Loading summary
Transcript7 lines
- [00:00]
Kyle Barber
The last two weeks have been by far the most exhilarating of my entire career. I mean, the momentum right now for Inflight is very real. But I'm not here to tell you about what's going right. I want to use Dive Club to provide unprecedented levels of access to life as a design founder and share. The messy behind the scenes and all of the winding turns that we've taken to get where we're at today, and I'm holding nothing back. This is going to be as raw as it gets, and hopefully by shining a light on some of the things I'm learning, it turns into a video series that you want to keep watching. So let's start by meeting one of the main characters of this story. Kyle Barber. He was actually the first designer I ever hired way back in 2013 when I was working on my very first startup. Spoiler didn't go very well. Anyway, he went on to design a ton of Tinder, did some time at Facebook, and most recently designed Air Chat. For the most part, we kind of lost touch. But then he listened to my episode randomly with Steph Golick talking about design founders. Except it took him a whole 40 minutes to realize that it was me on the other end. So we reconnected, went back and forth on his vision for a few months. He had a viral launch tweet, raised a pre seed, and eventually couldn't take it anymore. I had to join as a co founder. I like to joke with people that I flew a little bit too close to the sun, but here's the thing. I joined to work on something very different. Inflight started out as a marketplace for design feedback. So behind the scenes, I was building a network of all the best designers I knew, so. So that people in teams could pay them to get them to critique their work. I'm talking like design celebrities here, people that you almost certainly follow on Twitter. But pretty quickly we realized two things. The first is that marketplace unit economics are pretty freaking hard. Like, it's really difficult to get the math at work. I spent a lot of time in Claude building these interactive spreadsheet type things just to play with different levers of what a business model could be. And the second thing that we realized was that the software that we built to support the marketplace was actually way cooler than the marketplace itself. There was a very specific moment that I had sitting in this chair where I realized that I think I would use Inflight instead of using Loom the next time that I wanted to drop my work in progress into Slack. And this is a really big deal for me. I mean, I have been the loom champion at my last five roles. So starting in January of 2025, we set out to build a new type of design feedback tool. It started off incredibly simple. It was just a spatial canvas. With this new type of video cursor interaction, you could watch as people fly around and give feedback on your designs. And even in its simplicity, honestly, it kind of felt magical. But we started showing it to different designers and pretty quickly realized that people wanted to view transcripts too. So that UI definitely had to evolve. And this started a process of obsessive dogfooding and iteration. I mean, we rebuilt inflight three or four times while doing about 200 different demos with designers and teams all over the world. We jump on a 30 minute call, have someone tell us why our product sucked iterate and then repeat. We weren't even adding people to the product yet. And these were just demoing all kinds of concepts, trying to figure out the right mechanic, the right interaction patterns, what the role of AI even was in this product. I mean, we went deep down the rabbit hole of AI avatars for feedback and who knows, maybe it'll come back someday, but right now it's kind of mostly generic nonsense and a bunch of accessibility linting. But you feel this pressure to make something that's unique. We were doing a lot of things differently. One concept that stuck around a while actually was the ability to create your video in chunks and have people respond to different threads. But it was still missing something and we knew it. We couldn't put our finger on it, but we knew it wasn't ready for primetime. Which, let me tell you, it's not a great feeling when investors and advisors and everybody that you talk to is telling you to launch and iterate. You know, classic startup and MVP advice. You know that Reid Hoffman quote, the one where he says, if you're not embarrassed by the first version of your product, then you've launched too late. That thing haunted me. And deep down I had this mounting doubt that we were using craft and this incredibly high bar that we established to punt the real fear of launching down the road. It's really, really easy to convince yourself to just take one more iteration when you have two different design founders. But it did make me feel better remembering how Nate Parrott talked about the first six months of designing ARC when he was at the browser company in the early stages.
- [05:01]
Nate Parrott
We were kind of lucky in that we were building for ourselves. You know, people who spent a lot of time in the browser reasonably tec tech savvy, but not everyone engineers. We were kind of lucky in that if we built something, it was pretty obvious to ourselves whether it was useful or not. You know, you can try to build a feature into the browser and use it day to day, but you know, you have to use this tool for work. And so you can't really kid yourself about what works and what doesn't. And so we actually had, I think our superpower in the first six months or so was that we were the audience and we were the daily users and we were using this tool and we were fully dependent on it for so much that, like, we didn't really need to do user research or focus grouping or anything like that.
- [05:41]
Kyle Barber
So, like Nate, we kept iterating and kept dogfooding and eventually we had an insight that put us on a pretty different path. People suck at giving feedback. It's a pretty clear reality. You probably experienced that if you're a designer listening to this. So we had an idea. What if instead of reviewers making their own loom esque video, it felt a little bit more like having a conversation instead of. And this kind of came out of the way that I was personally starting to use AI more and more. I was just spinning up a voice conversation and prompting the AI to ask me questions as a way to help me think more clearly. And I was having a lot of success doing it. So we asked ourselves, what if that's what giving feedback looked like? So we started playing with GPT and 11 labs and using AI not only to help the designer ask better questions, but using AI to have a dynamic back and forth with the people reviewing the work. And it was pretty interesting. I'm not going to lie. At this point. We started onboarding people, and whenever I get to that part of the demo, I'd always peek up to see their facial expression because it was pretty cool. Like we had finally achieved that level of novelty that's almost impossible not to hope for. So all of a sudden, I'd spent whole days where my entire design process was happening inside of 11 labs. As I tinker with system prompts and language design, it kind of made me feel like a junior designer again, if I'm being honest. Which brings me to another dive club lesson that I learned kind of the hard way. During this time, it's really, really easy to over architect your system prompts because you find something that needs to be improved that the model isn't really handling well out of the box. So you had a little conditional statement here and maybe another one over there, which creates yet another edge case. And all of a sudden, you're playing whack a mole and defining every single possible use case. So this lesson from Tuan Kumar's time designing Luma AI really hit home for me.
- [07:40]
Tuan Kumar
We had to create so many of these, like, special case rules, like, if this happens, do this. In this case, do this. This is how you break it down, lo and behold. I think one of the things we are realizing is every time you approach in the LLM side of things a problem like that, you're putting a ceiling on the intelligence yourself, because the way you wrote those logic pathways and those rules are the limits. No matter how intelligent the model gets, that is all it can ever do, because you're literally telling it not to do anything else. And I think in hindsight, we should have, you know, this wisdom that I was spewing a second ago of, like, knowing the arc and everything. I should have taken that myself and been like, no, no, no. By the time we launch, we would have it. That's like a perfect case of, like, where clearly the methodologies got better very quickly over time.
- [08:27]
Kyle Barber
You really feel this timeline tension when you're working with AI. You know things are going to get better, but it's not quite there yet. So what do you do? Well, to further complicate things, we accidentally stumbled into potentially yet another pivot.
- [08:42]
Steph Golick
Real quick message, and then we can jump back into it. Look, I'm not a motion designer, but I'm working on a new product right now, and I've been using Jitter for everything, and it's honestly a cheat code. I've been creating Lottie files, micro animation showreels for Twitter, and I didn't want to use After Effects because it would have taken me all day to figure that stuff out. But there's no steep learning curve with Jitter. The whole product is super intuitive. They even have incredible community templates that I've been using to jumpstart my project. So even though I wouldn't consider myself a motion designer, I feel like Jitter gives me superpowers, which is why I can't recommend this product enough. If you had to dive Dot, Club, slash, Jitter, you. You can spin up something stunning in a matter of minutes. I'm all in on paper as the next great design tool. And this is coming from someone who's taught Figma to more people than just about anyone. But I'm ready. It's time for a tool that puts creativity at the heart of Everything, not systems and processes. The way that we design and create software is changing faster than ever. And the tools we're using were made for yesterday's tasks. Instead, Paper is putting designers first and building all kinds of features that I've always wanted. So if you want to be one of the very first people to use the next generation of design tooling, then head to Dive Club Paper to get on that list. Early Access launches in May, so that's Dive Club Paper.
- [10:10]
Kyle Barber
Okay, now onto the episode. Once people saw the AI interviewer, the first question they ask is, well, can we send this to external users? I mean, we had people saying that they would cancel Maze immediately to use Inflight, which created one of many existential crises because the underlying primitives of Inflight are very flexible. So there are a lot of ways you can slice the icps and core value propositions. Is this a tool for client feedback, internal team feedback, external user testing? Is the value prop, speed, quality, organization, alignment? Trying to answer these questions accelerated my bowling by half a decade, I'm sure. But there was one outcome that felt like it still had the highest ceiling, and that was the internal product team. Because if you can create a product that is tailor made for this one workflow of dropping work in progress into Slack or wherever you work, then you can capture a totally new context window. Because right now all of that is lost in random threads and loom links and scattered figma comments. So if you could really capture the feedback checkpoint, then that would allow us to build new types of AI experiences that we might not even be able to think of today. So even though it felt like the more difficult path versus carving out some chunk of the existing user testing market, we weren't ready to give up. So we iterated yet again. And keep in mind, we've been onboarding beta users for a couple months at this point, so some users have stuck through a hilarious amount of production changes. Shout out Nick Basile. But we thought maybe we could strip it down a bit. And this is what led us to this Lume meets granola idea. And you know what? It felt pretty good. Now obviously as a designer, there's a million things that I hate, but for the first time, it felt like we finally had the right foundation. My dog food didn't taste so damn bad anymore. Not only that, but we started receiving some pretty high quality inbound. So after five months of iterating, five whole months, we finally shared our vision with the world. Was it the perfect path? Would I even share any of this? As quote unquote advice? Probably not. But I'm glad that we trusted our gut because it's kind of working. And last week, I had one of the best days of my entire life. I kid you not, but I'll save that for the next entry. Ra.