Quadratic: The AI Spreadsheet

Hey everyone, welcome to the Latent Space Podcast. This is Alessio, partner and CTO at Decibel, and I'm joined by my co-host Swyx, founder of Smol AI.

Hello, hello. Uh, this is another lightning pod. We're in the remote studio with David Kircos over in Denver. Is that right?

We're based in Boulder, yeah.

Boulder. I'm sorry. They are, you know, close enough.

Close enough.

Close enough. So you're the founder, CEO of Quadratic. I, you know, have been, uh, lucky enough to see you for a little bit as an angel in the journey. Uh, it seems like you've really blown up. Like I, I invested before you were an AI spreadsheet, but now, now e- every- everyone is AI.

But I had, I had a really good experience recently because I had a lot of data munging to do. You know, we have three thousand people at our events, and everyone needs like some kind of data on our attendees.

So I just threw it in Quadratic and then I, you know, typed some English and it did some data analysis and produced some charts for me, and, uh, that was really magical to see. Well, welcome. Is, is that a good characterization, AI spreadsheet, or how do you, how do you talk about Quadratic?

Yeah, that's exactly how we characterize it today, and when you first joined as an angel investor, we were much more focused on building a technical spreadsheet. We were focused on building the next generation spreadsheet, so it speaks Python-

Yeah. That's all I needed.

Yeah. Speaks database connections, speak Python, speaks those languages natively, and so our early adopters were all more technical, and now with AI, that just bridges the gap. Everyone's able to write that code and do that more tech- technical analysis.

Yeah, I find like the people who had done the work pre-LLMs are suddenly enabled to do more. So I think this is a very similar story to StackBlitz and Bolt. StackBlitz worked on WebAssembly containers for four years and got...

was ... I mean, they were making some progress, but like they, they were nothing as compared to Bolt, which, which did the unlock. But like right now they can capitalize and just work on the agents because they already did the web container stuff, and it's kinda ...

I think it's similar to you, like you can work on AI spreadsheet because you already spent the last four years doing the WebAssembly stuff.

Yeah, we were focused for three years on building the high performance spreadsheet that speaks code natively as a next generation spreadsheet experience, and one of, one of our biggest challenges was educating new users on how to use the product.

And AI completely makes that go away. I'm sure it's a similar thing for Bolt and StackBlitz, where, you know, educating users on how to best use a technical product is difficult, and when you can have users sign up and just prompt it, just drop some data in and say, "Help me analyze this, help me visualize it," the learning curve went to zero and the adoption increased tremendously.

Yeah. Um, so maybe perhaps we could just kinda pull up a, a screen share just so people can see what we're talking about. I guess it's a very visual thing. While, while you're talking, so I, I think like the other thing I, I really like was in the original pitch was like it's kind of like a Figma for spreadsheets where you can kinda zoom in and out really nicely and it, it can support, you know, I don't know how many rows you support, but millions of rows.

Yeah. It's nice.

Yeah, we sup- we support millions of rows. Uh, of course, we wanna bring the whole team together around data analysis. That's why we chose the spreadsheet interface. Everyone knows how to use spreadsheets, and so when users first come into our tool, it looks really familiar, right?

You, you could see Quadratic here. It just looks like a spreadsheet. We want people to think, "Hey, I already know how to use this. I, I know how to get started." And then we have the left panel on the side, the AI chat, that can help you orchestrate and do everything in the sheet.

So how do you usually demo?

Yeah, I, I think, you know, everyone who starts with the spreadsheet is probably bringing in a data set, so I can drag and drop in a CSV.

Nice.

This is a small feature, but you can drop a CSV kind wherever you want on the sheet. This was always something I felt was more annoying than it needed to be in other spreadsheet tools, and then we could start with a prompt.

So help me visualize my revenue over time.

By the way, I'll, I'll say it, like you can't do this in Google Sheets. It's-

Yeah

... stupid. You can't even freaking drag and drop a CSV into Google Sheets. What the hell?

When we started, you couldn't even smoothly scroll in Google Sheets. It was quantized to the nearest cell for scrolling. You know, all these like minor product interactions were just really bad with Google Sheets. In Quadratic, you can, you could pinch and zoom, so if you have a track pad you can zoom out and you can look at very large data sets.

Uh, you can zoom out very far.

I don't know why you would want to but yes, yes.

Yeah. It's nice to be able to zoom in and out and focus on whatever you're working on. You don't need to be able to zoom out and see, you know, a million rows. That doesn't make any sense.

Yeah, so you, you visualize, uh, revenue here for people who are listening on audio. I mean, so this is what I call vibe charting. You know, you, you, you look at it, you type in some words and you're like, "Just get me something."

You don't even really look at the columns and you can make a chart. You can do some edits. Like, uh, something I always do as a first transformation is like give me the Y-axis in log form instead of, instead of absolute numbers and you can do that.

Let's do it.

For me, the actually ... The, the, the ... Yeah. Go, go, go ahead with that one. But like, uh, for me, uh, you know, the, the, the, the thing that really sold me was I had a whole bunch of categories of different talks.

I was, I was analyzing my CFPs, and for me, like learning Matplotlib and doing like colors and positioning of the labels and axes and legends and stuff, I just never did it because obviously it's just ... Like, you can do everything, it's just like it's hard to, it's hard to do.

But like because I can just ask for it now, it, I, I'm unlocked with like charting ability so much easier.

Yeah, absolutely. And everything that Quadratic is building in terms of adding analysis to the sheet is it's building a method of analysis. So if I go in and change some number that was part of my original data set, everything that it built is gonna update.

It's not giving you a static image output or anything like that. These are interactive charts that are based on the input data. So as that data changes, you can see how it reflects in your tool. And so a lot of our users are using Quadratic to model stuff, to build models to better understand data.

Drop a CSV in- ChatGPT and ask it for a chart, those aren't responsive, those aren't interactive, and those I don't think ultimately help people understand data better.

And, uh, from a product perspective, should people still think about it as traditional spreadsheets, so I could do traditional formulas in there as well? Or, like, is it just kinda like a spreadsheet-like UX on top of a Python runtime and, uh, AI?

Yeah. Some of the core interaction methods in Quadratic. So if you press equals, you can use formulas. Uh, we have similar formula set to Excel and Google Sheets. And then if you press slash, you can go in and code in Python.

So that by default will open it in a code editor in the right panel, and I can, for example, output some data to the sheet. So this is, I'm just counting to nine. Um, and that will return natively as a table in the sheet.

So you, you can mix and match formulas, Python, and even database queries. So I c- I can go in here as well. I'm connected to, uh, a database, and we can browse the database schema so we can see the tables that are in there, and we can also chat with it, so I can say, "Show me teams that signed up in the last hour."

And it, because it knows the context of the schema, it knows all the field names and table names to get that type of query right, and it will populate it on the sheet.

And, uh, all the Python is running in a Wasm container inside the browser? Or what's the... How does the magic happen?

Yeah. Everything in Quadratic is running in WebAssembly in the browser, and then the whole sheet is drawn in WebGL. We did this so you get these really fast response times when you're writing code. When I write, wrote that Python code and put it on the sheet, it happened in a tiny fraction of a second.

This database query took one second, but a lot of that time was on the database. Actually drawing it and putting it on the sheet was, was very quick. I, I, and I think it's, uh, Python and JavaScript. Does, does the JavaScript also run in WebAssembly, or does it execute in a normal JavaScript sandbox?

Yeah, we, we... Well, we run JavaScript in a web worker. Okay. So in an isolated environment, yeah. Yeah. If, if you were to open up a JavaScript cell and- So much easier. Honestly, yeah, like so many people run Python in spreadsheets.

They should just like have like Python web workers and let, you know, let like Chrom- if Chrome did that, that actually would cut a lot of work for people. Yeah. And the way that we think about all these things is like a native, natively putting them in the sheet, right?

Like, you can, you could run some JavaScript in Google Apps Script, but it's a separate thing. It's kind of running- Mm ... off to the side. It's not a part of your sheet. In Quadratic, I could go and run Python, and I could fetch data from an external API, and that just becomes a native part of my sheet that I can then go and, say, build a formula on top of or continue to do additional analysis with.

And- Yeah ... whenever I want fresh data, I can just rerun that API call. It's just a part of the sheet, right? Yeah. Awesome. I mean, that, that's, the- there's just like a ton to go in on, on spreadsheets.

Like, you know, I always say the, the big three categories of business apps, it's Docs, it's Slides, and Sheets, uh, which obviously maps to Microsoft Word, Microsoft PowerPoint, Microsoft Excel. And like Sheets is the hardest one, right? Like, I actually, like the, the reason I, I did nine rounds in my Google interviews, the last interview that I did was build a spreadsheet, and, uh, it was, it was very, very tough.

Yeah. I, I don't know if I ever told you that story, but... No, you didn't. Yeah, building a- Building a spreadsheet, um, is, is very tough too because people expect it to have all the features of Excel, right?

And that- Yeah ... that's a really... And all the bugs of Excel in some cases. Excel has all these bugs that people rely on, date bugs and formula bugs that are now features. They're now just the way that that works.

And when we set out- Yeah. What's- ... to build this spreadsheet, we, we set out to create something a little bit different. So we wanted to make it as familiar as possible, but not fully compatible with Excel. You can't export a Quadratic sheet back into Excel.

You know, we have code and we have database queries that live in our sheet. You can always pull your data out, of course. You could select any area on the sheet and get a CSV or get another file, but you can't just go back into Excel 'cause it, it doesn't make sense.

Yeah. What's your date policy? If I, if I type, you know, zero in a date field, does it give me, you know, J- January 1st, 1970 or something? I think one of our engineers wrote a blog post recently on the, uh, difficulties of date parsing and in spreadsheets, or like the cursed nature of date parsing in spreadsheets.

Uh, cool. Okay. So like, I, you know, I think that's the spreadsheet component. It's, I, I think it's very underrated the amount of data connectors you have, like it connects to Postgres. What else? Like Snowflake, RDS. Yeah. Today we support MySQL, Microsoft Server, Snowflake, Postgres.

We're adding- Yeah ... a bunch more in the next couple weeks. We're adding BigQuery, um, and we're adding easier support for connecting Neon and Superbase as well. Yeah. So like i- in the data engineer world that I used to be in, this is the ultimate BI tool, right?

Like you connect to your database, and you do an analytics on it in a spreadsheet. And like there, I think there have been a few attempts at this, but like obviously nothing, nothing this, uh, performant and flexible as like a full spreadsheet.

Yeah. And it's been amazing for our team to use Quadratic for all of our product analytics, so we pull our data around-

Wait, are you showing your real data? Is this okay, is this okay to show?

I can show our real data, yeah.

Okay.

Uh, we, we, we, we pull our data from a bun- from like Mixpanel, from Postgres, and from other sources, and we, then we th- bring it all together into, into dashboards that we can all work on. And so it's amazing to see people on our team who are less technical work on these dashboards alongside our, our most technical developers, um, and collaborate on analysis, which is something that I think is just missing in other BI tools.

They're either really technical and then non-technical people just press Export to CSV, or you're using a really non-technical tool and then the technical folks don't wanna engage with it.

Let's talk maybe a bit about AI and what the interface between the spreadsheet and AI is. I think everybody understands the one-to-one nature of a, this is a data source, make a chart. How do people interact with like the chat interface from a whole spreadsheet level?

Like how do you think about creating context? How do people maybe can direct the AI to certain pieces? It's almost like you can, you know, zoom in and out, you can put all these things, but then it maybe becomes a curse at some point.

I think one of the core challenges with building an AI spreadsheet is giving the right context to the model and giving the model the right tools to take actions based on what the user's asking. And so we've come up with a pretty clever way of giving the AI a summary of the entire sheet and then the ability to access additional details as it needs it.

So it can see all the different tables that are... And all the different code cells that are anywhere in your document and knows where they're located. It can ask for more information if there's not enough included in the first piece of context, and can then use a bunch of different tool calls to do actions in the sheet.

Our goal is to make it be able to see and do anything a user can do and way beyond. Obviously, it can see more than just what's in, you know, a little frame on the screen. I mean, it can take actions that don't really make sense as user actions, like I want to highlight all cells that start with D or something like that.

That makes sense as a computer to go do, but a- as a user, there's no button for that, right? We wanna give it all the actions a user can do, 'cause users ask it, "Hey, format, you know, that in bold," or something, so it needs to be able to do something that a user would ask like that.

But people also ask for things that they don't know how to do themselves, and the AI needs the tools to be able to a- to do those on its own.

Awesome. Uh, what are-- what was your journey in sort of implementing the agent? Like, uh, you know, like, obviously, I think when the ChatGPT moment happened, like everyone was like, "Okay, how do I, you know, put this into my product?"

Where did you start? Where did you evolve? Did you, did you run into any wrong forks in the road that people can learn from?

Yeah. W-we started as a code assistant agent, so we started with that little chat in the side that helps you rewrite database queries and rewrite code cells, and we found ourselves using that for all of our coding in the tool, and it felt like a very natural progression to bring it up another level and allow it to do a lot more things in the sheet.

So it was, it was gradual. It was really a pull from us and our own users, where we're like, "Man, this is the primary way we're interacting with our data already. We need to make this do everything."

Is it a simple sort of agent loop? Are there any guardrails? Is there any fine-tuning going on? Just any-anything around like the technical implementation.

Yeah. Um, can talk a little bit about how it works. Every-- it, it knows the context of your sheet. When it makes changes, it looks at the sheet after every change.

Mm-hmm.

So if it generates an error or something, or something doesn't look right, it can react to that, and the agent loop will continue until it thinks that it's answered the user's question. We don't want it to go too far.

We want the user to be able to sort of understand all the changes that it's making in a single block, and so it doesn't go on forever, or it shouldn't go on forever. We are working on fine-tuning a model.

We think we can get the costs of inference way down and the quality and specific... You know, AI working in a spreadsheet is not really what these models were trained to do, right? It, uh, there's a lot of 2D positioning, code errors, these, like, multi-dimensional overlaps of data on a spreadsheet, and the LLMs weren't really trained to do that from their core.

So by adding examples of what good spreadsheets look like and how to do the analysis in a way that's helpful to the user does help it give better answers.

Yeah. Um, I, I think, like, one of the interesting things that I see in is, like, sometimes supplying too much information, because you, you could go the, the distance of, like, dumping the entire spreadsheet row, row by row, cell by cell into the context, and it might get distracted.

So I think there needs to be something smart about, like, effectively just doing the schema of each table. And I-- but I, I've, I've also seen it basically kind of just, like, peek at the data and understand the schema or understand what it is before it executes.

I, I, I don't know. Like, it's, it's interesting, like that, that data problem.

Yeah. We tried to give it the whole sheet as the context initially, and it works really great when you're working on a really small sheet. And then if you d- you know, drag and drop 100,000 rows of data in, first of all, it's gonna be really expensive.

Your request is gonna cost multiple dollars, and it's a really bad response. It, it, it, you know, it's too much in context. Um, and so we've, we've done a lot of work to summarize what's in the sheet, what data you can find where, and then we put that in context by default, and then it has a tool so it can continue to read more if it wants to understand more about the data, which is probably the tool call that you saw when it was like, "Hey, I wanna understand the column names," and maybe it's-

That's very smart.

Yeah. This da- this column, this data doesn't just include numbers, but there's some values that are a dash, so I have to handle that in my code or formula that I'm writing. It'll only peek at those things when it thinks it needs them.

And so basically, each of the subsets, it's almost like its own object, and then you can, like, semantically describe them in the context, and then the AI can pick from there. And this is all done without... You're not spending any time on vision, right?

You're just kinda like, through the API, you're able to describe everything, um, and pass it back.

Yeah. And we, we do a little bit of vision. When users are creating charts, we pass back an image of what the chart looks like to the AI, 'cause a lot of times the user will describe visually the change that they want to the chart.

They'll say something like, "This doesn't look right," or, "Make this prettier," and it has-- it, for it to have, be able to answer that well and give the user what it wants, it has to see what the user's seeing.

The, the thought just occurred to me. Have you seen the Studio Ghibli version for charts?

No.

So there's, you know, a lot of people, when image gen came out for GPT-4.0, they started Ghibli-fying everything. But you can also Ghibli-fy charts, and I think actually it's, like, very underrated for creating infographics. To me, that's like the, as a content creator, that's like the goldmine, that I can deliver good infographics, but, like, it, it looks distinctive, and it's not, like, a default Google Sheets chart thing that, that doesn't look anything interesting.

So I would actually... My feature request is I would, I would, I would put, like, a little, like, style this the way that my company styles it, and here's a demo. Let's say I work at Studio Ghibli. I would style this as a Studio Ghibli chart.

But, like, really, like, there's, I think there's, there's some alpha there for image generation. But I, I hear what you're saying in terms-

Yeah.

Yeah. Go ahead.

And I've, I've done that a bunch of times, where I've taken charts in Quadratic where it kind of looks like more like a standard chart, you know. It does- it doesn't have a ton of styling with it, and I've taken it and dropped it into, you know, ChatGPT or something and said, "Hey, make this look better for a presentation.

Put it in this style."

Oh, you've done that. Okay.

"Remove these labels." Yeah. And but I keep the, you know, the chart in Quadratic 'cause I can update it really easily, right? Once you've turned something into-

Yeah

... a static image, you can't update it. So I, I would just take the image and, like, put it in the model-

Yeah

... to do that.

Totally, totally. Yeah. You know, I think the other thing that I would say is interesting, so I don't know, have you seen the, the new Canva, uh, data grid thing that they did? So Canva launched the spreadsheet, which I was like, "Oh, that's, that's kinda, that's kinda interesting."

But they don't obviously execute code, and you can basically build the Canva features inside of Quadratic because each cell can run code, and you can sort of drag and drop or map over a set of data to then call an API to then generate images.

So you could actually kind of build that entire workflow w- within Quadratic.

Yeah, I, I, I think that Canva's spreadsheet and tools like Airtable and Smartsheet, they're really geared towards data storage and process management. Quadratic is really focused on analytics. Um, and so our, our spreadsheet's really geared towards helping you transform and visualize and understand data.

And yeah, you, you can augment cells with data using AI, AP- AI ca- or API calls and AI as well, and so I think that, that they're maybe more geared towards that type of thing, where they're augmenting sales data or augmenting marketing data.

Um, do you see the spreadsheet as being more of a proactive UX that the AI uses to communicate to you? I, I think under, you know, I, I think the use case of, like, I ask AI to do something and the spreadsheet is clear.

I think there's maybe the future of, like, hey, the, uh, spreadsheet is proactively doing some analysis for you because maybe some data looks off or, you know, whatever. It's, like, the right time of the year. You see Quadratic, Quadratic almost as, like, a tool that even other agents might build on in, in the future.

Like, I know Anthropic, for example, did the example of, like, a financial research agent today, uh, at the update. You could see, uh, that agent then using Quadratic to, like, put some of the data and model in a spreadsheet so that you could, um, use it.

Uh, I'm curious if you've given it any thought.

Yeah. I think that the spreadsheet is the best interface for a human working with a computer to understand data. So if a user goes into an AI model and asks, "Help me understand what home I can afford," or something like that, the user's gonna wanna change different variables.

What interest rate do I have? You know, what's my monthly payment look like? They wanna play with and change a bunch of different variables and see how it affects the output and trust that the calculations were done in a way where it's correct.

I think that the best, most comfortable interface for working with that type of thing for the vast majority of people is a spreadsheet, and so I, I do see these agents generating spreadsheets to answer people's questions, particularly when they focus around the user understanding data.

Of course, it can give you a number, right? But does that really help you understand better? And do you trust it?

Yeah. Yeah, I, I think that auditability is, uh, is important. Any other use cases that we haven't talked about that typically is highlighting? We haven't talked about multiplayer, uh, and that was a big sprint.

Yeah.

Yeah.

Quadratic has real-time multiplayer. A lot of our users are bringing in their, their data and chatting with it. That's the primary use case. They, they wanna do analysis fast, probably very similar to your use case, where they just don't wanna mess around with formulas and buttons.

They just wanna get to the result really quickly.

Yeah. But like, just, like, generally, you know, what, what are, what are you seeing? What's going viral in terms of, like, you know, I think you're, you're exploring different ways to explain the product to maybe a less technical audience because of the AI push, right?

Yeah. We see a lot of, um, like, business owners and business managers who work at small companies with just a few people who have bring in data. I think they often don't have time to mess around in Excel and they see AI as, like, "Oh, AI's gonna, you know, help me do this really quickly."

I think one thing that occurred to me that I think, I don't think I see in Quadratic yet, but I'm curious on your thoughts, is kind of, let's call it form controls. Does that word mean anything to you?

Effectively buttons and, like, little, little form elements that you can kinda turn the spreadsheet into more of an app, where it doesn't reactively update every time a number has changed. You're, you're kinda changing the para- the data flow paradigm from, like, numbers have changed, you recalculate everything.

That's a spreadsheet, right? To more of a, like a, you know, like, we'll enter in some things, but, like, it, it's more of, like, a user input environment that then starts the recalculations whenever you're done.

Yeah. I, I think it makes a lot of sense to make it clearer what inputs you can change, like it... Or are, are some of those inputs not just boxes with text in them? Could you have a slider or some other types of inputs that then-

There you go

... can recalculate models and, and affect them? Absolutely, that makes sense.

Sliders. Yeah, yeah. I mean, the, you know, basically anything in the HTML form spec is, it's-

Yeah

... fair game, you know?

Yeah.

It's almost kind of like the roadmap is, like, done for you in terms of, like, what you can do. Um, I, so I happen to know that your team is pretty small. It's, like, six, seven people?

We're, we're nine people now.

Nine.

We're a little bigger, but yeah.

So how do you do this with nine people? This, this is, like, a very broad, like, it's a remote team, nine people, you know, uh, you hired my friend Jim Nielsen, who I have to shout out as, like, one of the best reactive operators that I know.

Uh, and he's over in Japan. Like, how do you run a remote, you know, company that's what I've been calling also a tiny team. Like, it, it, it's you're doing a lot with very little.

Yeah. Jim Wa- Jim Wa's our first employee. Um, and we've, we've been laser focused on hiring people who are obsessed about solving the problem that they're working on, who are self-motivated to solve a difficult problem over a long period of time.

I think that's really important for a remote team. About half of us are here in Colorado, and the other half are remote, so.

Okay.

Hiring, and a lot of our engineers are remote, and they're all very self-driven. Like, they, they wake up just thinking about the next problem within the spreadsheet- ... and, like, obsessed with solving it, and I, I think that that's critical for running a remote team, building something really difficult.

Yeah. How, how important is it to you to be, like, you know, you're not 100% open source, right? Uh, you're source available, but, like, do, does a lot of feature requests come from there? Do, like, does the, does the community contribute a lot?

Do they need a lot of grooming?

We found a lot of our developers through being open source, and I think in that way that's been incredibly valuable to us. And it, it also goes a long way in building trust with customers who care a lot about trust, bigger corporations, companies who are working with more sensitive data.

You can self-host Quadratic. We're source available on GitHub, and building trust to put your data in a tool is difficult, and I think us being open in that way really helps.

Awesome. Any

Kind of like future of data analysis and AI things, you know. I, I feel like this is almost like, you know, fitting the, the old UX and putting AI on top of it. Any, any hot takes?

I, I continue to believe that the spreadsheet interface is the best interface for humans understanding data working with computers. And so I think what will change is a lot of the buttons will go away, and the way that you interact with it is gonna become much easier and simpler.

Uh, you can say, "Make my document beautiful." You don't have to click a million buttons to, to make that happen. But users will still wanna be able to click on a number and understand where it came from to get that confidence in it that it's, that it is the, the number that they expect it to be.

Um, so I, I see our agents getting more proactive as you add data and as you start to do things in the sheet, continuing that work and making it a lot faster. But we're really, we're really building this tool that works with you to understand data.

Like, we, we want the human in the loop at every stage. We don't wanna set it off and then come back a lot, you know, a day later and have something done. That's like receiving a spreadsheet from a coworker or something.

It's like, "Okay, now I have to try to understand this all over again." I think the process of building a method of analysis alongside the, the tool is really what helps people understand data, and I think that's why they turn to spreadsheets in the first place, so they can get an understanding of the data, so they can make a better decision.

Yeah. Do you think that will be mostly Python once the model does most of the analysis? Or what's kinda like the future of formulas, I guess?

I think formulas will be used less in favor of Python, yes. Um, if you're summing a column, a formula is great. If you're trying to aggregate data of like how many users signed up last week and did this thing, formulas are a mess, and they're super...

If you had a complex query, imagine reading the formula that the AI generated. It does not help you understand what it's doing. Um, and so I, I think a lot of the analysis will go to Python over time.

Uh, so as the resident JavaScript guy, why not, why not JavaScript? What's your... Actually, do you have a split between Python and JavaScript right now? Is it like-

Uh, we, we-

... seventy-five, twenty-five?

I don't know the split right now. Um, it used-

Okay

... to be like ninety/ten.

Yeah.

Um, but I haven't checked in a couple of months. Python's much more used in the data science world, right?

Of course.

There's a lot more tooling around Python to work with data.

I mean, Pandas is amazing.

There's actually-

Yeah.

Yeah. We, we support Pandas natively and, um, that is a h- that is a big part of why it works so well. JavaScript is pretty immature in terms of working with data. It doesn't have a lot of packages or tooling around transforming and working with data.

There's quite a good, a lot of good visualization libraries though, which we support. If you wanna create charts using JavaScript libraries, you can do that at Quadratic.

Yeah. One of the things I, I, I'm interested in exploring is like data analysis in Python, but visualization and UI in JavaScript, right? So it's kind of like a mullet or a chimera of languages . And you can support that-

Yeah

... 'cause the different cells. Like data is like the universal interface that you pass between languages, and you can do each stage in whatever language you choose.

Yeah. You could query data in a database, transform it with Python, and visualize it using JavaScript libraries in three separate cells right next to each other.

Yeah. Awesome. Cool. Uh, I think that's all that we prepped in, in terms of just talking about it. Uh, any other things you wanna mention or call to action?

Yeah. Sign up. Go to Quadratic AI and try it out if you wanna use AI to understand data. Quadratic is free, and we charge after a certain amount of AI usage. Um, and so yeah, if you wanna use AI to un- better understand your data, check out Quadratic AI.

Thank you guys for having me on.

Thanks for joining us, David.

Thanks so much.