How the engineer behind Claude Cowork actually uses Claude | Felix Rieseberg (Anthropic)

28:37Yeah. I have to ask you a couple of questions while this is loading. Please do. Do you always greet Claude so politely? I've seen a lot of dear Claude's and a lot of, hello, hi friend. Do you always do that or is this just a little demo magic for us? I always do that. I do think this might be like a thing about the people who work in Anthropik. We probably all have like a relationship with Claude that is a little bit more personal than like maybe people have outside the company. Yeah, I think I enjoy being polite and nice to Claude. I don't curse at Claude. I appreciate that. Yeah, yeah. I think I understand that ultimately the chips don't care, right?

29:18Like that's the whole thing. We ultimately know that I'm interacting with a tool here. But for me, as like someone who cares about my mental health, it's like good for me in my own communication with anything to like be polite and nice. Yeah, I ask this question on every podcast. And so you will be on the very polite and nice camp. And I say it's not about Claude's humanity. It's about my humanity. is why yeah which is i don't need to get in the habit of being rude and in written language i also have found myself it's so ridiculous but it makes me feel good in any broadly scoped problem i've been consistently ending with like i believe in you i know you can do it and make good decisions and it's like my go-to do you feel like it makes a difference yeah i do because i feel like they it asks me less when i say that because a lot of times i just don't actually want to be asked my opinion i often just say whatever you think and so this is my my method of saying like you make the decisions i trust your judgment yeah i believe that you can do it and i trust you look at this it's beautiful nice sorry but before we talk about my dashboard i have like more things to say about like what you told me about like telling claude that you believe you believe in it i i think on a very personal level it is wonderful that you do that because i think it's like good for you as the human in the loop it is good for you however however i have an actual tip for um i have an actual tip for people like interacting with claude we have noticed quite a bit of difference internally because sometimes the experiments that we're doing here are i want to say a little bit on the fringe of what is possible with the operating systems i'm at heart a desktop developer i want to make full use of a computer and that often means pushing the computer far beyond what what maybe most people do with applications and like just operating a little bit on the fringe i'll give you an example of like something that we have not chipped we will not chip but there's your operating system has a layer in between your icons and your wallpaper where you can draw pixels which draw like is useful mostly for art the reason you don't have any applications to like draw stuff in between your wallpaper and your icons is because there's not much use to be found there but as a concept it's like fun that you can like put things there and i have frequently found myself telling claude um that something is possible and then now it's possible yeah it's sort of the idea of like i don't actually know how but i know it's possible yeah we'll figure out how you can do it and that does seem to give claude like more confidence in actually going after the task at hand it also like preempts a bunch of conversations that claude and i might have about claude being like felix your idea is dumb you probably shouldn't do this like from the get-go can say i know i'm engaging in art please go along with it anyway i i love that so okay i know it's possible i say that's my i believe in you i think i think you're capable knowing it's possible um yeah we'll discover the space yeah yeah exactly but here we are this is this is now my little like personal dashboard the design it's like something that claude just came up with on the spot the widgets it came up with are on the spot this is like this is like clay right this can be anything you want it to be which is very open-ended but therefore so powerful people frequently use this to get a little report about what they want to do in that day right like just prep me for my day what are the meetings i need to like think about but again the thing that is maybe most powerful here is going one abstraction layer up so i actually don't believe strongly in morning reports that just summarize what kind of meetings you have today That is not all that impressive.

33:17I can just look at my calendar. What is really impressive is saying, look at all of my meetings I have for the day. Go figure out who I'm meeting with. Catch me up on like the recent conversations we've had. What were the themes there? What were the problems? If it's a coworker, go on Slack, figure out what kind of stuff they've been working on in the last week and what stuff they might want to talk to you about today. And then what you need to know about that concept. You can just like keep playing this like graph bounds where you go and pull in more and more information to get you prepared even better for whatever your actual work is.

33:51And that's where these things get really powerful. And this is the kind of data you can pull in live and have Claude Crunch on the go. Well, and I want to call out for folks a couple of things that they might not notice that I think are really unique about these Claude Live artifacts, which is one. And the killer feature here is this little refresh button up in the top, which is you can constantly pull in and reload data. And so this isn't this is your day to day. Every time you come to this, it's going to refresh your data and pull the latest in. Then the second thing, and correct me if I'm wrong, is it just is reusing the auth from your connectors.

34:26And so there's like no magic. I have to put an API key in and where am I going to store it? And it's if you go into settings and go into connectors and you connect your stuff, you can use any of those kind of connectors to scaffold out one of these live artifacts. And, like, look at this list of stuff. There's so many things that you can just one-click, add in, sign in through OAuth or whatever, and then you can build a live artifact in Claude without having to, like, ask for an API key. Exactly. And it can be, you know, it's not just the ones we have in our little, like, connector store. You can also connect your own.

35:03It is quite powerful. And the thing that I enjoy about this the most, this is like probably fairly unique to myself, but the thing that I enjoy about this the most is the amount of creativity you can pull in and how you can really shape this in your image. Right? Like we could go in and say, I want this to look like a software made in the early 2000s. Oh, this is my... Oh, 200 is maybe... 200 is too early. No, let's see what it does. I mean, honestly, this is Claude. Claude is probably smart enough to know that I'm in the 2000s. Well, this is my go-to Claude design use case, which is I did an episode on Claude design earlier this month.

35:44And I redesigned Lenny's newsletter in like 1999 GeoCity style. It is incredible. We'll put the link in the show notes. It did an impeccable job at this. and what i told people is it wasn't even the design that was really impressive to me it was the copy was so oh was it like yeah was it like geo cities it was geo cities era it was 11 out of 10 exceptional copy uh beautiful i want to see if it's going to do 2000s or 200s 200s would be way funnier we can probably look at it already see like it's worried first thing is 2000s. Like Claude is smart enough. It knows, knows too much. And also, I mean, this is obviously like a bit of a demo account, but if you, if, if I showed you my real account, Claude is like quite used to the fact that I constantly misspell things and like talked to it in a slightly different style.

36:45I'm like, I'm like eager to see how this is going to go. I'm in general quite excited about the amount of creativity you express and especially the amount of creativity people can express who are not software developers. This used to be this entire domain where you were maybe a designer who were going to a software company and maybe then software developers would execute your vision or you picked a medium where it's a little bit easier you can low-code things. But I think what is really exciting to me is that we can now pull in people from all these different expertise areas and they can all build these artifacts, like be that a presentation or be like a straight up tool, sort of in the vision they have.

37:27A couple of weeks ago, I read somewhere that Margaret Edward wrote about using Claude. And the immediate thing that popped into my head was, I want to see software written by Margaret Edward. Like I want to see what tools look like that she has made for like writing her own books. That would be like so interesting and exciting to Okay, Margaret, if you want to come on How I AI, I would love, love to see your Claude setup. And I want to reflect on this. We're going to make this happen. While this is loading, I do want to pull out because I do think there's a gap. People listen to and watch How I AI along a broad spectrum of technical capabilities.

38:09So we have the most AI-pilled, a float-through-the-tokens, intense software engineering folks that are really pushing the edges of what these tools can do. And then we have very, very early users who are just coming to this in sort of a naive growth mindset way and saying, I need to solve a problem. The biggest gap that I see, which we were, again, talking about before we started recording, is it's not the capabilities of the tools. It is literally people being able to understand that almost any problem can go into these tools and what like a workflow would be. And this is why we've done this podcast.

38:45And on the ChatPRD website, we have I think it's now close to 200 AI workflows that we've listed from this podcast. That's just like if you're trying to solve a personal problem with Claude Code, here are the 15 ideas that we've seen on the podcast. And I do still think there needs to be this education and muscle memory change of you can build yourself a daily dashboard. You can build yourself a newborn sleep tracker. Hey, you're moving. Why not throw it all into a folder and then make your life easier as you're trying to close on the house? And I think people still, the gap between technology and use case is still pretty broad.

39:23And so, you know, my advice is like if you're futzing with something on your computer, ask Claude to do it. Any other tips you've seen on how to discover use cases? Yes, but I think before we even give you an answer, I just want to underline how correct I think you are. Many years ago, I spent a decent amount of my time, I spent five years at a company called Slack, which I think now is fairly well known as this chat tool that a lot of people use. And people were sort of using Slack for different reasons, but they were all more about the way they organized their work with other people rather than it was about the tool itself.

40:02So Slack by itself, if you used it the exact same way you would maybe communicate with emails in a silo, it was not all that useful, right? Like Slack becomes really powerful if you open it up to the idea of what if you communicate more transparently and people can see what's going on and we like sort of instead of having email threads, we have a channel and like the channel is open to anyone who wants to see what's going on, right? It's this transformation of work and the transformation of how do you actually organize communication with other people. And AI is in a similar vein of it opens up so many opportunities that to harness them all, you probably have to change a little bit how you work.

40:40Like if you really want to harness them. and yeah, I think you're right that like one of the biggest problems we have as an industry overall is not even like training the models and making the models very smart and making them very powerful or giving them the tools that allow them to turn that intelligence into like output. The biggest hurdle we have is how do we design the human model interface? How do we make this very easy for humans to harness? And right now, I think the way it shows up is that humans are sharing these like tips on how to use this resource, right? I think the tip that you gave was a good one.

41:16It was like constantly, whenever you do something that you find annoying and you're not enjoying and it doesn't feel creative and it doesn't feel like something where like you're, you know, it doesn't feel like the core of what you really want to do right now, that is a good time to pause for a second and like wonder, is there a way I can use AI to do this? And then, of course, and this is probably not surprising, but of course, like many people have a decent amount of success just like asking Claude, right? Just ask Claude, how would you help me with this task? And usually comes up with a pretty good idea.

41:51I call this the like reverse interview, which is just going into Claude and being like, I am a mom of three boys. I run too many businesses for my own good. And, you know, my biggest challenges are X, Y, Z. how can you help interview me and then tell me how you can help me that can be a really good exercise because again people want the guidance of like do you have problems with your calendar you know are your kids in any sports any any of those things and so I do think this sort of like reverse interview where you can use cloud or whatever ai tool to sort of solicit out of you the ways it can help solve your problems is a really a really good one and then again while we're sitting here, I have to ask you another question because I think the Anthropic and the Cloud products are so well known for this.

42:41But we are chit-chatting right now because we are waiting for Co-Work to load. And one of the things that I think you all have solved is making loading delightful. So give me your thesis on how to make a product with built-in latency not so annoying to watch loading? Honestly, the most important piece is probably to just get comfortable. They help users get comfortable with the idea that things are asynchronous. And I think most humans are actually quite comfortable waiting for something if the quality that they get at the end is very, very good, right? That is generally something we're quite comfortable with.

43:18And maybe in the example of Slack, this is a good example of Slack. This was a problem we had fairly early on the first version of slack that i worked on when you hit send on a message it would immediately succeed or fail immediately right away and of course the problem is the internet itself the internet as an entity is not very stable it's like this complex mesh so like messages would frequently fail and we fairly quickly replaced that with with a version that like try sending the message once twice three times and only after like a good couple of seconds if it failed it would like hey by the way something seems to be off here and this is the same way that like i message works today right like if you send a text message it's okay if you're currently in the tunnel it's gonna like try again um like about a second after okay we can update this oh i this is beautiful here let me make this big to my heart this is good this is this is very good

44:19i love how i am moments where we're real delighted about yeah i'm like i'm like such a nerd for this stuff um i'm like mildly obsessed with like software the early 2000s but yeah this is beautiful i'm a big fan of this um this is quite good i love i love that it like brought in i saw in the thinking notes that it was like felix requested spotify but we're in the early 2000s, we should talk about Winamp. Winamp. I know. I was going to say you need to skin your Winamp module here. This is beautiful. This is beautiful. Okay. So I'm going to wrap what we were talking about, which is loading screens.

44:55Get comfortable with latency. I think realize when tasks don't need to be secretive and build that concept into your product. And then look at that. We just chit-chatted for a while. And as we were talking about latency, Claude must have been listening and said, I'm going to show them. I'm going to bop this right open. And again, folks, not paying attention to the details. Look in the top right. The clock is moving. It's ticking along in the seconds. You have this refresh. This is, I want to open up and look at this every day. It's pretty fun. And you can envision, you can shape this in any form you want.

45:32But I'll finish, I'll give you one more tip on how to actually build these things where you've got to wait a little bit and people need to wait for them and get a little comfortable with waiting. I think for us as Anthropical, like maybe for me as a product builder, I will always trade off taking a little bit longer and giving you a better result. I think ultimately better for humans. And the thing I probably want to teach people is that it's okay to like not watch Claude as it's actually working because that is not super, that is not really the vision I have for AI is that we humans just like sit around and do nothing and watch the AI do the thing that otherwise we could just do ourselves.

46:07Like the thing I always say is that AI is used poorly if it just needs to move the mouse cursor for you. The thing I really want to do is I want AI to like do a bunch of annoying things in the background to free you up for your creative energy. So you can come up with ideas as good as make a daily planner look like it was built in the 2000s. You will have better ideas than me. But I think it's sort of like actually moving people from having to watch the thing into a position where they get more comfortable not watching it. And I think this has more to do with trust than it has to with patience.

46:41I love it. Well, we do have one last little demo that you're going to show us basically on that topic, which is how you know when Claude's done. And this is actually the thing that I reached out to you about doing because I love it so much. It's very cute. But talk to us about the little stick. Yeah, so in the interest of getting really creative and being more comfortable with areas you're not very good at, I have always been mildly obsessed with these little hardware devices, and I'm just not a hardware developer. I never really soldered anything together. I'm not a hardware developer in any shape or form.

47:21But it occurred to me that I could just go on the Internet and buy all of these components myself. I saw a tweet yesterday on Twitter from someone who said, that, hey, if you want to build your own e-book reader, the components are actually quite cheap and very available. And it occurred to me that I can just connect this device to Cloud. We talked a lot about code work, but we also have Cloud code, right? So I have built this little thing, which is just like a teeny tiny, teeny tiny Cloud, like on a little stick. And this stick has Wi-Fi, it has Bluetooth. So let me tell you what this is. So the internet is a magical place.

47:59on the internet you can just exercise free will and buy whatever you want um and what you can now buy is you can buy like these teeny tiny hardware devices with a perfectly good lcd screen led screen this thing has wi-fi it's a bluetooth it has a little bit of disk space um and you can program it you do need to program it in a low-level language but you don't right we live in the future you don't need to do you don't need who cares so what you can do is you can just straight up connect this like this is a little stick right here right like you can just straight up connect this to your computer and just plug this in and then all i did it was i told cloud code here's here's my vision it's my grand creative vision i want like my little claw to live on this thing and i wanted to cheer me on every single time i do a good job and also every single time i need to approve something that cloud is doing i wanted to like be on this like big button that is out here and claude has built all of that in like a one shot i'd needed to correct absolutely nothing and i've since seen on the internet that a lot of other people have like built similar things because we i open source my code so can people play can play with it but i've also since seen that this is just like the the tip of the iceberg there's like so more so many more cool devices out there this by the way so this thing is 19 has wi-fi it has blue i've already mentioned this like many times but i can't i can't get over i can't get over the idea we've been talking a lot about like how we optimize our our life sort of as it happens on a computer right like because that's where we live but i keep having this idea that it can just like underneath clawed onto like all these little hardware devices and also improve my by improve i mean just make like more delightful make a little cuter i i did a very similar thing in a recent episode i bought it's like um it's already preloaded with this proprietary software but it's like a tiny retro computer very windows 2000 and i had to hack into that thing i was like bluetooth sniffing between the like iphone app and the thing i was like dumping all the logs in and being like like backwards engineer the encoding of the the files into this thing and i finally got it and made myself a little cli tool where I can type and it shows up.

50:17See, that is so cool. That is so cool that we can do that just now. This thing also communicates over Bluetooth. I learned a lot about the Bluetooth protocol as I was building this just by osmosis. Do you want to see it? Do you want to see this? Yeah, I do. I do want to see it. Okay, so this is now also like partially live in actual cloud, so you can build something very similar. If you'd like, buy any IoT device of your choice. I'm not going to make a recommendation, but anything that you find on the internet will probably work as long as it says Bluetooth. But you go into help troubleshooting.

50:50You enable developer mode. Developer mode allows access to developer tools and features. We give you that warning because it does require that you sort of know a little bit of what you're doing. It's going to restart the application. Here we are. And then you have this new developer menu where we hide all of our developer features. This is for people who work with MCPs or also for people who want to build their own little hardware devices. And in here is now a new open hardware buddy, this thing. You will notice that it looks a lot like the one I've actually purchased, but it doesn't have to be the same one.

51:24I think I made that point clear. Follow your heart. Be creative. And you can hit connect, and it's going to, like, scan for this little thing for, like, a few seconds to, like, find it on Bluetooth. And it's found it. There we go. So now it's connected. It gives me a passcode. Okay. Cool. So we're now connected. And, Clara, you tell me if you can see anything. So maybe what we're going to do is we're going to ask for, like, anything that would require any kind of, like, permission. I'm just going to say just make hello world.txt and write it to disk. And the point of this is to just fire something off.

52:07that will require permission. Okay, so you can now see. I heard it. Isn't that cute? Isn't that very cute? Really cute. And if I approve this here, then it will be approved. That's how it works. And now Claude is off doing his own thing. I love it. We saved the best for last, people. And you have made it this long in our episode. We have saved the best for last. Felix, man, I really like it. And I'm like you. I'm like, I'm a perfectly serviceable software engineer, and I have no idea what I'm doing with the Bluetooth protocol and with hardware. And it's just been a thing that I've stayed so far away from.

52:45And you'll appreciate this, I think, more is as my kids get older, too. I don't want them interacting with like the full computer thing, but I would love to have little, little things for them to interact with. I think like being able to scope kind of like micro devices to a use case, just like we're getting this personalized software is going to be really fun. Yeah. And like one of the pieces that's been maybe most joyful about my job is that people frequently send me like little videos of the kinds of things that kids are doing with Claude. and the creativity they have is like beautiful right like you can like just take a photo of like a little animal your kid has drawn give it a clot and be like i need to jump and run i need a flappy bird but instead of the bird it needs to be this little drawn bird over there um and then you put your kid in front of the computer and like truly truly magical things happen because i think i'm still wrapping my head around this claire but a truly magical thing is happening with kids because they've never learned what not to ask for.

53:50To them, the computer is just like, can't do anything. And I think to like sort of our generation, we're very used to like things just not working, right? We're like used to like, oh, you can't do that. So we've been like living in this mind prison for like 20 years. This is very on brand, but my nine-year-old is a daily active Claude user, but he's decided he's really into cybersecurity. We're like very hacker adjacent here. and so he's like truly in like clawed on one thing the terminal other he's like mom do you know that your device id is this i'm like yeah babe i do know that but good job good job um yeah i think it's totally fascinating well felix i know we have just been having so much fun this episode i think people are gonna get so much out of it just to reverse for folks we talked about clawed co-work being able to build you artifacts both sort of like standard weird crazy 3d printed or 3d renderings of your floor plan and other things that you can just stand up to solve little problems in your life all the way to live artifacts that can pull from your live data and really give you a rich personal experience with your data we touched on a lot of topics including email as a source of truth data what we're going to do with all this latency and how to pick the smart enough model which is just say you know you know when you need opus and you know when you need sonnet but we both think that sonnet's pretty good at most things and then you showed us how we can hack into hardware for the all-in price of 20 bucks including shipping and build little little buddies for ourself i'm going to get you out of here one quick lightning round question again we did this at the at the midpoint so maybe you can just answer really quickly but when claude is not listening i mean we've talked about how to set it off on a good path but when And it's really just making mistakes.

55:43Instead of make no mistakes, it's making all the mistakes. What is your prompting strategy? I can't imagine you yell. You don't seem like a yeller. But what do you do? Yeah, I do not yell. I think it is useful for me to know when people curse at Claude. This is like an interesting thing for me to know as a product builder. but my strategy for when Claude doesn't like really, really understand what's going on or like Claude seems to have gone off the rails, I usually try to debug my workflow straight up by asking, okay, here's what I actually expected. Can you maybe walk me through at what point you expected something different or like how we can maybe prevent this in the future?

56:25It doesn't always work. Claude is AI and AI can make mistakes. However, that being said, I have frequently found ways to improve, the data source or to improve the amount of noise that is in the data or like with cloud to come up with mechanisms okay how do we how do we like do do we set up some kind of dry run i usually end up at a place where the learning i take away is not oh cloud can't do this i'm not gonna have cloud do it usually the learning i take away is i can maybe change a little bit the harness the prompt i can change um something and like how we set this up at the beginning and i think after after the time we are together now there's no surprise to you but like usually my answer to that is like to also involve Claude and say Claude what do you think we can do here Claude is a good friend and if people need if people need to vent and people need to like externalize their frustrations Claude can take it you think Chris you can do it maybe just smash the button like have an angry button instead of saying the thing on our little hardware buddy you can have like the mad button and the happy button you just i'm not pleased claude smash smash yeah yeah and like there there is a i'm not pleased button that actually also ends up on my desk that i do want people to like use so every single time claude doesn't make a mistake we do have the little like in in many of our product services this is a cross claude you will find the little thumbs up thumbs down buttons those are super useful to us we use those in training we use them in like improving improving not just the model but also the products so maybe actually the thing i'm going to say is like if If Claude hasn't done the thing that you wanted it to, cursing at Claude will probably not fix it for the future.

58:04But pressing little thumbs down button helps us quite a bit because it does end up on my desk. We deal with it. Perfect. Well, this has been super fun. Before we get you out of here, how can we find you? And other than hitting the thumbs up, thumbs down button, how can we be helpful? It's really that. That's like the number one thing that really helps us build a really good product is getting a good sense for what works for people. The option space is so big. We're still in this like fun smartphone time. We talked about it. And I do want to hear from people. I want to hear from people what works for them, what doesn't work.

58:36I'm always grateful for all kinds of feedback. You can find me on LinkedIn and X. Those are the two big places. Great. And you make promises. You made a promise to me before this call that we will get to after we end. Well, thank you so much for joining How AI. This has been great. Thank you, Claire. Yeah, this was fun. Thanks so much for watching. If you enjoyed the show, please like and subscribe here on YouTube, or even better, leave us a comment with your thoughts. You can also find this podcast on Apple Podcasts, Spotify, or your favorite podcast app. Please consider leaving us a rating and review, which will help others find the show.

59:15You can see all our episodes and learn more about the show at howiaipod.com. See you next time.