Wordware AI, an IDE for programming using natural language - interview with Filip Kozera

so could you tell us more about uh what is uh wordware trying Ware wordware is uh like software but with words so wordware uh and we are essentially solving um one kind of the same problem but in long term it means that AI Engineers who will be programming with English as their main programming language uh and using it precisely in order to achieve kind of AI agents um they don't have the right tooling and we are providing them with uh with the right kind of schema on how to work on this and giving it uh to them in a web hosted ID where they can interact with loads of different foundational models and orchestrate them in the right way we are not a chatbot Building Company rather we create kind of complex AI orchestrations in the short term what we're solving is all of the prompting is happening inside of codebase right now which means that the domain experts in order to close a feedback loop need to communicate with engineers and that can take anywhere between half and a week uh half a day and a week depending on how efficient your organization is we close that Loop to be couple minutes that's pretty impressive yeah sounds sounds good can you show us the demo sure let's go so as mentioned prompt thing is the new programming uh let's F in nice uh and uh let's just as in any programming language uh we're going to do um hello world first MH so here we'll just have you know typical hello world and we're going to say hi to uh name name becomes an input to this function MH and now we do a slash command this on purpose looks a little bit like notion and you get different um different function so generation is basically an llm call uh you get to create another input and those things Loops conditional statements and code execution essentially enable us to be like a complete touring uh machine you know okay uh so Loops are kind of the very important Concepts that we take from normal programming and we move them uh and we make them work with inherently non-deterministic nature of llms so let me just start with a generation mhm uh we have a bunch of different models to play around with okay you can select them on the spot right yeah so you know we have command R plus online mix roll on Gro uh we have clo free Sonet and they are some of them are image um enabled so we have this generation right and I'm going to click go right now I'm going to say hi to myself because I never remember people's names so you're providing the parameter right you're inputting the so we iner the parameter and then okay I understand so this is super simple right hello philp now this application is an API that you can input into your codebase so you don't have to play with uh prompts inside you can share it and have a hosted version of this application which looks like this and it will always just kind of show the outputs uh but you know it gets a lot more interesting once we are starting to uh say for example generate a uh Json in the form of name at name we're going to rename and uh and greeting add new generation and you know what this actually is not looking pretty good so I'm just going to call this greeting all of these have the same ID so you can once you change one you change all of them so you're referencing the previous uh exactly so you can kind of see how this is a little bit closer to actual programming you know and from here English but in English correct so now I can call a separate function and the way that we actually work in companies is Engineers create this these functions often with code here we actually have 11 lab speech syntheses this is written in code um yes so it's like foreign function calls right exactly it's like assembly in C yeah you're trying to like ingest or like I know uh you're incorporating actual code into this uh environment exactly so normally the people who can judge the outputs of the llm get to work with pre-written functions these functions can be database call uh V database call uh an API connection and it can also provide inputs and outputs right so now I reference the output and from here we've got a kind of a you know super simple application uh that will generate say hi to philli and it will generate the right Json and then it will read the 11 will do the 11 L speech sentences and output an MP3 oh nice so we created kind of like a pipeline right correct we call this we call this uh prompt chains or Cascades um and this is fully functioning word app right is there is there a plan for example because there's like a lot of things you could like publish in in a sort of like a directory right a Community Marketplace something like that we are doing a product L growth with community at it at its bottom the community is providing the fun examples of what is even possible we are swimming a little bit ahead of the wave here and it's important to have Community to show what's even possible with AI agents then there's smbs which use our product a bucket for the whole llm so they put the API inside of the code base and the third one uh Third Kind of category of people we work with is Enterprises that um they essentially take our product create poc's often their Consulting and instead of just outputting a power presentation to their clients they give them a working product right so let me show you just something a little bit cooler I don't know if you've read this paper called react it is basically prompting an agent to correct itself uh write code and use tools in order to solve an arbitrary task so it's formed in this specific way of question fa action input observation and it repeats that n n amount of times in order to actually answer your question right so you can see that here we using the things that you're familiar from normal programming like looping uh like tool s like conditional statements and actually trying to achieve something here I really like the interface it's like very fluid and like very convenient very minimalistic at the same time and it seem like it's like easy for people who are new to programming to start operating that also it's inspired by notion but it seems to work better than notion so here we I actually asked it what is the Bitcoin price today the thought was I need to find an API that provides the thing it reads Google it realizes coin gecko API is a free uh API writes the right code uh executes it on our servers and uh it actually knows when it's done that's very important actually it's a conditional statement that actually works with the llm nature of how non deterministically it it's working so and price of Bitcoin 60,000 we uh yeah I've got a plenty of cool examples maybe I'll just show one more so you can cut it cut it um this is because of the way that we execute code I can give you a input as an idea and a name for any startup it would then use HTML to write oppus code and and it will host it so it immediately gives you a website of your business so let me just idea I going to say help customers move their pets abroad and name is going to be Dogo it will now generate the right HTML it builds whole website for yeah so you know Devon yeah so we are like a more General version of uh the infrastructure behind Devon so way AI Engineers are always going to write their own tools because they have a Hummer and they see nails they're going to always code but a gentle system like Devon have a huge opportunity to help a lot more people we're giving the people who are writing these kind of tools um a place that they can execute it and they can uh let's see this website so see it works all right Doo P get started why choose Doo our services very simple you would have to do a lot more prompting to make it work of course shows the possibility the beginning right it's going to be better over time so pretty amazing I really like the aspect of like combining different things because you're opening yourself to to the com right people can provide yeah and then we can build on top of that it's a big company to build uh big company to build because right now people think about AI as retrieval of knowledge either from documents or from the weights of the network itself from the training data set I think that's going to change in the the coming years to actually think about it as a reasoning engine and these reason reasoning engines can be connected in many different ways so for example I when I do brainstorming I use three different models that argue between themselves and they have a different perspective on each one of these these is it maybe last question is it possible to use it now it's possible to use it now and I've actually enabled anyone who is uh anyone who is uh this this might go wrong but anyone who has an IP address from Poland gets unlimited free credits for now ah wow so just use VPN and register damn it you guys going to ruin my company you know uh yeah but enjoy it and uh I we are super happy to hear any feedback uh we it's tough tool to like at the beginning it we do say it's a programming language hence people expect to need to learn a little bit more about this and uh yeah it looks pretty impressive thank you for showing us around andk you invite everyone to Reg yeah come join us

Michal and I decided to go to the Unstructured Data meetup because it sounded interesting, and it was at GitHub HQ, which we had never visited. While we were there, we accidentally ran into Kamil from Poland, and we quickly connected. To our surprise, Kamil and Michal had attended the same high school in Poland, one of the best in the country. As we talked with Kamil, we discovered he was working at Wordware, a company that builds an IDE for programming using natural language. We asked Kamil to share more about Wordware on camera, but he suggested we meet one of the co-founders instead. The following day, we met Filip, the CEO of Wordware. Filip was super friendly and energetic. We went for lunch together, enjoying Japanese-style curry outside of Shack15. We had great conversations about our travels, the differences between startups in Poland and the US, and life in general. Filip also gave me some cool insights about the O1 visa procedure. We then decided to record a short interview with Filip on the spot. Finding a quiet place at Shack15 was challenging, but we made it work. Despite my initial concerns about the setup, the interview turned out great, and here is the result. A few weeks after our meeting, Wordware announced that they had been accepted into Y Combinator. Congrats!