteKit.
We enable the world's largest brands like Under Armour, eBay, and Nintendo to iterate
faster and create quality software.
Try out Vercel today to experience the easiest way to use Svelte.
All right next up what's new in Svelte?
It's the so we're going to talk a bit about the newsletter that came out at the beginning
of the month and I think we have one one important thing in here.
Well I'm sure there are like a hundred important things in this newsletter but like one thing
that stands out to me is the optional parameters thing that you can now do.
So previously you weren't really able to do language specific sites in SvelteKit without
like a lot of extra work but now there's an optional parameter which means that you just
add like double brackets and then a word inside so you would say a bracket bracket lang and
then bracket bracket and then you could have like a if you had like an English and then
a Spanish site you could have like the default would be the English and then if you have
a domain slash ES slash whatever then that would be your Spanish site.
Is that in the link in the what's new in Svelte?
Yep it says new SvelteKit feature.
Yeah it's the second bullet point under what's new in SvelteKit.
Oh okay I was searching for lang when you were talking about how to use it I was hoping
that that would be in there but that might be in the doc.
It was just my my example.
I think this is an interesting is a question of like yes you could do it but should you
I'm not really sure.
So I don't know what you mean multiple websites or multiple languages?
Multiple languages with this kind of you of name schema right like so you could do it
by subdomain so like let's say you know Spanish would be es.domain.com right or you could
say and what this is letting you do is you know www.domain.com/es/home and I'm not sure
Yeah so I think that I assume the issue before was that if you had it on a different subdomain
you would have like a would that be a separate site?
I guess you could you could read it.
Yeah it would essentially be a separate site.
Yeah so this way you could you could do it on the same site I guess.
Right and then should you take a cookie and then edge render based on what language that
they have so you don't like basically like should you leak language implementation details
to the URL right like I think some sites do that some sites don't and I don't know what
the trade-offs are.
I think for apps you do not and then for blogs you do but I'm not sure that actually that's
better compared to just having everyone linked to one canonical URL regardless of language.
Yeah it kind of ruins SEO right?
I don't know.
Maybe not right yeah because maybe people specifically want nothing in some language
and it is very confusing to have one page represent multiple languages which you could
make it do.
So I just don't know what the best practice here and yeah it's an interesting debate.
Yeah so I remember I was thinking about this a while back and like there's some Swedish
site that will like force you to use the Swedish site even if I don't want to use the Swedish
site so there's no like language picker or anything it's just like I guess they're looking
at where I'm.
Your headers your browser your browser is with the language.
Yeah exactly which is that is super infuriating if you want to like get to the English website
or whatever or the other way around.
Yeah I think basic principle is you should let the user choose and yeah I'm not really
sure how to deal with this.
But let the user pick yeah.
Okay but you know and then another interesting angle so this one essentially opens up one
vector of clashing so if you have like optional paren slash dashboard and then you have slash
dashboard with no optional paren you will clash and I don't know how it should handle
There's some logic in there about how it matches the route.
I don't remember exactly what it says.
Yeah so you can't have an optional route after a rest parameter.
Okay right that makes sense right rest is greedy so dot dot dot rest is greedy that
So yeah I don't know I guess you could have you could always composed dot dot dot res
into this feature but it's nice to have it out of the box.
Yeah I mean you'd have to do a lot of work if you did it yourself.
No it's probably like 10 lines of work I don't know maybe I'm over overestimating it.
But okay so you know if you had dot dot dot res you would have like a layout page that
would then branch based on whatever was in the res param right.
So that's all with it.
So you'd have to rebuild the entire route system as Felkit?
If you kind of.
You're just taking a chunk of the route system and then do you have to rebuild it?
Oh yeah you may have to rebuild it.
Okay I take it back.
Yeah well so now you don't have to rebuild it because you have it built in the optional
param go check it out if you build the international websites I'm sure this is more for you than
for for those of us just building in a single language.
Has anyone tried the use enhance action the the new form action?
Yeah so I'm not sure if I've tried the latest version of it so there's a new update method
that lets you easily get back to the default form behavior while augmenting your right.
Oh I thought you could already do that so I've not tried this particular feature out
the update one.
Yeah it says in the docs update is a function which triggers the logic that would be triggered
if the callback wasn't set for the use enhance method so okay interesting I don't know when
I will use it.
I'm gonna have to try this out before I comment on it.
I've used the like the apply action one and call for tutorials if anyone needs.
See if I screaming at this podcast saying like of course it's so obvious please write
a tutorial so we can see.
We need tutorials.
Is use enhance not the form thing that it is the form thing but there's a new method
on the form thing.
I was gonna say one of my picks.
We're trying to figure out what update does.
Function which triggers the logic that would be triggered if this callback.
Yeah that's a confusing sentence.
Yeah all right so well Brittany we're doing a stream on forums right?
Yeah that's what I was gonna say one of my picks is gonna be our stream November 30th
I think we have it coming up.
We're gonna build out right now we're using sessionize for speaker submissions on Svelte
Sirens and we're just gonna build it in Svelte Kit because why not.
Yeah sounds good.
What are we gonna use as a back end?
I didn't even think about that.
You have to have somewhere to like get your form replies.
Yeah could use BuddyBase.
They recently released 2.0.
I love Netlify forms so like I could just push them all to Netlify.
So yeah I mean that works as well.
Probably don't need anything that's too complicated right?
Yeah that's that would be.
Actually we wouldn't even have to use Svelte Kit.
I have a like a vague memory of someone releasing like a Sass service that literally just receives
Which could be interesting.
I might need to do some research before we do that stream because I don't know what we're
Yeah all right and so that's the the new update method that we have no idea what what it does.
Okay no I was thinking a little bit about it.
So essentially like use enhance is like a prevent default.
It cancels any normal form submission callback that you might use and then and then you sort
of replace it with the function that you pass into use enhance.
So the update is kind of like a callback to to to go back in once it's enhanced to go
back in to the regular unenhanced.
That's at least how I explain it.
So maybe not super clear but like I think that the analogy of prevent default and then
maybe not preventing default is the right one.
So I still I'm still gonna need to try.
I'm just too confused.
All right and then there's something else here.
Yeah I dropped in a last-minute one that I realized yesterday.
AppRite which is a database company went on Product Hunt and announced that they have
their console 2.0 out and it is built in spell kit and I saw it in action yesterday and it
is really sharp looking.
I like it a lot and so I've put a link to the Twitter post about the console and the
Product Hunt launch and then there's the package.json link that shows that spell kit is being used.
I can't actually get into the console right now so I've been a little confused for the
last few minutes looking for that.
So I thought right yeah so AppRite is like a like a super base Firebase kind of alternative
yeah or pocket base as well I guess.
Well that's cool that they're using Svelte.
They sponsored Svelte Summit as well.
They did yeah our lanyards have AppRite on them.
Yeah yeah it's a great company that's where my coding cat partner Alex works and one of
my picks too is also going to be about something that we're working on that's built on AppRite.
So cool exciting.
All right I think I think that's it for for topics right unless you guys have anything.
Yep that's it.
Yeah all right um unpopular opinions last week I my pick was Twitter is not going to
Yeah because I said Twitter was going to fail and then he had to counter me and then I said
I actually wanted to ask Sean because you have more of that like marketing like money
brain from your financial background.
So I'm wondering what your perspective is on Twitter failing and just going bankrupt
in the next year.
Oh I don't know everything everything keeps changing.
What is happening with the world?
So up till yesterday I would have agreed with Kev.
Yesterday really changed a lot.
So what happened here today was Elon sent an email to the entire company talking about
Twitter 2.0 and saying essentially putting a link at the bottom saying where everyone's
going to work extremely hard to make me good on my 44 billion dollar investment sucks to
be you but if you're signing up to do this incredibly hard journey of making a social
media app click here and so you have to opt in to keep your job in 24 hours or you get
a three month severance and you know be on your way in the tech economy.
In an environment where there's already been a hundred thousand people laid off something
So like this is not something to take trivially like anywhere between like this is opt in
to keep your job.
This is not opt out.
I'm sure a lot of people are looking at the button today and going man I mean three months
severance I can get a new job in three months like no big deal.
But if you're in a visa situation or any other number of like situations where you can't.
Well I mean so you know as a visa holder myself you have some leeway to get a new job.
Some people yes they get shipped out immediately but a lot of us do have some grace particularly
if it's a layoff.
So I think that's okay.
I mean not okay but like whatever it's that's the way things are.
But like seriously 90 percent of the existing employee base could disappear today this week.
And I don't think Elon cares very much.
And so yeah I think there's a good chance.
I've already backed up my data on Twitter like I'm prepared for shit to go down.
I requested mine and I don't think I've gotten it yet.
Their 2FA system was down.
So when I requested my data it said we're sending you a text message put in the code
and that was one of the microservices they got rid of.
So that did not work.
And luckily the email one still works so I was able to request the data but I don't think
I've received it yet.
I'm sure it's a very popular button.
You know I'm sure everyone clicks on that.
I'm sure they have a lot of data to prep.
I'm sure it's a very slow and unoptimized scripts because most companies do not take
export of data seriously right.
Like it's not something that they're optimizing for.
So yeah I mean it's a very unfortunate situation.
The only thing I'll say is like Elon seems fairly serious about Twitter.
Like this is his job now.
I don't think it's the right thing.
I don't agree with it at all.
But he does take it you know I mean he does want to make a success of it.
I don't think he's trying to run the company to zero as some people are suggesting.
So we'll see what happens with that because I mean like if you do that you don't do the
things that he's been doing.
So it's very hard to reconcile his stated intentions versus his actions.
Yeah I saw something about what's her name something shot well at SpaceX is taking over
like the CEO role there from.
I think she was always the CEO.
Yeah she's been running SpaceX for a while.
It was Tesla.
Tesla's the like why would you drop Tesla to run Twitter.
It just makes no sense.
It's way more important.
Yeah the world doesn't make sense.
Things are such a mess over there and I heard when you do get your data if you had anything
that was in like a private circle or something that you didn't want public that it's just
kind of mixed in with your other data.
It's not actually like secured away like you thought it was.
Of course because no one bothered to think about that.
So unpopular opinions then.
Do you have any?
No you go first.
So okay so all search is AI.
I feel like people don't treat this seriously enough.
What search ultimately is is you typing in what you want and the computer trying its
best to get it back to you.
So a very simple search would be a filter right.
Like take existing list do exact string match.
And then you're like okay but sometimes I do typos.
Sometimes I need stemming.
Stemming is the process of taking out the words.
So what's an example of stemming?
I feel like I need some examples pre-baked.
But essentially let's say you have the word params and param.
If I type in params I want param to also show up.
So that is stemming which is like let's reduce everything to like sort of unit word and then
search based on that word.
So that's a form of AI.
I want a fuzzy search which means I need it to be resistant to typos.
So if I type in svelte or savelte I need both to show up.
And there's some formulas to do with Liebenstein distance and all that.
But even that is not that great because sometimes you actually just want like the rough domain.
You don't know what exact keywords to search for.
So let's say like if I type in frameworks I want svelte and react to show up even though
the word framework is not mentioned in the exact document.
So what that requires is vector search meaning in the mental space of concepts the word framework
and the word svelte, the word react all exist within a nearest neighbors function to each
other and so on and so forth and all the way until you essentially get AI where you interpret
what the user's intent is and try to serve out the best intention for them.
So I think a lot of people like kind of don't get this mental progression of like search
is AI and so I just wanted to offer it.
I don't know how obvious this is to people.
Is this like the first time you've heard about something like this?
I mean I haven't heard it in that way but I think it's obvious in hindsight once you
So it's like hey you know the combo box that Nellify shipped or the combo box sorry not
the combo box the command E like the you know the command key bar that every app has right
linear even Alfred or like the command palette.
Yeah the VS code command palette.
You know that feeling where you type in something that you want and then it doesn't show up
because you typed in like you used the wrong word.
Essentially everyone needs some kind of language model essentially to serve you what you want
and I think that this is not taken seriously enough and or this is the way things are where
everything's going to end up.
We're going to have very small language models embedded in anything with a text box because
taking things literally is not the way that humans really want to interact.
Humans want to interact by saying what they want and then you give it to them right.
That's ultimately what it is.
I guess it's kind of like moving from an imperative way of thinking to like a declarative way.
With some allowance for imprecision but like I think there are many use cases where this
is totally fine.
Like it doesn't have to be precise.
So you know programming is a precise format like Copilot is pretty much the state of things.
In my picks I have a Copilot competitor but yeah I think you know basically anywhere there's
a text box can you imagine a better user experience?
Unless if you and if you sort of offer some AI feature that basically instead of taking the user literally
try to understand what the user wants and give them what they want instead of what they
Very very subtle difference.
So to sum it up your unpopular opinion is that search is AI and not just...
I had a different I typed in a different one at the start as well.
So all if statements eventually become AI as well.
So if X more than Y for example you know that is an arbitrary hard-coded activation function
that you could train with data right.
So like that's it.
Sorry am I losing you?
I don't know if I should.
No no I think I am.
Okay so all right let's say if like percentage of positive responses is more than 0.65 do
So like that just an arbitrary like anytime you've hard-coded any if statement at all
that is hard coding a decision node that someone made some product manager someone somewhere
decided that you know if some expression is equals evaluates to true then do this.
But that decision or that decision node is essentially some baked-in opinion that you
might change based on data based on how people actually interact with your service.
And so instead of doing instead of hard coding in your code why don't why don't you replace
that with some feedback some back propagation mechanism where you observe how users respond
and you know you update your if statement based on that right.
Like that's ultimately what we do with human PMs that you can automate with inhuman PMs.
Poor product managers.
Yes not only that so most of this most if statements are done on an app-wide level but
with AI you can kind of do it on a personal level.
So like an on an individual basis how can you optimize for what people want.
Yeah so so I have a I think a kind of interesting thing to say about this.
So there was a some kind of project a while back about doing analytics and using analytics
to prefetch websites depending on where you go.
So that would be like like a very rudimentary version of this would be like oh you go to
the home page 90 percent of users then click on the product page.
So we prefetch the product page and if you go to the product page we prefetch the first
product or whatever.
You could you could probably train like an AI there because that's kind of what this
is right the like you have if statements basically but then instead of an if statement you just
feed analytics data into it into some kind of AI model and out you get like a weighted
score of some kind that you can use to prefetch and then depending on right.
So Gatsby did a demo of this.
At Chrome Dev Summit I think maybe four years ago using guest JS.
Yeah I think the potential is there obviously the hard part is one getting the data and
then two implementing it in a lightweight fashion right.
So I think there's essentially a so you know it doesn't have to be AI it can just be feature
flags where someone manually turning them on and off without changing the code which
is essentially a CMS right and it could also be yeah just like personal settings flags
like essentially like I think there's a data framework somewhere to be made that offers
this as a service like turns every branching logic in in an app into something configurable
and optimizable whether by human or machine.
Yeah sounds like a billion dollar.
I don't know yeah I've been thinking about this I was like what's the potential of this
I don't think it is I think it's just a very simple small yeah feature flagging type of
thing and it's too small or like it's too easily built by everybody you just have to
make it an open secret that this is the best way to do to make apps.
Right okay yeah I have no unpopular opinion this week.
I only have a opinion about the.
We talked about the Twitter stuff.
Yeah yeah yeah last week we talked about it.
Yeah all right and Brittany you you don't have one either I guess.
No I just wanted Sean's thoughts on the Twitter stuff that was kind of like just bringing
all that back up.
I don't have another one.
All right and now let's let's move on to picks.
Okay I know I have quite a few in here but really it's like one thing the app right thing
I realized why I couldn't get into the console by the way it's because right now you have
to install it locally and then host it somewhere else or self host it but in a hopefully a
month they're going to have a cloud version of it so right now you have to run the console
yourself or host or whatever but that's why I couldn't find it earlier.
My first pick is for DevRel social which is something that Alex was working on in Next.js
and he has a working version on his Twitter where on the banner you can add an emoji and
it will add the emoji to the banner and so we are going to be moving that to SvelteKit
and the first one we worked on yesterday is in there as a link for our stream.
I'm going to take a look at this later.
Yeah and all right the the stream as well on the yeah yeah I have the stream in there
but we did kind of mention it earlier but we're streaming on November 30th for form
actions I just have to figure out where that back end is going to be or if we're going
We can talk about it later.
Yeah sounds good.
At some point right Sean.
Yeah still on AI.
Have you have you fallen down the AI rabbit hole?
A little bit.
We'll see how far that goes.
So last week while you were at Jamstack Conf there was also GitHub universe that was going
on and they demoed Hey GitHub which is a voice assistant for essentially GitHub Copilot
plus OpenAI Whisper.
It's really nice for voice coding because if you've seen any voice coding demo in the
past five years it's all been precise.
It's all been again like enter in this key if I if I say um at all that might be interpreted
as a command but I think what Hey GitHub is is essentially like kind of like an Alexa
type thing where there is some AI layer that tries to understand what you want and gives
you the code that you ask for including navigating around VS code itself.
And so I have RSI repetitive stress injury which means sometimes on my left hand I get
pain when I type which is which sounds like the most first world problem.
It's like oh I type too much my hands are hurting.
Kind of rough when your job is about typing.
So I think I've been very excited about voice coding.
I think you know Scott Hanselman has this very famous blog post where he basically says there's
limited amount of keystrokes that you have in your life.
So you should use them wisely because your body was not evolved for typing on keyboards.
And so I think voice coding could be a lot better with Hey GitHub.
It's on a waitlist right now but essentially if you talk to me I can get you off the waitlist
if you're serious about giving feedback.
And yeah I think I think that's super useful.
On the non-GitHub side because I don't want to be too much of a GitHub show there's Codium
which I saw which I was demoed yesterday.
Codium is a independently created GitHub co-pilot.
If you're not a fan of GitHub or if you want to see what alternatives to co-pilot look
like a small team of like four people coded this in a month and created a complete GitHub
A complete co-pilot competitor.
So I'm pretty inspired by like one the fact that it was a small team that did this.
But two also it seems to be faster than co-pilot which is pretty cool.
I'm going to check it out.
That sounds really interesting.
So my I'll be quick here because I know Brittany has to has to run.
So my my pick is a game called Stacklant.
It's five bucks on Stitch.
What's it called?
It's itch.io not stitch.io.
And it's it's basically just a game of you build a village.
That that was the worst description of the game.
Yeah it is just like cards and you build stuff and you fight monsters and it's pretty cool.
Five bucks like huge value.
And yeah that's it.
All right I think that's that's it for us.
Next week we'll probably be talking to Penguin about Gradio.
So it's going to be interesting.
So yeah until next week.
Check out his talk from Soil Summit if you haven't seen it.
It's one of the most unusual talks I've seen.
It's very good.
Thanks for listening and we'll see you next week.
See you next time.