Get in touch

Postlight joins Launch by NTT DATA! Learn more.

Recently Postlight’s Head of Engineering, Aaron Ortbals, had an eye-opening experience with ChatGPT, an AI tool that can write code for you. This week Chris and Aaron chat about new AI tools being developed. They discuss the benefits, concerns, and future of these tools and how we can use them to move more quickly without replacing people in the process. 


Aaron Ortbals: But then I was like, you know, “Write a song about my pitbull, my dog.” And it wrote a pretty good song. (Laughs) And it was like, what is happening? It’s just these mind-blowing things.


Chris LoSacco: Welcome to the Postlight podcast. I am Chris LoSacco, the president of Postlight, and today on the show I am joined by Aaron Ortbals, our head of engineering at Postlight. Hey Aaron, how’s it going?

Aaron: Good. How are you, Chris?

Chris: I’m great. I am excited to talk to you today, because we’re going to talk about artificial intelligence.

Aaron: I know, it’s going to be exciting. I think there’s a ton to talk about and I sort of intro’d the topic and you were like “Let’s do it,” and so, yeah. I mean, we could talk about this for hours, but I think we’ll hit some of the highlights.

Chris: There are a lot of hot takes out there on the internet about (laughs) what’s going on right now in AI, and how it’s going to irrevocably change the world, which may or may not be true. But I… we had a conversation about it, because you told me this crazy story about how you used one of the open AI tools that’s out there, ChatGPT, to write a piece of code. I would love to just have you tell that story, and let that be the sort of inroad here to the conversation that we’re about to have, ‘cause it’s fascinating.

Aaron: Yeah. So, you know, when we were talking the story I told was… a lot of us have Apple watches these days. And there’s this incredible amount of health data being stored locally on our phones through HealthKit, so the iOS ecosystem. We’ve got all this data. You can go look at your heart rate over time, or your, you know, heart rate variability, which kinda tells you about training, and are you ready to work out or not. And…

Chris: Yep, yep.

Aaron: I was talking with my wife, and essentially, she’s an Excel nerd, and she wanted to look at the data and graph it. And so we tried to export it, which luckily Apple lets you do, and what it does is it dumps out a giant XML file. Which, for those that don’t know what XML is, it’s sort of like HTML for data, you know, broadly speaking.

Chris: Yep.

Aaron: But, you know, what do you do with a 2 gigabyte XML file? And it’s not as easy as a CSV to import into Excel. I thought, there’s gotta be a tool out there to do this. And so I go on GitHub, and I’m looking around, and I found a parser that would spit out a SQLite database.

Chris: Perfect.

Aaron: So, SQLite is a local relational database, and you can easily explore database, so it’s actually a really nice, natural thing for somebody that’s good at excel but isn’t a programmer to tinker with. There’s a tool called Dataset where you can, you know, point it at the SQLite database and use a browser to click around and look at the data and run a query. So it’s really cool. And I found a parser that would take the XML data and parse it into a database. And so I was excited, I pull it down, it’s written in Rust. So, Rust is a programming language that’s in the same realm of C++. So it’s a low-level systems language that I don’t typically write much code in, or… we have done some work at Postlight with Rust, but it’s less common for our day-to-day work. But I’ve tinkered with it, but I haven’t really written code in it before. And I ran the importer, I built it and ran it and it failed. And the reason that it failed was an issue with a column in the SQLite database, and I thought “Hey, I’m smart, I can fix this…”

Chris: (Laughs)

Aaron: …even though I don’t know rust that well.

Chris: Okay.

Aaron: And I started to look at it and I knew how I would fix it in other languages, and I thought “Let me try to do this.” And Rust is a pretty complex language as far as types and how it deals with data, and it has this thing called a borrow checker for memory, and it’s a very, very interesting and innovative language, but it’s not as easy to just jump in. And so I hop on ChatGPT. So, for those that don’t know what ChatGPT is, OpenAI has a Large Language Model, so an LLM, GPT 3.5, so they’ve evolved this over time. And what it does is, it takes a ton of language data. Millions and millions of words. Essentially, all of the internet has been fed into this language model. And it can understand what’s called a prompt, you can give it a question and it will answer it. And so, what blew up recently was, OpenAI released ChatGPT, which is just a chat interface to GPT3.5…

Chris: This model. Yeah.

Aaron: …yeah. With, yeah, with some special sauce sprinkled on top of it. And it’s pretty amazing to tinker with, it’s a little scary in that it can answer you like a human would, and it can both be very close to correct, 99% correct, or it can be totally wrong and be as confident that it was right, and it’s just wrong.

Chris: (Laughs) Right.

Aaron: But it’s a very cool tool for programmers to tinker with, because it can write code for you. So I described my problem, which was essentially filtering an array. And I described it in plain language, and it wrote almost a perfect implementation of it right onscreen in front of me.

Chris: Hold… hold on. I want to back up.

Aaron: Okay.

Chris: I want to unpack all the different things that ChatGPT can do. Because there are some obvious things that you might think to test an artificial intelligence, and it does some interesting things there, but then you can also do some very… you know, like, advanced things that you wouldn’t think it would get right, and it does. So that’s kinda wild. But what you’re saying, is you gave it a prompt in plain English that said, like, “Write this importer in Rust”? Or how did you… how did you kick this off? Like, what did you… how did you describe it?

Aaron: So, I think one of the interesting things about these tools is, you do have to be technical to use it to solve a technical problem. So you do have to be able to frame it or narrate… narrate it in the right way.

Chris: To get the right output.

Aaron: So, use the right words. Essentially, the way I would describe it too is, Google’s a powerful tool, but if you don’t know the right term to Google for you’re gonna struggle. And I think there’s some similarities to, like, a ChatGPT-style interface, where if you don’t know… actually, a good example was, the other day I was trying to find essentially a HomeKit button that would allow me to run a HomeKit scene in my house to turn lights on. And I didn’t want a button that I could slap on the wall, I wanted a button that would go where a light switch would go. And it turns out those are called “scene controllers,” and I didn’t know to Google for a scene controller or search Amazon for it. And I think that’s a good example of, until you know the right terms…

Chris: What you’re looking for. Yeah, yeah, yeah.

Aaron: ChatGPT can’t help you. And I think that’s also similar to some of this… what’s been dubbed “prompt engineering” for ChatGPT or Dal-E 2, the image tool where you can describe art or whatever, a picture, and it will generate an image. Getting the terms right can really make or break the result.But anyway, to come back from my tangent, I described filtering essentially an array of strings, like text, and doing a couple complicated operations on it, which is what would fix my bug that I was trying to fix.

Chris: Right.

Aaron: And it implemented. It explained it, line by line. You know, you need to implement this type and then you can use it this way, and this is an example of how to use it. And then I asked it to write a test, to test it, and then…

Chris: (Impressed laughter)

Aaron: It has the state of the conversation. And so, I said “Write a test for the thing you wrote,” and it wrote the test.

Chris: Wow. 

Aaron: And ironically, it did have a small bug in it of course. Which I think, again, this is not gonna replace people’s jobs, and I think that’s a separate…

Chris: Well, we need to talk about that. (Laughs)

Aaron: You know, there’s gonna be lots of validation.

Chris: Yeah.

Aaron: And I, of course, had to try to understand this to use it. But I tinkered with it and it didn’t work at first, and I made a small change, and I added it to the code base, and my importer worked. And I had my health data in the SQLite database. And so I was able to solve the problem. And I went to one of our engineers who’s an expert in Rust, and I said “Look at this implementation. A chatbot wrote it and I modified it, is this how you would do it?” And he said “Yeah, it’s near perfect, that’s a great… you know, you could do it other ways, but that was pretty good.” And so, it was kinda jaw-dropping to see that. And I’d done a… you know, experimented with it in other ways. You could see the benefits of using a tool like this at your disposal. It’s also a bit odd and scary and kind of new. And I think OpenAI, the reason ChatGPT has exploded and sort of made the news is that they found an interface to the model that was a good way to experiment…

Chris: Yeah.

Aaron: …and so it’s capturing people’s minds like mine.

Chris: Right. So, just to finish the thread on this programming example, I mean, this saved you hours of time?

Aaron: Probably not hours, but it may have saved me, I don’t know, 30 minutes.

Chris: But the point you made before is, you don’t typically write Rust, right? This is not your toolset of choice. But you understood where the problem was and how to fix it, enough to be able to describe it and write this prompt. And the gamechanger for me, in hearing this story, is you didn’t actually have to solve the problem, you just had to know enough to frame the problem. And then the bot took you the rest of the way there. And that is… that is fascinating and terrifying. If you extrapolate this into the future, do you think the nature of programming is going to change? Almost in a fundamental way, where it becomes more about reasoning and framing problems than it does about writing the code to solve the problems, you know what I mean?

Aaron: Yeah. I mean, I totally agree, and I think there’s a lot of… I think there’s a fear that this will… you know, could replace jobs. I think there’s an example right, with… let’s say an icon designer. An icon designer may produce these incredible icons for, let’s say, an iPhone. And somebody that would have hired a designer to create an icon will just go to DALL-E and generate an icon. And I think that it sounds scary, but the reality is, the creative folks that can utilize the tools to generate different versions or inspiration for, essentially, a human to use it as a headstart, I think that’s going to be a lot of what the practical usefulness is of these tools, is that programmers are going to need to be more disciplined, in that they need to not just take the output of a tool like this and just throw it into a code base and then ship it to production and then break stuff.

Chris: Right.

Aaron: You know, there’s still going to be, folks are going to have to focus in on…. yeah, framing the problem and using the tools to help them accelerate. And essentially move quicker. But they’re not going to… it’s not going to replace the big picture, in that folks that are really good at framing problems to your point, or like… it’s going to shift away from pure keystrokes to taking some of this output and putting it into a code base and looking at an overall application and how it works, and validating that it still works. Those kinds of things.

Chris: What about modernizing a legacy code base, right? You’ve got something that was written in COBOL decades ago, and you wanna rewrite it in Java, or you want to rewrite it in .NET. And, you know, typically what you would do, depending on the size and complexity, you would hire a team. Right? You’d go to, maybe, a consulting firm like us, or you’d go and hire a group of people and say “You’re going to spend the next 12 months delving into this code base, understanding all the weird edge cases and cleaning it up and writing it in a much more maintainable, modern system.” With something like ChatGPT, or whatever the evolution, the future point releases of it are, does that work become automated? Like, could you essentially tell an AI tool, “Please go rewrite this code base and make sure that it passes all the same tests?”

Aaron: Yeah. No, I think there’s an interesting thread there. It’s interesting on, like, an infinite time horizon, it seems possible.

Chris: Yeah.

Aaron: But I think there’s an interesting part about this, which is like, the correctness level of the code that it outputs. Where, yeah. I think there’s going to be certain types of professions or use cases where you have, let’s say, a self-driving car, where it’s safety critical.

Chris: Mm.

Aaron: Where it needs to be 99.99% correct, and if it’s not, you know, it’s a safety issue, and…

Chris: Well it’s literally life and death, you know, in those cases, yeah.

Aaron: Yeah. But then there’s other examples where, you know, mostly correct. You know, if you’re 90%, and then with a little bit of human validation you’re all the way there. I think, for example, like… lawyers are an interesting use case.

Chris: Ooh, interesting.

Aaron: Where you think about, you know, what a new lawyer may do at a law firm is they may read a lot, and then summarize that for their team. You know, I’m not a lawyer, but I do know that happens. And I think these tools will be able to do a pretty decent job of summarizing past cases or whatever it may be pretty quickly. Again, it’s a tool that they can use to accelerate their work, but you’re of course going to want a person that’s educated and knows what they’re doing and is experienced to validate that work. Similar to a programmer that may take, to bring it back to the COBOL example, I think you could absolutely give one of these tools a chunk of legacy code and have it translate it into another language, and it may be 95% of the way there. But yeah, through testing and validation, and just understanding, you know, syntactically it may be correct. It may compile, if you will. But does the business logic still work?

Chris: Exactly.

Aaron: Does the payment processing… this could be a bank system.

Chris: Right.

Aaron: Does it still process payments correctly, or you know, this run… 50-year-old piece of software is critical for planes to fly. So I think there’s going to be… again, it’s a tool to get a headstart, but it’ll be interesting to see how close they can get, especially years from now. How close they can be from a correctness standpoint. 

Chris: Yeah.

Aaron: To where, essentially, your translation is actually reliable enough to be replaced with little human input.

Chris: How do you think this changes, you know… people who are going to be getting into engineering in, say, 5 or 10 years? Or maybe even shorter, maybe in the next 2 to 3 years? Like, would you tell college students who are listening to this right now, “Hey, you better get familiar with these tools, because they’re gonna change… they’re gonna become critical to how you do your job”? 

Aaron: I think at this point, you know, most day-to-day work hasn’t really changed at all. And, you know, it’s… occasionally you pop a question into this tool or another tool that’s, I would say in much more common use day-to-day for engineers is GitHub Copilot. 

Chris: Yeah. So can you describe Copilot?

Aaron: Yeah. So GitHub Copilot essentially is a tool… so, to connect the dots here, Microsoft is a big investor in OpenAI. Essentially, they’ve helped pay for a lot of the compute that OpenAI has used to build these models, which are very expensive to train. I think it’s also worth saying that models have two main separate things. There’s training, which is when you train the model itself on all the data, and then there’s what’s called inference, which is like… the runtime. Like, the actual query is sent through the inference part of the pipeline, which is essentially like the part where you would be searching, like in a Google search box.

Chris: Got it.

Aaron: And so, there’s Microsoft who’s a big investor in AI, or in OpenAI. They own GitHub. And so there’s a tight relationship there. And GitHub Copilot is also a Large Language Model powered tool that you can use in your code editor that will… one good example is, you can write a piece of code, let’s say it adds some numbers together. And then you can write a code comment and say “Write me a test that will test this function.” And essentially it keeps track of the context that you have onscreen in your project, locally, and it will show you – essentially let you tab-complete – what it thinks you wanted to write. 

Chris: Mm.

Aaron: And so, what I’ve found so far is, most of the time it is helping me save from actually typing keystrokes, where it’ll do what editors have done before with autocomplete, but sort of autocomplete on steroids. And so I think it’s a productivity booster, but it’s in the same vein as, you know, Copilot used in the right way can have some of the same benefits as what we described with ChatGPT writing code. It’s just a different environment. But it’s a really cool tool. But again, much of the output it produces is wrong at first glance, but could be as simple as the brackets are wrong, or it gets the wrong variable name. So it gets things wrong, but it’s still overall, I think, helpful.

Chris: Do you think as we progress, we’re going to get further and further away from the actual implementations? Meaning, is there a future where you have product managers or designers or even, you know, non-technical executives… they describe an interface in plain language. “I want a report that shows me these four data points updated over the last 12 months,” and something is generating… essentially like a program, an app, on the fly, that is not… you know, it’s writing the code and it’s writing the interface code to actually draw the complete picture. Like, do you think that’s where we’re headed in the… I mean, I was going to say long term, but maybe it’s medium term or short term, I don’t know!

Aaron: Yeah. I mean, I think in the short term, and it’s easier to see, is… let’s say you’ve got a product designer that’s very good but doesn’t quite know the right CSS rule to implement an animation, for example.

Chris: Hmm.

Aaron: I guarantee you, you could ask ChatGPT to translate, you know, some plain language into “How would I do this style of animation with CSS?” and I bet you it would help them answer that question so they could hand it to an engineer or implement it themselves on a website. I think the higher-altitude you get, the longer the time horizon is that it’s possible.

Chris: Yeah, that makes sense.

Aaron: So I think an executive describing, you know, “I want a full app that does all of these things” or whatever, we’re probably quite far away from that. I mean, far away meaning years but probably not decades. But I think there’s gonna be, yeah. There’s gonna be lots of little improvements that build upon professionals doing their jobs faster. Another good example, like on the design front, is what Adobe’s worked on where you can like, drag a line through a Photoshop, in Photoshop, drag a squiggly line through a painting and it produces a river or whatever. 

Chris: Yeah.

Aaron: You know, stuff like that where designers can use it to enable them to iterate or test ideas, and then fine-tune it into the level of quality that they expect it to be. I think there’s gonna be a lot more of that.

Chris: This reminds me, it’s a great example in the sort of design, the creative space, right? Like, content-aware fill in Photoshop, and other photo editing tools, right? People are like, you know, AI, machine learning, I don’t really understand what this is. But there are these practical implications, right, that light up. Like, I mean, everything we’re talking about with ChatGPT, but also things like, if I have a photograph and there are trees and a waterfall, and I wanna delete the waterfall and put the trees in the whole background, there was a time when you had to do that by hand. There was a time when you had to do it physically, like with an airbrush or something.

Aaron: Mhm.

Chris: And then it was by hand in Photoshop, like manually putting pixels around. Now you can lasso around something, hit delete, and then say, you know, I’m gonna use the fill tool but I’m gonna use the content that’s around what I’m filling. And Photoshop will figure it out and put the right stuff there. And it used to be kinda janky, but now it’s like, pretty good. Like, it gets you 90, 95% of the way there, depending on, again, how granular you’re looking. And then you just have to go in and tweak a few things and clean it up. So it’s the creative version of, I think, what you’re describing about writing a piece of code. You know, these tools are getting better every year, every month. And I think we’re going to continue to see these productivity boosters in a lot of different areas, which under the hood are based on machine learning. They’re based on, you know, training a model and then understanding what the substance of something is and figuring out how to amplify it.

Aaron: Yeah. And I think there’s an element of this where… in the same way that, CGI and movies, we got to like 95% pretty quickly, with a long tail to get from the uncanny valley part of it, where you…

Chris: (Laughs) Right. There’s a lot of effort there.

Aaron: It’s pretty hard to fully get to that 100% effectiveness. And I think that’s the same thing with self-driving vehicles, where there’s an argument that you need, like, full AGI, so artificial general intelligence, where essentially you’ve got a full AI in the sci-fi sense where it can drive a car as if it was a human. I think the long tail, I think a lot of these scenarios like CGI or self-driving, there’s gonna be a curve here where you’re gonna need a human to validate and… They’re gonna move faster, designers are going to move faster, programmers are gonna move faster, but you’re still gonna need that human component for a while. But we’ll see. 

Chris: I know. I mean, as we think about evolving these implementations to inject some creativity, or… whatever that means. I don’t even know what that means. But that’s… it feels like that’s a new frontier that’s going to be explored. Which is, again, kind of scary. (Laughs) But we’ll see! We’ll see what happens.

Aaron: Yeah, and I think maybe we should talk about the scary parts or the less rosy parts of this. Which I think, one of the things on my mind is… it comes back to, yeah, the word “discipline.” And I think, when you think about, let’s say children in school trying to learn, actually truly learn things…

Chris: Totally. 

Aaron: These tools are very tempting, right? To use to complete your homework or whatever it is.

Chris: Oh my God. Write me a 3 page essay on Abraham Lincoln.

Aaron: Right.

Chris: And it’s gonna get it pretty good. 

Aaron: And this is a tool that has, essentially, a lot of the works publicly available on the internet to write the essay, and it’s gonna do a pretty good job. I mean, I’m pretty sure I read it passed the bar recently.

Chris: The… like, the law? (Laughs) Oh my God. The bar exam?

Aaron: Yeah. The law. It passed the bar to get a law degree. And I think the reality is, schools may try to ban these tools, but kids will get access to them. And I think it’s gonna be parents and everyone else to try to convince kids that, yeah, you’ll use these tools like you used a calculator to do your math homework, but it is important to focus in and learn a lot of the components of, like, what allows you to be a professional or whatever. Master a skill. And so it’s going to be challenging as these things get really good, and it’s tempting especially when you’re a young kid and you want to go do something else to not just use these tools to skip that learning part. It’s gonna be a challenge.

Chris: Yeah.

Aaron: But on the flip side there’s a lot of upside. So, I mean, this is gonna be… for humanity, a really tricky thing to navigate. In the same way that things like social media have been really tricky to navigate. I mean, AI’s gonna be another challenge. 

Chris: That’s right. I mean, with the advent of the internet we had all the world’s information at our fingertips and we had to adapt as a species to figure out how to wield that power effectively. And this is just another leap. It’s a big one, it’s a generational leap, and it’s going to continue to get better and easier and more accessible, and we’re going to have to figure out how to adapt. There will probably be missteps. There will probably be things that it’s like, “Oh, we didn’t anticipate that it was going to go in that direction, and we need to curtail it in this… you know, in this way, or make a law that says you can’t do…” I don’t know! There are going to be a whole bunch of situations that we encounter and have to respond to. But I think, I mean, education is a great example where we’re going to have to figure out how to make proper use of these kinds of tools, as opposed to just outright banning them. Because the reality is, there’s immense power here and this is where the world is going. Like, this is… it’s filtering in in all of these various places. So we have to figure out how to teach students to use things responsibly. You know? Which is going to be tricky.

Aaron: Yeah, very tricky. Yeah. And I think there’s also, I mean, there’s going to be the question of regulation and letting the free market do its thing with AI as it grows, and what governments are going to do with this, and how they’ll regulate it. I mean we talk about schools and if they ban it or not, but governments are going to be… you know, there’s going to be politicians that want to be aggressive on this and ban it, and there’s gonna be…

Chris: Yeah.

Aaron: …folks on the other side of that. And we’ll see what happens there as well. I’m sure it will be complex. And I think one thing, you know, for nerds like us where we write software for a living, and that’s… we know computers and technology so well. It’s easy to forget that a lot of people see, like, just pure magic, where a computer is spitting out what a human would. And it’s…

Chris: Right.

Aaron: It’s easy for those tools to get misused, and I think… yeah. There’s gonna be a lot of education needed about it. I think ChatGPT really opened the floodgates because it’s such an easy-to-use interface, it feels like you’re talking to a person.

Chris: Well I mean, this is the thing. And for those listening who haven’t tried it yet, you can. I would encourage you to go to And you’ll get… you know, sign up for an account and you’ll get a prompt and you can type stuff in, and, you know, I’m technically literate. Like, I consider myself a technologist, and I loosely understand the idea of building a big model and then figuring out how to, you know, query it to get this kind of answer. That’s kind of the extent of my… so maybe I am a little bit out of my depth. But it’s still mindblowing. Like, it is just… it’s unbelievable to see it come together in the interface that ChatGPT, you know, exposed. And I think the same way that when I saw content-aware fill in Photoshop for the first time, I was like “That is magic,” you know? And we’re going to see more and more of these things, I think, as these technologies get better. 

Aaron: Yeah. And I think… I don’t know where I first heard this, or if I came up with it in my own head, but I mean there’s going to be trillions of dollars in value in businesses created on the backs of all this stuff over the next decade. It’s going to be really big, and there’s gonna be use cases from design to engineering and everything in between, and people are gonna… someone, if it hasn’t already, is going to write that tool for lawyers that I already talked about that helps them move quicker.

Chris: Yep.

Aaron: There’s gonna be tons of value created over the next decade, and it’s gonna be really wild to see.

Chris: So I want to come back to the question of advice, maybe to close this out. Are there things that you would encourage people to look at and learn more about, or things to maybe avoid or treat with extreme skepticism, besides the things that we’ve mentioned so far?

Aaron: Yeah, I mean I think it’s good for, especially people in technology, I think if your job could be affected by this, which a lot of different jobs could be, I think it’s good to have some awareness of what’s going on. Or experiment with asking ChatGPT a question or two and seeing what it produces. I think it’s going to… whatever changes it makes to different industries, you know, there’s going to be a lot of, I’d say… you know, it’s like if it affects programmers, I think if you look at technology, there’s just going to be more people building software in 10 years, not less.

Chris: Yeah.

Aaron: So it’s not like… the jobs will be a little bit different, in that people will be narrating and conducting these tools a bit more. But they won’t, you know, it’s not like the jobs are going to go away. So that part doesn’t scare me. But I think, yeah, people should play with it. And I think really it’s a very exploratory time at this point. I think as a programmer, you know, testing out Copilot, it’s a good idea. It’s a good idea to tinker with tools and not be afraid to, you know, that this is something scary. I think this is something that’s fun, and it’s a new frontier.

Chris: It’s a new frontier.

Aaron: And that Rust… that Apple Health Rust code snippet made me slightly uncomfortable, and I think there’s gonna be some impostor syndrome…

Chris: (Laughs) Totally.

Aaron: …for folks when they use some of these tools. But then I took a step back and I realized, you know, I do have to be disciplined when I go in and look at that snippet and try to understand what it wrote. Using it as a learning moment. And I think everybody’s going to have to go through that phase of, I feel weird a machine did something quicker than me and pretty well, but then using it as a way to learn and help you accelerate what you’re trying to do. I think it makes a lot of sense.

Chris: That’s a great place to leave it. We have to come back again in 6 or 12 months and see how the world has changed and evolved. I’m sure there will be a lot different to talk about in this world of AI and tooling that it’s enabling. If you’re listening to this and you would like to talk about how this might affect your role or your business or something you’re working on, please reach out. We’d love to talk about it. We can get you on a call with Aaron and our engineering team and we’ll figure out how to, you know, help or at the very least commiserate with what is ahead with us for the brave new world that we’re about to enter. Aaron, thank you so much for being on the show and… we will talk to you all soon.

Aaron: Yeah, thanks Chris.

Chris: Alright. Bye.