Jev, Jev, Jev. Welcome to Jev week onhow I AI. We have seen a lot of newmodels be released in the last 5 days.We saw Opus 55. We saw GPT6 Soul, GPT6
Luna. Muse is blowing up the timeline.Everybody still loves their Grock bots.And yet there is one thing that I want
o talk about in AI right now and thatis this fast cheap doesn't speak systemone decision model from Typesafe AI. As
oon as I saw this trending on X, asoon as I saw it launched, I immediatelystarted testing it. Now, what I will say
is more than any other model I'vexperienced lately, Jev has been the onethat has explodeduse cases, personal productivity use
cases, code use cases, product usecases. Today I'm going to give you avery quick whirlwind tour of what Jebis, what it will give you back and what
it won't, and then how I have used itover the last week to do work that Ithink is worth hundreds of thousands if
not millions of dollars. And I haveprobably spent sub $10 on Jev tokens. Alot of it has been subsidized because
Jeb is currently free on the AI gatewayby Versel. But even at list price, it isa very inexpensive model. It is a very
effective model and it is going to be inthe middle of almost everything I buildfrom here on out. [music] So, let's get
o it. This episode is brought to you byOpen Art Arena, the global leaderboardfor creative intelligence. Every week,
new AI models launch, and everyoneclaims to be the best. But best at what?
Open Art Arena is built to answer thequestion that actually matters. Whichmodel is [music] best for your specific
job? Instead of one overall winner, Open
Art Arena ranks models across realcreative tasks from advertising and filmto animation, product, graphic design,
editing, and lip-sync [music]covering both image and video. And theserankings aren't based on hype. They're
judged by professionals, industryleaders, and working creators throughblind evaluations. So judges never know
hich model produced which output. Thatmeans you can see how models actuallyperform when it comes to the creative
work you're doing. So stop guessingwhich model to use. Explore rankingsbased on real creative work and find the
right model for your project and savetime and cost. See the rankings at OpenArt Arena. So if I were to explain Jev
to you, I would go to this table on thetype blog post announcing Jev. And itbasically compares normal LLMs on the
left, Jev style LLMs on the right. Bothtake in unstructured data as inputs.Both take in text as inputs. Jev does
not take in images, but it takes in textand it takes in text descriptions ofimages if you really need to get there.
The outputs though are very different.With the standard LLMs that you're usedto working with, you are getting strings
and generated text out. So you'regetting text in, text out. With Jev,you're getting text in, type safe values
out. And what I mean by type safe valuesis these are values that are predefinedthat then Jev picks from and selects and
returns to you. We will show you whathose values are, but essentiallythey're like it's this or that. It's yes
or no. It's 1 through 10. Like it'spretty simple. Now it sounds simple, butit is incredibly powerful when you put
it against the right problem.The other thing that you will noticeabout Jeff is it is cheap AF and it isfast AF. And so if you look at the
current cost of input and output tokens,it can be pennies to tens or hundreds ofdollars per million output tokens.
Jev only charges you on input tokensbecause it barely outputs anything. Andit is 4 cents per million input tokens.
It is like dirt freaking cheap. Andbecause they output basically nothing,they don't even charge you for output
okens. Um whereas the standard LM aregoing to charge you tons for the outputokens. And then I'll go into use cases,
which is how you would use like astandard LLM versus why you would useJev. You know, standard LMS, chat botswhere you want text in, you want text
out. Coding where you want, you know,code to be produced. Um, and so it'sgreat for things where you needstuff generated, but Jev is really great
for making decisions. That's ultimatelyI've been calling this it's like adecision model. So you give it adecision and it makes a decision and and
it's high agency I would say. So if youneed smart if statements, should I goleft or right? If it's good, put it to
this person. If it's bad, send it tosupport. All that kind of stuff Jeff isreally good at. It's really good at
classifying data. That's a lot of whatI'm going to talk to you about today.And it's really good at real time. So,
it is fast, fast, fast. So, you can putan LLM in a real-time loop in a way thatwas not performant enough with these
other models. I do want to explainexactly what Jev outputs. And so, Iwould highly recommend if you're going
to use Jev, go to their docs. The twoplaces that I'm using the most are theprimitives, so what it returns. And then
I use some of the patterns in cookbooks.So what will Jev return to you? It will
essentially return one of three things.It will return a choice which means youcan give it text, you can give it a list
of choices and it will pick a choice.So, if my question is, what should Iwear on this date and my choice is adress, jeans, workout clothes, ski gear,
or nothing,it'll like maybe pick the dress, right?
And so that is choice. If I selectscore, it's going to say like how severeis this or on a score from one to five,
what it is. This is really good forseverity rankings. So say the examplehere is a great one which is like if you
get a bug, you want to rank it likecosmetic, broken or blocking, Jev willtriage that bug and give it a score. And
then finally, there's a new. It's aversion of a boolean. It basically tellsyou what the likelihood that the answer
to a question is yes. So if the answerto a question is yes or no, is Claire apodcaster? it's going to give you like a
99% nule because there's a 99% chancethat I am a podcaster. Okay, so this ishow you know you would think about the
three things it can return. And again,very simple, but if you've been asoftware engineer, this is like 90% of
software engineering is like doing thesethings, returning a choice, scoringsomething, routing, saying yes or no.
And so it is just so so so powerful. Andthe number one thing I have been usingit for is classification of a vast sets
of unstructured data that would havebeen annoying to classify but is veryhigh value. So I'm going to show you
those use cases and hopefully this willinspire you about how you can use Jev.Okay. So I'm going to pull up Codeex and
show you like two or three examples thatare super simple to run, incrediblycheap, and incredibly powerful. This is
one CTO's, VPs of product, V, you know,chief product officers, listen up. Thisthe one that three years ago I would
have paid truly $100,000 for. So what Ihad Jev do is look at thousands of PRs.I got PRs. I have connection to to
GitHub. Pull every single PR. And thenwhat I had it do, which would have beenso impractical in the past, is I said
categorize all the PRs. And so use Jevto do pair-wise relations. And so thissomething that I found Jev is really
good at is compare like P to PRB andsay, are you the same? Are you working
on the same thematic area or not? Again,
Jev's not going to tell you what thematic area is unless you give itchoices. is it's going to say like yesthese two are related or no they're not
and I've been doing this a lot where I'mtaking vast amounts of data and I'msaying like yes related not related yes
related not related and it's doingclustering which I find very very usefulso I create these clusters of data sets
and that's what Jev did and then I useda really cheap model I use Gemini flashlight to then take those clusters and
categorize them and I first ran it onour marketing site where we only havelike 112 PRs. It cost me 1.1 cent. Iguess they wouldn't round down that.1
penny. And you can see here, it lets mesee how much my work is related tomaintenance, content, tools, site, and
conversion services or docs. Superuseful. Again, it ran very, very, veryfast and very cheap. But then I ran it
on my chat PRD app, which has about2,000 PRs on it to date in this calendaryear. because it has a lot more PRs. It
cost me a lot more money. And by a lotmore money, I mean nine whole cents. Itcost me 9 cents to do this. It tookprobably about 2 minutes. What it did is
it analyzed 1,700 PRs. It did 17 found17,000pairs in those PRs of things that couldbe matched. And then the themes were
labeled by Gemini Flash Light. And itpulled all of the data out. I mean,again,CTO's, CEOs, like I know you're asked
this by the board and by your team andby your boss all the time, like whatpercentage of work is going to whatinitiatives? Like, tell me the
percentage to tech debt. Tell me thepercentage to this product or thatproduct. And you can see here it gotdown. It excluded our docs PRs because
those are completely separate. almost30% of the effort the PRs that we do atChat Pured are around platform security
and infrastructure. on being a goodcitizen and investing in securityperformance and infrastructure and then
of course because we're a chatbot aconversational AI reliability and thensome product areas that we invest in are
like data and integrations documentediting and prototyping which is new andso this was so cheap so fast it is
accurate like I can just tell you yesthis is where we're investing our timeand you can even see as we invest more
in different things over the last couplemonths it goes up. And so this is justan example of how powerful, cheap, and
good Jev can be on large data sets. Andthen I want to give you all one thateven if you don't have like a big repowith PRs everywhere that you can run,
which is you can run this on your localcloud code and codec sessions. So all ofyour cloud code and codec sessions are
stored locally. And so you can actuallyrun this analysis on everything storedon your local machine. So I did that
while Codex kicked off a thread to dothat. And you can see depending on whatyou count as like effort how much I'm
spending on different things. So I do alot of product engineering. And so forgrouping by user turns in January I was
almost exclusively doing engineeringtasks in cloud code and codecs. And thenas you kind of like come into this new
orld now in September less than 40% itlooks like of my tasks are actuallyengineering tasks. I'm doing a lot more
work with agents which makes a lot ofsense. And then I'm doing a lot morelike publishing media. I'm doing a lot
of our videos through. And then you cansee I have a new business. So clientdelivery, family and personal stuff. And
then where it couldn't categorize or itwas unclear. If I take it to sessiondays, you can see again most of mysessions were coding. And now it's like
more equally split across different usecases. And so if you think about this,it is just super useful to do meta
nalysis on all this data that's sittingon your desktop that LMS can totallyparse. that will cost you basically no
money and give you a lot of insight thatI think previously would have been hardto get. So these are two use cases I
think everybody who is coding should do.The third one and I won't show itbecause it's a lot of my privateinformation is I did run this on my
personal Gmail. So I basically said likegiven a subject line and a snippetcategorize whether or not I can like
delete this email like score whether ornot I can delete this email. It did itvery fast and then it gave me very clean
tags that I could go through and thenhave another model work throughcategorized emails. And so this is where
I would say like Jev alone is okay. Jevwith an LLM buddy is super powerful. Sowhat I like to do with Jab is I like to
take a big corpus of information, tagit, categorize it, cluster it, filterit, and then apply really precise AI
actions to the right clusters. And sothat could be take your bugs, take yourhigh severity ones and really triage
them deeply. It could be take youremails, group them into like you candefinitely delete them and ignore them,
delete them, and then work your otheremails with an agent. It's just allthose things that you can do where avery fast but accurate filter can be
helpful. So, I think this is sopowerful. I hope it unlocks your mindson what Jev can do. I want to show youlike the biggest version of this that
I've done just to like kind of give younumbers and ideas of what I'm working onand then I'll show a couple like fun
little apps that you can build with Jevthat I think would have been hard tobuild before. So the example I want to
give is you all have heard me talk abouthis product. So chat PD I'm trying tobuild like a product insights graph.
Basically, I'm like trying to suck ineverybody's data and tell you reallyinteresting insights about it. And I
have used every frontier model there isto try to figure out. And it's justreally hard. There's just too much
nuance in all the data to kind of likegro all this, get it in structuredformat, do the right like structured,
unstructured. It's super expensive. I'vespent thousands of dollars if not tensof thousands of dollars trying to
prototype this. It's just hard. And Ithink I cracked this baby with Jev. Ikind of want to show you what thatmeans,
which is it's not that I like Jev hasdone it all at all. It is actually Ihave figured out where classification,
clustering, and mapping are importantand I put Jeff there. And then I figureout where like the big brains like
analysis,strategy, insights, extraction, and Iput like Astra there. And then likewhere generation is important, I use
like a soul or a Luna. And so if youlook at this map and it's like probablynot that interesting to you all because
it's very internal to chat pd but youcan see like I extract things, I extractlabels, I then quickly assign everything
to those labels with Jev, then I pairthose clusters in those groups withAstra and I say like what the heck'sgoing on across all of this. And then it
extracts more insights and all of asudden I have this really cool product.And just to give you a sense of like the
scope of data here, I'm probably pullingabout a,00and um, 1100 individualsignals into this. So these are PRs,
these are support tickets, these aregranola conversations,these are um, linear tickets. They'relike kind of like all the signals in our
business about what people are worriedabout and what is going on. And then I'mgetting now like this gap between what
are customers telling us they want andwhat are we actually working on and Ican get those trends over time I can get
hem categorizedI can get details about them so it'sjust really super fascinating veryhelpful has made this feature I would
say probably more margin accretive to mybusiness if you all know what that meansbecause before I was really like trying
to brute force with these like brainymodels, what I needed to work on and howI could do this data. And now I can like
combine these like very cheap decisionmodels like Jev with a really brainymodel and get this like interesting
completely hard to generate set of datacontext like auto wiki. It's like reallyunlocked this product and it has taken
about,00raw sources of data and done over like200,000classification and pair-wise groupingsand it probably cost me four bucks on
the Jeff side. It's cost me a lot moreon the Astra side. But I just think likethinking through where classification
like super smart classificationclustering decisions could unlock reallycomplex products and how you might use
as I'm showing here Jev alongside somesmarter models. It's like really blowingmy mind right now and I think is it
should be interesting for those outhere building interesting dataproducts. So, so far I've told you likewhat Jeb is, how you can use it on PR
review or data analysis, how you can useit on your local sessions, how you couldeven use it on your email, but I'm going
to show you two things that I built withit in an afternoon that I think areally cool. And this is my attempt tolike go against the like eyepopping
demos that you're seeing on X. I thinkpeople have shown interesting things,but they actually haven't shown how you
have to build it and how it works to gethat real-time effect. And so, I'm goingto show you these apps live because I
think it's important to kind of likeunderstand where the latency comes fromand what the real experience is as
opposed to like kind of a 30-secondond demo that goes goes viral. ButI do think it shows some like prettycool stuff. So the first one I want to
do ishow I AI audience signals. So youwonderful people give us comments on thehow I AI podcast and I go through them.
I read them every day and I reply tothem when I can, but I've never doneanalysis on them. And there's about4,500 comments and I just wanted to know
like are they good, are they bad, arethey happy, are they sad? And I also
wanted to know if you all had anyepisode ideas for me where I couldextract them and come up with episodesthat I could do like this Jeff one. And
so I hooked up the YouTube V3 API. Youjust have to enable it in Googleconsole. And then I said pull all thecomments and categorize them into
positive, negative or neutral. So thatwould be a choice, right? Or a score
probably. And then I said also use Jevto identify whether or not includes aidea for a future episode and then
analyze all the data and give me adashboard to look at the data. And soyou can see here about half arepositive. There are 58 comments in here
with episode ideas and then there's likequality praise which is praise for ourproduction team which is like how nice
the podcast looks. And so then it mademe this dashboard which is telling me isit positive, neutral, negative or mixed.
And then it gave me an audience requestboard. So you all want like openw weightmodels, you want instinct, you really
want a comparison of Grock and Muse forpersonal use. That's coming soon. Andthen do I use Grock or Brockbot or like
what am I using? And then you can also
see by episode the sentiment which isRyan's very popular three-step AI codingworkflow. 61%positive. We had an awesome cloud code
for product managers episode. 80%positive comments. And then we canactually go through the commentsthemselves. And I using Jev built live
search against these comments. So, if Iwanted to find comments about screenshare, it would search and find very
quickly something that says screenshare. If I wanted to say, "What are thecomments about slop?" I could do slop
and then very quickly it scanned all4,400 and found all the comments thatare related to slop. So, again, this is
very performant, very fast. just forlike behind the scenes, you have to likebatch the results and then score them
and then push the high scores up to gethat kind of performance. So, there isome architecture here. I don't want to
pretend like Jev you just like slap Jevin the middle and evaluate all 4,400that quickly, but you can imagine this
very useful. And so, one, thanks forall the comments. And two, this wasreally, really cool to build. And then
finally, I have to give Astra props.Really nailed the Howa AI podcastyling. So, good job on the front end,but I want to show one last one before I
get us out of here. And then I said thiswas Jev week. So, like and subscribe fora second follow-up Jev episode with one
of our How Ia guests that's going tocome later this week. But I wanted toshow like kind of a fun realtimeuse case for Jev. So if you have been
seeing any of these like real time applications of Jev like playing a videogame or you know playing playing Tetris
doing live search all these things youcan do a lot of real time stuff wheremaking a very quick decision oreturning a set of choices can be very
powerful. Okay, so I built a real timeapp that takes in voice and then usesJev to determine what color is like my
emotion and then returns a quote inreflection of my emotion. It usesOpenAI's real time voice API. It usesJev and then it uses like API ninja
quote API. There's apparently an API forquotes. And so we're going to see ifthis works and how fast it is. I'm
feeling so tired today.I'm feeling very happy today.Okay, the quotes. Oh, there wesee the quote worked. Okay, hold on. Oh,
it's excited.I'm in love today.I can't wait for the weekend.Doing this podcast is the best job ever.Look at that. So, we got all these Okay,
I'm going to turn it off because it'sgoing to like keep listening to me. Lookat all these quotes where likeimmediately it picked the right color.
It picked the right quote. You can tellthe quote API has a little latency init, but you can imagine how fun it would
be to be able to build these real timexperiences. And again, when it's likepick the right color to match thesentence, it can do that really well.
Just for you all to know kind of how itworks, I gave it a list of hex values.So, a list of colors. Every sentence or
like phrase it ingests from thereal-time API, it asks what color or itscores the colors. It gives the topranking score color and then it also
picks from a set list of filters for thequote API. So it filters the quotes andthen it scores them against my sentiment
and then it shows it on the screen. Soagain like running locally not thefastest. I'm sure I could optimize it by
caching a bunch of this stuff and youknow doing all sorts of things but againlike a little like magic use case that I
think is really indicative of the kindsof things you can build with Jeb. Sothat is my Jev 101. How to use it onyour own data how to use it with other
models and how to build some funrealtime experiences in your app intro.Again, as I said, this is Jev episode
number one this week. So, we're going tohave another one midweek with one of ourmost popular guests. So, if you are
xcited about that episode, subscribe,comment, let me know what you want us totalk about, what questions I can answer
about Jeb. And until then, it has beenso nice showing you my favorite newmodel. Thanks for joining How I AI.Thanks so much for watching. If you
enjoyed this show, please like andsubscribe here on YouTube, or evenbetter, leave us a comment with yourthoughts. You can also find this podcast
on Apple Podcasts, Spotify, or yourfavorite podcast app. Please considerleaving us a rating and review, which
will help others find the show. You cansee all our episodes and learn moreabout the show at howiaipod.com.See you next time.