Jev: 8 real use cases this fast, cheap model

1. 0:00 – 4:32

   John Lindquist returns for Jev week

   1. CVClaire Vo0:00

      In some ways, managers are really at risk. This concept of, like, org design and role design, you can put those skills to use when crafting your agents, which is why I am currently running, no joke, 40 GroqBots [laughs] right now.

   2. JLJohn Lindquist0:13

      Whenever you have constrained inputs that are interacting with apps, it's... Jev is a good thing to reach for. This one is, is a favorite of mine because it represents, like, a workplace where you have multiple agents that could all have their own tasks, and avoiding collisions so that they never end up on the, the same space. At each step, it's gonna be able to control these different agents and make sure that they get to their specific task as quickly as possible, and that they never collide or interfere with the other agents. This one's really fun to me because it's a presentation coach. I click the microphone, I'd start talking. I'd have a list of bullet points in either a podcast or something that I wanna make sure I cover, or an interview or anything. It listens through everything I say, and then starts checking the boxes to make sure that I covered everything.

   3. CVClaire Vo0:56

      This model, which is basically free and incredibly fast, allows you to do discovery over data in a way that feels like it's opening up my opportunities and allowing me to, like, look at things that I thought weren't high ROI before. [upbeat music] Welcome back to How I AI. I'm Claire Vo, Product Leader and AI obsessive, here on a mission to help you build better with these new tools. Today, I have repeat guest John Lindquist, who has done one of our most popular episodes to date on the podcast. And here he is joining us on Jev week at How I AI to talk about our favorite new model and all the amazing product-facing use cases that he's discovered testing Jev from TypeSafe AI. Let's get to it. John, welcome back to How I AI. I have to say, this is Jev week on How I AI, and the reason it's Jev week is a lot of models have come out in the past seven days. You and I do a lot of AI things, and yet when we're DM-ing back and forth, we're talking about one thing and one thing only [laughs] which is, which is Jev. So, you know, folks who maybe listened to my episode earlier this week about why I'm hyped on Jev, I would love to hear why y- why you're so excited. I... You know, I'm looking up here. You have 23 demos we could probably look at, so you are going deep Jev as well. Um, what about this model, this framework, this, this way of thinking about building has gotten you excited?

   4. JLJohn Lindquist2:32

      Fast and free are both amazing. Um, if you've ever been tired, uh, sending a basic request to an LLM and waiting around for a bit just to have it, like, call a function, this is now essentially instant. And I think of all the time I've spent on this and the hundreds of demos I made, I think I've spent 73 cents on it. And I, I love the one quote someone said, uh, "I think I'm gonna pass this inheritance of Jev on to my children once I die."

   5. CVClaire Vo3:01

      Yeah, I like the one where it was like, "TypeSafe, TypeSafe is really taking off. They have to have made double digits in revenue by now." [laughs]

   6. JLJohn Lindquist3:08

      Yes. [laughs]

   7. CVClaire Vo3:09

      So yes, fast and-

   8. JLJohn Lindquist3:10

      Yeah

   9. CVClaire Vo3:11

      ... free. Um, I also have had that experience where it ran over, like, I don't know, 20,000 records and it was like, "I have charged you point four cents." [laughs]

   10. JLJohn Lindquist3:21

       Right. And, and it, it exposes those scenarios where I thought... Like, I had five gigabytes of JSON. I'm like, "I had never passed this into an LLM. I just ne- Like, I don't have that budget," on and on. I'm like, "Let's see what happens." And it was, it was 40 cents, you know. And I, I wasted it on a really stupid thing, like trying to organize this or whatever. Um, but five gigabytes of JSON is a lot of text, and it went through and did the whole thing in, uh, probably a couple minutes.

   11. CVClaire Vo3:51

       That's been my experience too, which is I'm looking at corpuses of data that I would never just dump into an LLM-

   12. JLJohn Lindquist3:59

       Mm-hmm

   13. CVClaire Vo3:59

       ... both from a cost perspective, and honestly from a time perspective. It's like, this is not worth-

   14. JLJohn Lindquist4:04

       Yep

   15. CVClaire Vo4:04

       ... my time to even see if something's here.

   16. JLJohn Lindquist4:08

       Yep.

   17. CVClaire Vo4:08

       But then this model, which is, is f- basically free and incredibly fast, allows you to do discovery over data in a way that feels like it's opening up my opportunities and allowing me to, like, look at things that I thought weren't high ROI before.

   18. JLJohn Lindquist4:27

       Yeah. Absolutely. Same. It's just opening doors that weren't, were closed before.

2. 4:32 – 6:15

   What Jev actually outputs

   1. JLJohn Lindquist4:32

      Yeah.

   2. CVClaire Vo4:32

      So just quick reminder for folks, uh, Jev is not like a, the LLMs that you're used to and love. It does not output text. It outputs basically, like, decisions and scores and yes/no probabilities. And so you kind of put information in out, in, and then you get out very structured TypeSafe, very structured-

   3. JLJohn Lindquist4:51

      Yeah

   4. CVClaire Vo4:52

      ... limited set of options. But the ways you can use this are almost unlimited. And so John's gonna walk us through some of his favorite demos of how he's using it. And hopefully we'll get a combination of inspiration on where a real-time, fast, cheap model can build cool things, and a little insight of, like, what else is around this. Because what I've found is I see a lot of these, like, shiny demos on the timeline, and then you're like, "You're hiding your-"

   5. JLJohn Lindquist5:18

      Yeah

   6. CVClaire Vo5:18

      ..."you know, your embeddings, and you're hiding, like, how this is actually working." And so I think if you can lift the veil for us on how some of this works and where it's Jev alone, where it's Jev plus, that would be awesome.

   7. JLJohn Lindquist5:30

      Yeah. I, I would... Just, just to add to your definition of Jev, of it being TypeSafe, I think of LLMs being unstructured to unstructured-

   8. CVClaire Vo5:39

      Mm-hmm

   9. JLJohn Lindquist5:39

      ... where you put text in, you get text out. And this one is similar. You put in unstructured sentences and data, but then you get structured data out. And the power there is it's humans talking to machines, and there are the scenarios where you have functions and APIs and everything set up, but I wanna communicate with them using unstructured data. And I- Uh, we're so used to chatbots.

   10. CVClaire Vo6:05

       Mm-hmm.

   11. JLJohn Lindquist6:05

       And it's just not a chatbot anymore. You need services, you need functions, you need things you wanna throw data at first, but you can talk to it however you want, which is incredible.

   12. CVClaire Vo6:14

       Yep. Well,

3. 6:15 – 8:17

   Demo: real-time voice to-do app

   1. CVClaire Vo6:15

      let's talk about talking to it.

   2. JLJohn Lindquist6:17

      All right, yeah. Uh, so let's go ahead and try this out right away. Uh, so traditional to-do app, we have a list of to-dos, and the fascinating thing here, which I'll try right away, is I'll just say something like, "Book dentist appointment, remove. Buy oat milk, complete. Review pull request, low priority." Now, the speed there was all done through Jev. It does a few passes of looking through the text where it ensures that it lines up. If there's any, like, dictation mistakes, and you say... It doesn't get oat milk quite right, it'll go through and say, "Does, does this quite match up with something, or do I need to do a different... Do I need to categorize this into something that exists?" Like a, a confidence score. Like, it's all a lot of confidence scores there. Um, it matches it with the thing in the list, and then it matches it with the function that it's gonna call, right, based on, based on the way that, uh, the operation that you wanna perform on top of that data. So it's, it's multiple layers of analyze the c- dictation itself to see if that's, like, a valid dictation.

   3. CVClaire Vo7:30

      Mm.

   4. JLJohn Lindquist7:30

      Then analyze if it matches up with a task, and then analyze if it matches up with the operation you want to perform on the task. And it can do that all live, and it even, while you're dictating, it classifies if you've spoken to a point where it can take an action. So as the words come in, it's like, "Is now a time where I can... where I have enough data to make a function call, where I can reliably keep on talking?" 'Cause you notice that entire time I was talking, I never paused, I never hit Enter. It was able to classify, "Is this a moment where I have enough information to take action?" So it's just showing all these pieces of Jev fitting together to give you this real time, like, give me kind of streaming data in, like, unstructured streaming data in, and I'll start putting it in all the places it needs to

4. 8:17 – 10:38

   How sequential Jev calls chain together

   1. JLJohn Lindquist8:17

      go.

   2. CVClaire Vo8:17

      I have a question for-

   3. JLJohn Lindquist8:18

      Which is-

   4. CVClaire Vo8:19

      For folks-

   5. JLJohn Lindquist8:20

      Yeah

   6. CVClaire Vo8:20

      ... that have, just have no idea how to actually put this into place and want more of the tactics, which is, are those a, you know, are those, like, sequential Jev calls? Are you like, "Yes, no, it's time to analyze. Yes, no, it's, um, this is a good transcript. Score, here's the top task that relates to it, and score, this is the top tool call that relates to it"? You know, is this, like... 'Cause people are try- I'm sure trying to put in their mental model, like, like a tool call where you get all that structured data back. Like, y- you know, how, how have you set this up so people can kinda understand how they chain these Jev tools together?

   7. JLJohn Lindquist8:56

      My mental model is, uh, remember back when we used to write code by hand.

   8. CVClaire Vo9:00

      Ugh. [laughs]

   9. JLJohn Lindquist9:01

      There was-

   10. CVClaire Vo9:02

       I do remember

   11. JLJohn Lindquist9:02

       ... if conditions, if, if else switch statements, all these things that were the moment of classification, and if you just think of a, uh, a line-by-line program where things execute, you think of, "Where do I need those conditions?" And so as you're breaking the problem apart, whatever the thing is that you're building, just think of all the conditions, the if else, the, the moments where, um, it either needs to be put into this bucket or that bucket. And so as I was going through this, I built a lot of it backwards because I started with what I was building, marking the to-do as either done or removing it or adding a new one and such. And then like, "Oh wait, nope, now I want the dictation to be able to select it." And so I just kinda built it backwards, and with... It was so easy with, uh, Opus 5.5 recently just to kinda wing it and, like, s- take it different directions and go through scenarios and see, see the cases that, that were exposed. 'Cause I had that final idea of I just want to... If you've ever used a Siri or a, a Google Assistant or any of those, and you've tried to manage, like add to my to-do list, and they're all terrible. And like, how can I build... Like, does Jev enable this to actually be good? And I think it, I think it does. Like, it's one of the most common things, like managing. You could integrate your calendar into this. You could integrate it... Because they're all behind function calls, right, or MCPs or however you wanna build it out. You could integrate anything with this and just have this top level dictation live inference running, and it's something anyone could build cheap and free. Like, it's-

   12. CVClaire Vo10:36

       Amazing

   13. JLJohn Lindquist10:37

       ... amazing,

5. 10:38 – 11:50

   Demo: plain English to function name (grocery cart)

   1. JLJohn Lindquist10:38

      yeah.

   2. CVClaire Vo10:38

      Okay, so you and I both did live voice apps. Mine was much more emo, which was just talking to the screen and getting a color and an inspirational quote. Yours is much more practical, which maybe reflects our different personalities. Uh, let, let's see what other use cases you have, and what I like about what you're showing us is I talked a lot about, like, internal data analysis and clustering and all this stuff. You have such great-

   3. JLJohn Lindquist11:00

      Yeah

   4. CVClaire Vo11:00

      ... inspiration for how you can use this inside products or to do things that were previously really tedious. So I'd love to see another-

   5. JLJohn Lindquist11:07

      Yeah

   6. CVClaire Vo11:07

      ... one of your favorite demos.

   7. JLJohn Lindquist11:09

      Yeah, let's, let's dive in. So, so this one kinda breaks it down into its smallest pieces. Um, if you think of a grocery store and if you think of I have a function call, like getCart, if we remember programming, and you type in something like, "What's in my cart?" And that's how... If you think of classification of regular text to functions. Like, this is just demoing back if, if the previous demo didn't make enough sense. Think of, like, plain English language mapped to a function name, and that's one of the core foundation, foundational pieces of what Jev does. That's that demo. Um, classifying documents sounds like you kinda covered that. This,

6. 11:50 – 13:45

   Demo: data deduplication and record merging

   1. JLJohn Lindquist11:50

      this one I think is incredible because if you've ever run a scenario in... I use Google Contacts, and often you put in the, the same contact twice, and you've probably gone through the process of, like, merging them together. Uh, very common scenario where you have messy data of someone typed in something wrong. And they hit submit, and now you have two, uh, company entries in your database or whatever. This can go through and go through all of the records and kind of merge those records together. So let's click on that. And it can go through and say this one wants to merge, so Cedar Grove Office Products obviously matches with Cedar Grove Office. Ridgeway Data is Ridgeway Analytics, and it can go through and apply those merges for you by matching together and classifying is this one close enough to that one. And again, it's huge data sets merged together in, uh, milliseconds.

   2. CVClaire Vo12:47

      That-- This is, like, my favorite use case of Jev-

   3. JLJohn Lindquist12:50

      Yeah

   4. CVClaire Vo12:50

      ... which is, like, pairwise comparison of a lotta data to-

   5. JLJohn Lindquist12:55

      Mm-hmm

   6. CVClaire Vo12:55

      ... create grouping and clusters. And so in the episode that I just did, we-- I, I did that over PRs, over, like, were these PRs on the same product area or the same thing-

   7. JLJohn Lindquist13:03

      Yes

   8. CVClaire Vo13:03

      ... or the same theme? But you could do data reconciliation, I think is a really useful one I went through recently, and I had, like, 2,000 passwords in my one password. And you know how, like, whenever you do-

   9. JLJohn Lindquist13:16

      Yeah

   10. CVClaire Vo13:16

       ... your one password, it's like update, save, make new, and I'm just, like, all over the place. Had a whole bunch of different ones.

   11. JLJohn Lindquist13:22

       Yep.

   12. CVClaire Vo13:22

       And, like, I was just thinking how much faster it would've been to use Jev and some of the metadata to just pair those up and merge them. Um-

   13. JLJohn Lindquist13:29

       Absolutely.

   14. CVClaire Vo13:30

       And again, it's not like you're, you're giving an example where there's, like, six, you know, things. You're talking about when you have 60,000 or 600,000 things.

   15. JLJohn Lindquist13:40

       Yep.

   16. CVClaire Vo13:40

       Those pairs feel very expensive until you use-

   17. JLJohn Lindquist13:43

       Mm-hmm

   18. CVClaire Vo13:43

       ... something like this.

7. 13:45 – 15:06

   Confidence scores and multi-model validation

   1. JLJohn Lindquist13:45

      And, and something we haven't shown is that there is a confidence score that comes back.

   2. CVClaire Vo13:49

      Mm-hmm.

   3. JLJohn Lindquist13:49

      So if there is a North Star clinic and there's North, North Star clinic services, and it's not-- and they don't quite line up. Like, if you want, you want them only to merge if you're, like, 99% or more confident, you can s- you can set those parameters in there. So that, that, that is a knob you can turn if, as you go through a few checks. And I think, like you mentioned before, is after you do this pass, you can send a smarter LLM against the result and say, "We were-- The data did look like this. Please," you know, "either random sampling over the entire data set, ensure that this was done correctly." And then you can gain more and more confidence over time as you run more and more, uh, you know, large data runs. So.

   4. CVClaire Vo14:32

      One other thing you can do on that, in addition to validating the data pairs, is you can describe them. And so what I've done-

   5. JLJohn Lindquist14:39

      Yeah

   6. CVClaire Vo14:39

      ... is do the pairs and then run... It doesn't even have to be an expensive model. But say, like, "We've decided these two are the same. Explain to me why." And it can give a short-

   7. JLJohn Lindquist14:49

      Yeah

   8. CVClaire Vo14:49

      ... kind of like, "It's the same because they both say services," and, you know, whatever it is. Um, and so you can get sort of a, a quantitative matching and then a, like, qualitative explanation with two relatively cheap models once you've narrowed that down.

8. 15:06 – 18:23

   Demo: Jev as a multi-level app router

   1. JLJohn Lindquist15:06

      Yeah. Yeah. Great stuff.

   2. CVClaire Vo15:08

      Okay. Great stuff. We are just psyched about all [laughs] this.

   3. JLJohn Lindquist15:11

      Yeah.

   4. CVClaire Vo15:11

      I was like, "I could watch all of these." [laughs]

   5. JLJohn Lindquist15:14

      Th- th- this one's fun. It's kinda hard to, to show off, but essentially if you think of having, uh, your entire application, and say you have-- go back, going back to the to-do list, and say that's one part of your much larger app. You can build additional abstractions around that so that if on your landing page someone comes up and either has a command bar, like a command K, or other demos... Or sorry, other, other parts, omnibar, any place to put in, uh, text, and they start typing in the, uh, you know, dictate or voice or something. They can't remember what the to-do app was called. And you notice this is to-do, to-do, and I typed in to-do. It's able to go in and match against the tool, and then you could, uh, type additional things where you'd say, "Go to the to-do app," and then take this action. Like I, I would know in my head, "Oh yeah, there was that pull request." And, and then take all the pull requests and, um, mark those as low priority. And so if you think of this multi-step, uh, multi-step classification where you can build your smallest tools and then build another abstraction of Jev around them where it can pick which of the smallest tools to use. Um, this kind of goes back if you've, if people have set up MCPs where there's, uh, abstractions around them. I think Executor and others where it picks which MCP to use. There's the tool itself, and then there's a layer where you're picking which tool to use, and you can build that as, um, as many layers of that as you want. That, that's what this is kinda talking about. It's, it's difficult to show off, but just imagine saying, "Go to the to-do app, do this," and yeah.

   6. CVClaire Vo17:00

      What I would say is, like, w- r- Jev is a router is maybe, like, what-

   7. JLJohn Lindquist17:04

      Yes

   8. CVClaire Vo17:04

      ... what's helpful.

   9. JLJohn Lindquist17:05

      Jev is a router. Love it.

   10. CVClaire Vo17:05

       A very, very fast router. And so, you know, this use case you're saying, like, as I have text, route to the right tool, and then if I can infer from the text the job to do in the tool, I might as well infer that and, like, route as far down-

   11. JLJohn Lindquist17:20

       Yep

   12. CVClaire Vo17:20

       ... the user journey as I can infer from these, like, structured options. Other practical-

   13. JLJohn Lindquist17:27

       Yes

   14. CVClaire Vo17:27

       ... use case I'm thinking, a lot of people start to build these, like, Jev-powered coding harnesses where it's like given the first input, let's route to the right model, let's pick the right tool, all this kind of stuff. And so you can-

   15. JLJohn Lindquist17:38

       Yep

   16. CVClaire Vo17:38

       ... imagine very fast configuration, very fast routing, very fast, like, I don't just have to take the first step and then think and take the next step. I can actually just build the chain and then execute it. Um, and so I think these, these decision models like Jev, w- if they remain fast and cheap, can be a new way to think about navigation of your-

   17. JLJohn Lindquist18:02

       Yes

   18. CVClaire Vo18:02

       ... user experience, whether that's a kinda like front-end experience or a dev tool or whatever it might be.

   19. JLJohn Lindquist18:09

       Yeah. E- even search-

   20. CVClaire Vo18:11

       Yep

   21. JLJohn Lindquist18:11

       ... to a certain extent of, um- Search for a long time has been, has had a lot of really expensive options out there.

   22. CVClaire Vo18:18

       Mm-hmm.

   23. JLJohn Lindquist18:19

       And I don't know how far you could push this into replacing certain

9. 18:23 – 19:35

   Architecting around Jev

   1. JLJohn Lindquist18:23

      search engines, but yeah.

   2. CVClaire Vo18:24

      You know, I gave an example-

   3. JLJohn Lindquist18:26

      Yeah

   4. CVClaire Vo18:26

      ... um, when I did this, like I had 4,000 YouTube comments, and I was trying to do live search through this. And I want to demystify some stuff for folks because it actually isn't super fast to like search over all of these results and say like, "Yes, no, this is a good result," and rank it. But you can like-

   5. JLJohn Lindquist18:44

      Mm

   6. CVClaire Vo18:45

      ... cluster like groups of like 30 or 50 or 60 and score them and rank them. Um, and that is a much faster way to build, quote unquote, "real-time search with Jev." So you know, for folks... A- again, I just wanna like give people a little bit more specifics about what you're seeing on the timeline, which is you do have to think about how you're gonna architect around Jev. It is fast-

   7. JLJohn Lindquist19:07

      Yes

   8. CVClaire Vo19:07

      ... it is efficient, but it's not like completely latency free. And so you may need to g- go look around, see how people have done search, and then if you find one you like, go into GitHub, figure out how they actually made that surf- search super performant, because I, I've seen a couple different strategies out there of to how to do this.

   9. JLJohn Lindquist19:25

      Yeah. And caching and all that stuff can help as well.

   10. CVClaire Vo19:28

       All, all, all the caching.

   11. JLJohn Lindquist19:30

       But like-

   12. CVClaire Vo19:30

       Yeah.

   13. JLJohn Lindquist19:30

       Yeah, all-

   14. CVClaire Vo19:30

       It's easy when everything's locally cached and you just go over the-

   15. JLJohn Lindquist19:34

       Yeah

   16. CVClaire Vo19:34

       ... the full

10. 19:35 – 24:29

    Demo: Jev vs. traditional LLM at chess (speed and cost benchmarks)

    1. CVClaire Vo19:35

       corporate.

    2. JLJohn Lindquist19:35

       Yep.

    3. CVClaire Vo19:35

       Let's talk about games, because this is something-

    4. JLJohn Lindquist19:38

       All right

    5. CVClaire Vo19:39

       ... I haven't covered yet, but I think is a really cool demonstration of some of the, um, ideas inside Jev.

    6. JLJohn Lindquist19:48

       Yeah, I, I built this one 'cause I wanted to show off Jev versus a traditional model. I picked a free model on OpenRouter. This is like a, a low reasoning model. I, I think it's a Kimi or something. I'm not sure, but, um-

    7. CVClaire Vo19:57

       And for people that aren't watching, it's pl- it's playing chess.

    8. JLJohn Lindquist20:01

       Yes. It... This, this is playing a game of chess, and Jev are the white pieces, the LLM is- are the black pieces. And here, this is set so that, um, Jev is able to, uh, think through all the possible moves, and then it ranks the highest three moves, and then it thinks through all the next possible moves from there, and based on that two-step reasoning, picks the best next move. And so kind of like router, it's kind of like, um, spidering into the po- potential best moves, um, and then coming back to what the next best move would be. And it's still doing this in sub one second times and using AI and inference. Whereas if you pass the same data set to an LLM, the amount of reasoning it takes... This, this is playing blitz chess, so there's, there's a one-minute timer. Jev would be able to complete this entire game and beat it, and the LLMs would eventually get there. They'd be much more expensive, and they might not even perform better. Because in scenarios like this where it's focused on routing and the next best thing to happen, the, like the amount of data that you can crawl into and find the, the next best move can kind of balance out between these two.

    9. CVClaire Vo21:22

       Mm-hmm.

    10. JLJohn Lindquist21:22

        Um, unless you're getting to like really super high level, like creative chess where you're trying to stump someone by feinting certain moves or whatever, but I'm not at...

    11. CVClaire Vo21:29

        [laughs]

    12. JLJohn Lindquist21:30

        I, I'm decent at chess. Um, I can beat my kids.

    13. CVClaire Vo21:34

        Perfect.

    14. JLJohn Lindquist21:34

        But-

    15. CVClaire Vo21:35

        For now.

    16. JLJohn Lindquist21:35

        Sometimes.

    17. CVClaire Vo21:35

        [laughs]

    18. JLJohn Lindquist21:36

        For now.

    19. CVClaire Vo21:37

        Yeah.

    20. JLJohn Lindquist21:38

        Um, and if you look at the, um... It has some benchmarks over here that Jev was 10 times faster in the average move, and it was four times cheaper, and this was a, a very, um, space money alphas on, uh, low reasoning, so with very, very tiny context windows.

    21. CVClaire Vo21:55

        Yeah. What, what I think this-

    22. JLJohn Lindquist21:56

        Yeah

    23. CVClaire Vo21:56

        ... demonstrates is, one, there are scenarios in which speed is a, is an advantage. Um, this is, this is putting in the context of a game, but, um, it's very obvious how much... I mean, 10X faster is not incremental, and so it-

    24. JLJohn Lindquist22:13

        Yes

    25. CVClaire Vo22:13

        ... really does unlock this concept of real time. And I do think about, uh, going back to, uh, when I was a, a young, a young product manager and did a lot of like conversion testing and, and growth work, and like speed did matter in a lot of user-

    26. JLJohn Lindquist22:32

        Yeah

    27. CVClaire Vo22:32

        ... experiences, especially consumer, on what won. And so I do think... You know, we, we hear a lot on the coding side about model optimization from a cost perspective, but I think we're gonna start to see like latency wars really heat up as well-

    28. JLJohn Lindquist22:47

        Yeah

    29. CVClaire Vo22:47

        ... which is like the faster the response, the more, like more magic you can unlock that feels like an if/else statement, the, the better.

    30. JLJohn Lindquist22:56

        Yeah.
11. 24:29 – 28:21

    DOM interactions as a decision set, not an infinite canvas

    1. JLJohn Lindquist24:29

       so far.

    2. CVClaire Vo24:30

       And I, I think this can demystify too some of the other kind of fun eye-popping use cases that I have seen, like browser use, right? If you look-

    3. JLJohn Lindquist24:41

       Yes

    4. CVClaire Vo24:41

       ... like at a, at a website, navigating it with a mouse feels like an infinite set of possibilities, right? You can like move it anywhere.

    5. JLJohn Lindquist24:48

       Right.

    6. CVClaire Vo24:48

       Any pixel on the screen is where your mouse can go and click. But if you actually look at the DOM, there's probably like 10 clickable buttons on a page.

    7. JLJohn Lindquist24:56

       Yes.

    8. CVClaire Vo24:56

       And so if you can very quickly say, "There's 10 clickable buttons on this page, which one do I click? Like which one do I want to click?" That, that demystifies that concept of browser use through Jev, 'cause you're really just taking a limited set of interactions available, um, to us and our human brains. It looks like an infinite canvas of pixels-

    9. JLJohn Lindquist25:16

       Yeah

    10. CVClaire Vo25:16

        ... but to, to a model like this, it can just be a set of decisions. You know, I've also seen on gaming, um, you know, Jev plays Tetris. It's because Tetris is four options of your shape-

    11. JLJohn Lindquist25:28

        Yeah

    12. CVClaire Vo25:28

        ... and then like 10 options left or right. That's all you, that's all you have. And so-

    13. JLJohn Lindquist25:33

        Yeah

    14. CVClaire Vo25:34

        ... um, you know, these like horizontal scroller games are just you can like jump, you can walk, you can do, you know, any of those things. And so because these decisions can happen so fast, if the interaction of your app, even if it's a game that feels complex or a web app that feels complex, it probably can be distilled down to a dozen choices. And then you can-

    15. JLJohn Lindquist25:54

        Yeah

    16. CVClaire Vo25:54

        ... explore those choices, and because Jev is so fast, it almost feels instantaneous to chain those together.

    17. JLJohn Lindquist26:00

        Yeah. It... I'd almost say in, like you mentioned in a game, in a platformer or in, in a 3D game, you might be thinking about all the things the character could do, but the input is actually the controller. And so the, the limited amount of options you have are all the buttons you could press, how you could press them in various ways, which is still a lot, but there's a very constrained input there. And whenever you have constrained inputs like, uh, that are interacting with apps, it's, uh, Jev is a, a good thing to, to reach for. And, and to contrast the, the DOM example, if you think of taking a screenshot of a webpage and saying what on the screenshot or what on our do- what on our product or website might be confusing to a user, that's not a Jev thing. That's an LLM that can look through an image, it can compare it to all the images and products in the past. It can say... It can look through the layouts and the colors and the contrasts and all those sorts of things, and come up with a discussion for you of here's, here's some points that might be, uh, because the fonts are, uh, gnarly or whatever. And so that, that's where trying to delineate between the two when, when to go which way.

    18. CVClaire Vo27:12

        As AI adoption grows, so do your company's security risks and requirements. New frameworks, audits, and vendors keep piling on, but your team isn't getting any bigger. Most compliance tools promise automation, but you're really stuck doing all the work by hand, chasing screenshots, validating controls, and playing catch-up when audit season comes around again. Vanta works differently. Their agentic trust platform is built to scale with you, not slow you down. With over 1,400 automated tests across 400 plus integrations, Vanta collects evidence and monitors your controls year-round, cutting your audit prep by 82%. Plus, you can access the Vanta Agent everywhere you work, whether your team is in Vanta every day or living in tools like Claude or Cursor. Want proof? Over 16,000 fast-growing companies like Ramp, Writer, and Harvey save time and resources while scaling fast with Vanta. My listeners can get $1,000 off by going to vanta.com/howiai. That's V-A-N-T-A.com/howiai for $1,000

12. 28:21 – 30:28

    Demo: Wikipedia “path to philosophy” route mapper

    1. CVClaire Vo28:21

       off. Okay. Let's, let's do one more. Do you have a, a last favorite one? Um, I'd love to-

    2. JLJohn Lindquist28:28

       Oh, yeah, me.

    3. CVClaire Vo28:29

       I know there's so many.

    4. JLJohn Lindquist28:31

       This one I'll run... This is super fast.

    5. CVClaire Vo28:33

       Okay.

    6. JLJohn Lindquist28:34

       On Wikipedia, there's every page leads to philosophy.

    7. CVClaire Vo28:38

       Uh-huh.

    8. JLJohn Lindquist28:38

       If you've ever done that. So you can put in any person in the world or any topic, so like LeBron James or whatever.

    9. CVClaire Vo28:44

       [chuckles]

    10. JLJohn Lindquist28:45

        Run to philosophy. So you give it an end goal, and it looks at the entire page and, uh, it's going through the Wikipedia API and mapping that route-

    11. CVClaire Vo28:55

        Oh, that's cool

    12. JLJohn Lindquist28:56

        ... and finding from the person to philosophy. Um, so mapping routes as well, like giving it a final destination, and it can crawl its way there. This one is, is a favorite of mine because it, it represents, if you think of a, like a workplace where you have multiple agents that could be doing multiple th- uh, that could all have their own tasks and avoiding collisions so that they never end up on the, the same space. Um, so this is going to... The job brief is deliver the fragile blue bin to staging, bring medicine to the ward, and inspect the spill. But then you have three people that you can assign to. So this can move one tick, and you can see at each step it's gonna be able to control these different agents and make sure that they get to their specific task as quickly as possible, and that they never collide or interfere with the other agents. So if you think of this in like agentic programming or any sort of parallel work you'd be doing to ensure that, uh, you have things running and you wanna like analyze to help steer them in different directions so they never collide or disrupt each other or touch the same files or, or whatever, you can ask Jev to take in like every single s- It's fine if it goes in every single step and it checks. It doesn't have to plan out the entire thing ahead of time. It can do just in time. Ob- obviously, planning out ahead of time with a decent route probably helps it out, but I, I think as we get more and more parallel agents and swarms and whatever words you wanna use for that, uh, this, this concept will become even more and more important.

13. 30:28 – 33:36

    Demo: multi-agent coordination and collision avoidance

    1. CVClaire Vo30:28

       Well, and what, uh, you know, this has made me think is Jev unlocks efficient inefficiency, which is like it is kind of like inefficient to map out all possible routes, rank them, double-check that they're not gonna collide. Like, that is actually pretty inefficient. It's, it is effect- it, it's an effective way to solve the problem, like it's an accurate way to solve the problem, but it feels inefficient. And in the past we were just like kinda like tossing these problems... In the past, last week, we were tossing these problems to these like big, brainy LLMs and being like, "Think really hard. Just think really hard and come back with a plan." And instead here you can really very quickly evaluate all those options, stack rank them, and go through the sort of like inefficient universe of options and come to the right, kind of like the right, right conclusions. Very... It's, this is why I'm like very excited about this model, 'cause it's all these things that in the past I think with, maybe with regular, regular LLMs, you know, were trying to do this delineation. It felt, it feels like you can do things you never imagined before, right? Like-

    2. JLJohn Lindquist31:37

       Mm-hmm

    3. CVClaire Vo31:37

       ... I can create a 3D video game. I can like generate images out of words. Like, these LLMs are generative, and they allow you to create things you never be a- before, create before. Jev, I feel like allows me to do all the things I wanted to do, but they felt too cost ineffective, too slow-

    4. JLJohn Lindquist31:55

       Yeah

    5. CVClaire Vo31:55

       ... and not really worth it. And there are a lot more, [chuckles] a lot more things for me on that pile-

    6. JLJohn Lindquist32:00

       I love that

    7. CVClaire Vo32:01

       ... than there are on my like 3D gaming idea pile. And so that's why maybe as a builder I'm so excited about this particular model.

    8. JLJohn Lindquist32:09

       Yeah. I, I love that. I 100% agree. It's even like I have so many APIs and so many apps and so many things available, how could I... And the exploring with this is quick, and the turnaround is quick, and it's exciting to see. Like you feel, you feel progress as you're working with it, whereas with gen- with generative... And I, I, I feel the almost because if you're a developer, you understand the APIs, you understand the reasoning, you understand classification, whereas with the LLMs a lot of it is guessing, and if something goes wrong you need to think of a different sentence to say. With decision models you can like, you can think the decisions it made, and you're like, "Oh, well let's take this different route." It feels much more like programming, to be honest, where you're, you have... I feel much more in control when I see something goes wrong because I know all of its options and I know what I want, and it's fun. It's just so fun.

    9. CVClaire Vo33:05

       It's so fun. And I love that you say it, it feels like program- it feels like the smartest function. Like, it just feels like a function-

    10. JLJohn Lindquist33:14

        Yeah

    11. CVClaire Vo33:14

        ... that has a lot of intelligence built into it, but it's still a function. Like, it's still, I input specific things-

    12. JLJohn Lindquist33:19

        Yeah

    13. CVClaire Vo33:19

        ... and I get out very specific things. I can just, um, reason with it a lot more. Okay. So if you, if you've stayed with this podcast this long, it's just John Clare, two Jev boys, [both chuckling] excited about, excited about-

    14. JLJohn Lindquist33:35

        Welcome to the club,

14. 33:36 – 36:56

    Demo: real-time presentation coach

    1. JLJohn Lindquist33:36

       yeah

    2. CVClaire Vo33:36

       ... decision, decision models. Uh, John, okay, I'm gonna give you... 'Cause you, we're at, we're at example 9 of 23. I'm gonna give you one last chance to pick a, a favorite from your remaining demos before we do lightning round and kinda get you, get you outta here.

    3. JLJohn Lindquist33:53

       Th- this one's really fun to me because it, it's a presentation coach. And so I would, I, I click the microphone and start talking, and as I was talking I'd have a list of bullet points on either a podcast or something that I wanna make sure I cover, or an interview or anything. And as I start talking, it, it listens through everything I say and then starts checking the boxes to make sure that I covered everything. And so let's say this, this real time, uh, as someone who likes to go off on tangents, you know, I'll get passionate about something and I'll go a dif- different way. You're under the clock, and Jev can be your, uh, stay on track. And I think it's kind of a almost a life coach of like, "You need to make sure you hit all these things. I'm gonna watch every single thing you do, every single word you say, and make sure that..." And you could start doing like red flashing lights of, you know, "You have limited time left. You still need to say," all this sort of stuff. So this, this one speaks to me as a, uh, a teacher, presenter, workshop giver, um-

    4. CVClaire Vo34:53

       I love this. I just-

    5. JLJohn Lindquist34:54

       Yeah

    6. CVClaire Vo34:54

       ... did Lenny's, Lenny's, um, summit recently and, and gave the kinda like opening talk, and I hate those monitor... 'Cause they're like, "Hey, you can only have three bullet points on here." I'm like, "That's fine, whatever."

    7. JLJohn Lindquist35:07

       Yeah, yeah.

    8. CVClaire Vo35:07

       But then I like to ramble. I like to work the crowd.

    9. JLJohn Lindquist35:10

       Yeah.

    10. CVClaire Vo35:10

        I don't know if I've said that, that bullet point or not. And so imagine you-

    11. JLJohn Lindquist35:13

        Yeah

    12. CVClaire Vo35:13

        ... speaker notes could check off as you go, and then even progress your slide for you, like, "You're done here."

    13. JLJohn Lindquist35:20

        Yep.

    14. CVClaire Vo35:20

        "Let's, let's move on."

    15. JLJohn Lindquist35:22

        [chuckles] Yeah.

    16. CVClaire Vo35:24

        Um, and, and does this just take in... I, this, I do have a question about your real time, again, because I like to make this very tactical for people and I'm curious how you approach this. On these real time voice ones, are you constantly putting in the long stream of text to, to the moment? Like, are you... Or, or are you doing like a-

    17. JLJohn Lindquist35:44

        Yeah

    18. CVClaire Vo35:44

        ... moving window of snippets? I'm just curious how you're doing that.

    19. JLJohn Lindquist35:48

        Yeah. It, it's doing a... I can't remember exactly the logic, but it is analyzing what you've said up until then. It is like concatenating each word until it reaches a point where it can take an action, and it essentially transforms that into the payload that it's gonna send over to the function.

    20. CVClaire Vo36:05

        Yeah.

    21. JLJohn Lindquist36:05

        And so that one is... It's stored in the, it's in the history. Like, you could have undo and whatever because they turn unstructured into structured data.

    22. CVClaire Vo36:13

        Yep.

    23. JLJohn Lindquist36:14

        Um, but yeah, it, it, it's a window which then gets chopped off and then you start your new, uh-

    24. CVClaire Vo36:20

        Well, and it-

    25. JLJohn Lindquist36:20

        ... window of, of text

    26. CVClaire Vo36:21

        ... this is a good moment to tell people exactly how much this thing costs. I, I feel like last time I checked it was like four cents per million input tokens, and nothing for output tokens unless you, like me, are using Vercel AI Gateway right now, and then it's Completely free.

    27. JLJohn Lindquist36:37

        Free.

    28. CVClaire Vo36:37

        Completely free. [laughs] Um, so-

    29. JLJohn Lindquist36:41

        Yeah

    30. CVClaire Vo36:41

        ... so again, like, I can talk, I can yap, but I don't know if I can yap a million tokens quite yet, so. [laughs]
15. 36:56 – 39:54

    Quick recap

    1. CVClaire Vo36:56

       real time voice use cases, sort of like pairing, deduplication use cases, routing and navigation, and like multi-turn, multi-path kind of scoring, which I think is awesome.

    2. JLJohn Lindquist37:09

       Mm-hmm.

    3. CVClaire Vo37:09

       And then this one, which is like just keeping you on track. Did you do the things you said you would do while you're, while you're talking? Did you present the point?

    4. JLJohn Lindquist37:19

       Yeah. The mon- the monitor, right? The thing that's always watching you and making sure you're-

    5. CVClaire Vo37:23

       Love it. Um, let's-

    6. JLJohn Lindquist37:25

       It's fan... Like-

    7. CVClaire Vo37:25

       Let's double check. Did we do it? Did we, we covered what Jev is, why speed-

    8. JLJohn Lindquist37:30

       [laughs]

    9. CVClaire Vo37:30

       ... what it costs. Does the app stay in j- where does it fall short? Did we call, did we talk about where it falls short? A little bit about-

    10. JLJohn Lindquist37:38

        Uh

    11. CVClaire Vo37:38

        ... where you'd wanna pull in an LLM, but have you ever tried to throw Jev at something and didn't do a good job?

    12. JLJohn Lindquist37:46

        Uh, not yet. I don't think I've spent enough time since it's been out for a matter of days, um, really analyzing how smart it is versus what... I, I would say there, there's times it falls short where I add another layer of classification in, so I do multi-step. And I've seen people critique it where they only do a single pass-

    13. CVClaire Vo38:08

        Yep

    14. JLJohn Lindquist38:08

        ... and they think, "That's not good enough. I'm not gonna use this anymore."

    15. CVClaire Vo38:11

        Yeah.

    16. JLJohn Lindquist38:12

        When I run into those scenarios, I'll do multi-pass where I'll do one classification layer, and then another, and then another. So I think it might be a disconnect there between some people who try it out first and they're like, "This isn't good enough a classification," or they don't, like, define the classifications, or they don't have strong enough APIs or... So I, I think just when it falls short, um, because it's fast and free, don't worry about adding just another layer. And then, uh, over time you could compress those down into a single layer as you get more and more data that reinforces what the exact flows are that you need to go from, from step to step. Again, it's all so brand new that-

    17. CVClaire Vo38:54

        Yeah

    18. JLJohn Lindquist38:55

        ... uh, and it's getting smarter. Like they, they said, we, these models are as, you know, they're as dumb as they'll ever be.

    19. CVClaire Vo39:01

        Yep.

    20. JLJohn Lindquist39:02

        And even these models will get much, much smarter.

    21. CVClaire Vo39:04

        Well, and I'll, I'll give folks one other tactical piece of advice, which is you may use a specific type of Jev call, like a, like a null, and you may say, "I think this is a yes/no question." And you may run it and you may be like, "Actually, it's a choice question," or, "Actually, it's a score question."

    22. JLJohn Lindquist39:21

        Yeah.

    23. CVClaire Vo39:21

        And so you can also test different ways to ask for different types of decisions to get it more accurate.

    24. JLJohn Lindquist39:29

        Mm-hmm.

    25. CVClaire Vo39:29

        Um, so I've, I've had that experience a couple times where, like, my first concept of how I would classify or make a decision was just ultimately, like, not the right one, or I had to layer one and then, then the other. So that's something-

    26. JLJohn Lindquist39:39

        Right

    27. CVClaire Vo39:39

        ... just for people to look at. Okay. As, as you said, we could just... I could go on truly forever about this, but I won't. I will spare our audience. Um, let's get to lightning round, and then you and I will go back to just classifying all the things for

16. 39:54 – 46:04

    Lightning round and final thoughts

    1. CVClaire Vo39:54

       $0.

    2. JLJohn Lindquist39:55

       Sure.

    3. CVClaire Vo39:55

       It has been, I don't know, a y- I was reflecting, it's been almost a year since we last chatted. It was so cute. We were talking about Claude Code and, like, aliases to dangerously skip permissions, and just like babes in the wood, and now life is completely different. You know, what, what are you, other than Jev, what are you really into these days? What has changed for you in terms of AI engineering? What are the big, big moves in the last year?

    4. JLJohn Lindquist40:22

       Some of the funniest stuff recently with the latest models, when it gets into multimedia and you're working with video and 3D and everything, is buying extra hard drives and distributing work between Mac Minis and such, which is not a pro- like, it's a problem.

    5. CVClaire Vo40:39

       It's a real problem.

    6. JLJohn Lindquist40:40

       You're li- it's a problem. And like mana- having AIs, like, watch my disk space on my laptop, being like, "Bro, it's time to hand some over to the SSD that..." Anyway.

    7. CVClaire Vo40:53

       Um-

    8. JLJohn Lindquist40:53

       Um.

    9. CVClaire Vo40:54

       Hold on. I have to pause. You and I are-

    10. JLJohn Lindquist40:56

        Okay

    11. CVClaire Vo40:56

        ... exactly the same person.

    12. JLJohn Lindquist40:58

        [laughs] Okay.

    13. CVClaire Vo40:58

        Because truly yesterday, I put the SSD in. I have a pinned Codex chat, and the pinned Codex chat is-

    14. JLJohn Lindquist41:05

        Yeah

    15. CVClaire Vo41:05

        ... clear up disk space, and it has two, um, skills in it I call on a regular basis. One is move media files onto the hard drive. So it's like-

    16. JLJohn Lindquist41:15

        Mm-hmm

    17. CVClaire Vo41:16

        ... it takes, it finds all the places I store all my MP4s, and it goes, finds them all in different places, and then it organizes them on my SSD. And the other one is it goes, like, in the crevices of my computer and finds work trees that are, like, still running local servers and, like, clear, like-

    18. JLJohn Lindquist41:31

        Yes

    19. CVClaire Vo41:31

        ... busts them of the cache, clears... I, like, I was at a client's, uh, office and they saw my computer, like, freeze up and you're out of memory, and it's like, they were like, "Can I buy you, like, a new computer?" I was like, "Dude, this is a new computer. This is just-"

    20. JLJohn Lindquist41:44

        Yeah

    21. CVClaire Vo41:44

        "... the problem." And then I set up, uh, I killed the open clause, RIP, but then-

    22. JLJohn Lindquist41:50

        Yeah

    23. CVClaire Vo41:50

        ... I set up this stack of Mac Minis as remote Codex machines to run all my, all my code on because I-

    24. JLJohn Lindquist41:59

        Yeah

    25. CVClaire Vo41:59

        ... it's, it is a real problem.

    26. JLJohn Lindquist42:02

        It's a problem.

    27. CVClaire Vo42:03

        I, I'm sure-

    28. JLJohn Lindquist42:04

        It's a problem

    29. CVClaire Vo42:04

        ... I'm sure we'll all just, we'll go to the cloud, right? I l- I like-

    30. JLJohn Lindquist42:08

        I like-

Episode duration: 46:06