Julian: Hey everyone. Thankyou so much for joining the Behind Company Lines podcast. Today we have ColinTreseler, CEO of Supernormal, where you can create detailed meeting noteswithout lifting a finger. Supernormal AI automatically transcribes and writesto the meeting notes. Saving you hours each day.

Colin, I'm so excited to chat with you,not only because your entrepreneurial journey and, and so excited to learnabout, your structure. I won't, I won't break into anything right now, butoverall, excited about where you've come from and what you've learned over theyears, but also what you're doing in the AI space.

I know we, we chatted a little bit aboutit and and pre-show about the, the, the increased popularity of ai. But, yourcompany has been working on it for a while and, and I'm curious. What you, whatyou've not only seen in terms of development of your product, but also in thespace overall. But before we get into all that good stuff, what were you doingbefore you started the company?

Colin: I've, I've been intech for almost 15 years now, and I moved back from Silicon Valley toStockholm, Sweden. In late 2015 and jumped in just building stuff, right? Wasdoing some consumer social, was doing some video apps. But time kind of went byon all of those. We, we scaled things up to 10, 15, 20,000 daily active users,but nothing exploded.

Come 2019. Got my best friend in mybasement of my house thinking, okay, we're gonna build a distributed team. Whatdoes this look like? He had just rolled out of GitHub and, they were famouslydistributed, like 60, 70% of their workforce was not in the office. I was like,what worked well, what didn't, what was, what was great about it?

Right? And and that was kind of the, therole into, into where we landed. My kind of general experience as aprofessional sailor for a while of racing sailboats around the. And, and kindof realized that I wanted to use my brain a little bit more. Yeah. And, and tryto drive some impact. I, I thought that would be at getting people to do moretime spent on Instagram and, and help them help the world get more open andconnected.

In retrospect, probably. Move thosenumbers a little too much. I mean, when, when I started working on, on productthere, I think we were at 35 minutes per day on average for every single personon the platform and scaled it up to like 47 minutes per day, right? So onemight say that, that it helped humanity waste a lot of time.

I mean, billions of hours a day. Justwild. I think most of the, the work since then has been, okay, can we do timesaved? Instead, can we use our computers and these incredible machines to dowork for us?  

Julian: Yeah. Yeah. Andthinking about, the work that you did, obviously it, it is impressive and Ithink a lot of companies try to capture their audience and, and really offer,whether it's it's value entertainment, something where you have a return of,of, whatever your main customer base is.

And, and on a more repeated and moreactive basis, what are some things that, did go well in that experience thatyou've learned in terms of keeping people engaged, keeping people retained andthen what are some things as you. Some outcomes that, that you weren'tnecessarily so, so now I don't wanna say proud of, but you know, weren't on the,the cutting edge of what you were doing.

And I know you mentioned some hoursspent and things like that, but go to the first question. What are some thingsthat you, that worked really well? Learning about how to keep peopleengaged.  

Colin: Oh, wow. Yeah. So whenyou're working at one of the large platforms, you learn optimization, right?Yeah. So you don't usually go zero to one.

It's let's make something better. One ofthe biggest takeaways I had was when I redesigned was coming out and we wantedto roll it out to 1% in test, right? Yeah. Like the standard playbook of you'regonna ship something inside of meta, what is that gonna look like? Right. AndKevin, the CEO of Instagram, Kind of came into the room was like, guys, this isthe right strategic direction, like the right look and feel, the right design,just ship it a hundred percent.

Yeah. Just go out the door. Right? Yeah.So this, this feeling of trust your gut of, you know what, sometimes you justneed to make bold steps and not measure on 1% or 2% before you decide, hey,this is actually gonna be what the discover tab is gonna look like, right?Yeah. You learn a lot. Challenging your, your biases, right?

Yeah. Okay. So you think something'sgonna happen? Well tested, find out, right? Sure. So it's a real big likeoptimization game. Right. Yeah. And, and, and famously from like 2010 or evenlike 2008 to 2019, it was human capital. Just throw more people at it, moreproduct managers, more engineers. Right, right, right.

And then you see these massive layoffsat, at the large companies, because that's not actually probably the fastestway to ship. Right. Yeah. So one of the things that I, I learned still onworking on Instagram specif. It was, it was a fascinating project because it wasmachine learning, right? It was fixing search.

It was getting search better. Yeah. Butthe team was tiny, right? With 300 million daily active users. At the time, Ithink there were, in total 25 engineers. Right. Wow. So the scale at which youcan build things and, and do things. I've, I've copied that model in, in thework on Supernormal. Our, our team is still small.

We're about 13 people, right? Yeah. Sowhat you can do on a team of that size and how quickly you can iterate and, andchange and move is is critical, right? Time to market is everyth. Yeah, wewere, we were fortunate enough to be, when we got into the Supernormalexperience, first to bring large language models into meeting documentation,right.

Writing your notes Right and ahead ofentrenched players and even startups that were in this space by four or five,six months. And that leg up has given us such a, an amazing position to justgo, continue to iterate faster than the next.

Julian: Yeah. Now, do you,would you attribute that to, is it, is it a, a access to better technology?

There was a huge push in microservicesthat allowed way more efficient deployment. What in particular did you see thatwas. That has been so, because a lot of the people, a lot of founders think asyou scale up, you gain more customers, that you need more people, you need moresupport, you need more team.

Like you said, there's huge humancapital. But I mean, you, you, you, you literally worked on the antithesis ofthat idea and what was, well, how were you able to still be successful and,deploy at a high velocity.  

Colin: So there were threelong, dark years of us saying, we, we made this bed back in 2019.

Right. As Covid. Yeah. Was just rollingout and happening. Mm-hmm. We were like, look, AI's gonna be good enoughsomeday to write our meeting notes. It's not there today. Yeah. What's thesteps? What are the precursors for it so that when AI finally gets good enough,we can just jump on it. And until that point in time, we're gonna stay smalland nimble.

Right. Yeah. Three, four people max. Andwe did that for three. Right. Yeah. So team size if, if you need high velocityand, and being able to not like do full large pivots, but micro pivots and, andreally test things quickly. That was really helpful. We also just kept on asteady stream of 25, 30, 40 new users coming in each day from Google Ads.

Right. So just great. We've got peoplecoming in. What's this cohort doing? What's the feed? But we also, we did a lotof the unscalable things, right? So I would offer onboarding calls to everysingle person coming into the platform. And by the time I stopped doing that, Ithink I was, I had 15 to 20 calls a day for 15 minutes each.

Just talking with a new interestingperson, what they're trying to get out of this, right? Yeah. Yeah. Following upwith them in, in 20. So a lot of those like mechanics of, okay, this is theright way to connect with people and, and just be so customer centric learnedat the, the large platforms and then just carried that forward with the work,but still with the mindset of let's keep this team small and tight.

The other aspect was that we were allengineers. We were all builders, right? Yeah. Designers. So even on a team ofthree, the three of us were all had product roles in the past, had design rolesin the past, but were all engineers at. And that's, that's super critical,right?  

Julian: Yeah. Yeah. And onething you mentioned as, as just a second ago, this prerequisite to ai at thetime when you, you doubled down on this idea that it'll be better, it'll be,the models will improve, the systems will improve, they'll have moreadoptability.

What was the prerequisite of theprecursors that you saw that the AI needed to get? Up to where it is now, whereobviously it's become hyper popularized ChatGPT. Other companies are startingto either expose the fact that they've been AI all the time for a convenienceno, I'm just joking. But they, they've been able to just, it happens.

It happens, right. But, but a lot ofcompanies have been working like yourself in this space for a while, and nowit's, it's almost like the perfect timing to then scale and. But what werethose prerequisites? What did you need in place? Was it, was it technology? Wasit awareness? What was it in particular that that needed to fall into place?

Colin: So we basically saidthat, okay. At the time g p t one or two, right? Yeah. It was okay at formingsentences, but maybe, maybe a small paragraph. But that was about it. It stillcouldn't summarize in a meaningful way. Yeah. So what we did was we gave peoplebuttons in their meetings and said, Hey, hit this button whenever there's anaction item or a follow up or a decision or a key point, right?

Yeah. And we would tag the part of thetranscript and that. Just a giant data set for us. So after a few months, wehad 60. Thousand hours of meetings where people had gone in and tagged, Hey,this is an important part of the meeting, we would surface just thattranscript. And they would rewrite it. Yeah.

They'd rewrite it into their own notes.Right. This is what the action item should be. Ed gave us a, an amazing set ofdata that when AI finally got good at abstract summarization. So abstractsummarization is the idea that it entirely rewrites in its own words. What youfed into it that came to be around July of last year in 2022, right.

When the models actually got goodenough. Right? So Da Vinci, free from Open ai, that was just eye-opening of, ohmy God, here it is. Right. So in the, in the meantime, while we were waiting,we had this head of AI who was a good friend of ours but he was working atStripe and running Stripe radar, I think at the.

And we'd kept, we kept on feeding him,anonymized data and saying, Hey, what can you give us from a summarizationperspective? He called me up late July and said, Colin, it's there. I'mquitting stripe. Let's go. Right? Wow. The AI can actually write your notes foryou, and it's the dream that we've had since 2019.

It's time to go, right? Yeah. So fullyproductize. Get it up and running and rolling and, and the hard parts aren'twriting the prompts or getting high quality notes out of, let's say, GT four.The hard part is to be able to do it in a cost effective way, right? Mm-hmm. Soif I'm gonna deliver you an hour of notes if you run that all through GT fourtoday it, it'll end up costing you, I dunno, three, $4, right?

Yeah. Yeah. Just per meeting. And thatdoesn't really work. We need to be able to do this at 10 cents per meeting.Right. Right. So there's, there's a whole stack of, of machine learning modelsand pre-processing and post-processing that goes into making all of that workat the right price point where people are like, oh, I want this on all.

Julian: Yeah. And how are youable to, when you say the cost of it, is that cost and time, cost andcomputing, power, cost, and energy, what in particular is the cost that you'rereferring to and how are you able to decrease that cost?  

Colin: Sure. So, cost ofcomputing, right? Yeah. So if you're gonna run An entire transcript through GPTfour, for example.

Yeah. There's a cost for basically eachword that you're sending in, or a token Yeah. Is a simplified way to, to talkabout it. And, 30,000 words, 40,000 words in the meeting. Like, these thingsadd up really fast. Yeah. So how do you, how do you predictively use it? How doyou take fine-tuned LLMs that you've built out yourself from open source modelsand run them on hardware that ends up being 10 times cheaper?

Sure. How do you work with, like, I needthis in real time, or, you know what, we can process the first half of themeeting now and it'll be ready by the time the end of the meeting happens.There's so many different aspects to shaving off and making it more efficient,so, You've got the best product possible at the least cost.

Julian: Yeah. And for thosewho don't know, just add a little bit of context to the, the developmentprocess of ai. I think a lot of people may, assume that companies are buildingit and it, it's very much more of a company-centric kind of thing. But thenthere's a lot of research and institutions that actually does influence andimpact the, the models that are being built and how language is beingcategorized and labeled and all the nuances that people don't understand in therecent.

You, you mentioned something I think theDa Vinci came out with, with, with more. Was it, was it a more scalable model?How's the, the development of ai come to this point? And also what new recentupdates, I know you mentioned some but just to reiterate, them, have been ableto give more individuals access to that type of technology.

Describe, to give a little bit morecontext.  

Colin: So if you, if youbreak one of these models, these large language models that we have, ChatGPTis, is probably the most consumer, like everyone knows ChatGPT, right? I was ina hotel in Copenhagen and, and there were all these people walking through thelobby and at least half were talking about chat.

G p T. It was mind blowing. If you breakthese models apart, you basically think of them as a brain as neurons, right?Yeah. Several years it was okay. Like more neurons. More neurons, more neurons,and we'd see it. Okay. How many parameters do they have? Yeah. And it was,okay, 65 billion, 500 billion.

Right, right. As. That grew as thosemodels grew, all these things called emergent behaviors just started happening.One of them was at summarization. All of a sudden you could feed it a bunch ofcontent and say, you know what? I want this in tweet length. Right? Like,summarize this down. Give me just the importance stuff.

Right? The new wave of models whetherthey're even larger and larger or larger. There's two things that arehappening. One research out of, out of meta was like, you know what? We can getbest in class performance on a smaller model. Like it doesn't need to have 500billion parameters.

We can actually do it on 65 or ahundred. Yeah. We just need better data to train it on. Right. So, so that'sone big piece of, of the. The other is that these models don't activate theentire model anymore, right. They're, they're sparsely trained. So that looklike there's, there's only a thin slice that's actually handling what you'reasking it.

Mm-hmm. We don't have to fire up thewhole thing across 16 TPUs and, and off it goes. Mm-hmm. So the efficiencieshave gone through the roof in the past six months. Open AI slash pricing on oneof their models by 10 times. Right. That's made it super access. People arebuilding it into all of their products.

Yeah. And just trying to figure out theright prompts to use.  

Julian: Yeah. And when youthink, just a little bit, a quick question here. When, when you say you needbetter data, is that better label data? Is that more data points? Is that justa huger pool of data? Is it all the above? What are particular do you mean bybetter?

Colin: Yeah, think of it as,as better better labeled data. Right. So, got it. Yeah. You could say, Hey,this was the instruction I gave it. This is the prompt I gave it, and this isthe output and it's really, really good. Right? Yeah, yeah. What you can do there,there are many different techniques, but you can take a smaller model and say,you know what?

This, this. All of this amazinginformation that came out of G PT four, that's now the, the parent and, and yousmaller model, you're the child and we're going to train you on this is what weput into the big model and this is what came out. And by the way, you can nowadopt that same quality. It's, it's one of like the common things that's putout alpaca, which is Stanford.

Data set that was applied to Meta'smodel, basically was able to train based on a ton of great data out of thelarge model. We can train as a small model to do just the same thing. Yeah. Andwhen you think about models and models size in particular, The larger, themodel typically costs a lot more.

So if you can get the same performanceout of something smaller, it's gonna be a lot more efficient. You can do thingsfor 5 cents or 3 cents instead of 50 cents. Yeah. Per, per hour of content. Andthat really changes the game.  

Julian: Yeah. And what are thechallenges in just like transcribing audio, I mean, that that's something thatlike, that, I think a lot of people think about, which is taking somethingthat's physical in nature, whether it's visual or audio, something that's,organically produced, not synthetic or anything that we input, but then takingthat and extracting that information and making it communicable and, and, Iswhat, what are the challenges about building around that?

Like do you have to continue to trainyour own models or are you just implementing like, newer updates and newer,sophisticated materials? What goes into that process that a lot of people justaren't aware of?  

Colin: So, transcription isa, is. Is a classic problem, right? Because you can say, look, there's, we knowwhat it's supposed to be.

We have, there's, there's a word everrate is what we call it for a given transcription. And for a model, and you'relike, well, it gets 92% of the words correct, or 94% of the words, right? Yeah.If you're still operating in a world where it's like, I need a transcript forthis. It's, it's imperfect.

The cool thing about going and saying,Hey, I actually am gonna push this transcript even though it's imperfectthrough a large language model. To summarize or get information about it. Largelanguage models are extremely forgiving. Yeah. So you don't need a perfecttranscript anymore. We, in Supernormal we still keep the transcript around ifyou want to see it, but the, the rate at which people click on that tab ontheir meeting notes to go see a full transcript is so low.

It's like less than a percent. Right.Uh, It was like, yeah, cool. It's there. It's a mountain of. Whereas you justkind of want your notes.  

Julian: Yeah. Yeah. And tellus a little bit more about Supernormal. Not only, you were, you were, it soundslike, in development for quite some time. You launched the product and, and youwith a lot of updates and things that change in July, 2020.

What's been exciting about, the, therecent traction you've seen, but also what are you particularly excited aboutin, in the near future? How many people are you working with? How manycustomers, how many clients, and how many are you looking to grow and expand toin the next few years?  

Colin: So we have a massivefree tier of users who are just using us on, on their meetings day in, day outon Zoom and Teams and Google Meet.

And then on the paying side, 500companies. Yeah. And that number is growing between somewhere between 30 and50% every month, right? Yeah. So it's we're, we're doubling at a, at a prettyfun rate. Yeah. Uh, In terms of time and from. From a product perspective,Supernormal is just all about never have to take your notes.

Ever. Yeah. Right. Like we're gonna doit for you. You can focus in on the meeting, focus on your conversation, andthe moment you get out of your meeting, within two seconds you've got the fullset of notes, action items, decisions, summaries. If it's an interview how didyou know, and I'm doing the interview, how did the person answer yourquestions?

Yeah. And that's just the tip of theiceberg, right? Yeah. Because you end. All of this text that is super usablefor the next set of, of large language model interpretation. So if you have anaction item that happens to say, oh, I'll send a, I'll send a follow up email.Great. Yeah, we wrote the follow up email for you, right?

Julian: Yeah.  

Colin: Not just with thecontext of the call. We can pull in things from your chat or from your emailand create a, a much larger. Awareness of what you're working on. Yeah. Therelationships of the people around you and, and pull that into a large languagemodel and say, Hey, really craft the, the correct response given all of thisinformation about this, this communication thread or what's happening in mywork.

Yeah. So that's exciting because thenthose. That content, whether it's a follow up email or you need this stuff togo into your CRM cuz you're doing a sales call, right? Or your applicanttracking system for hr. It's just the tip of the iceberg. I, I consider.Meeting data to be this new like first class data object in an organization oran enterprise that we've never had before.

Yeah. Right. Like we've only ever had itdisparately through someone writing a Google Doc while in a meeting. Yeah. Andhoping that they catch all of it. Or writing it down with pen and paper. Butthe shift in kind of time saved that we're seeing for people of, Hey, not onlydo you not have to write your notes again, We've got the action items, they gointo the right system.

Your tasks end up where they belong.Yeah. And then at the end of the week, we can kind of give you a great summaryof, Hey, these are all your tasks this week, or these are your conversationsthis week, or this is the status of the project that you talked about fourtimes. Right?  

Julian: Yeah. It's incredibleto think about the, the amount of, not only time you'd save, but.

The, the level of insight that you cankind of carry through those conversations and see, key points that were talkedabout and, and really compare between two meetings, key points that were talkedabout, action items, what was successful, what was not successful. You become alittle bit more intelligent in the way you're not only, going through thesemeetings, but also training your teams and how that can really impact theoverall organization.

And, thinking about externally orinternally, what are some of the biggest challenges that Supernormal facestoday?  

Colin: So scaling the team,right. So we, we were three, we're now 13. I think we'll be 20 to 25 in thenext quarter. Which means that we're in interviews with, with candidatesprobably about 10 a day.

Right. Yeah. Being really thoughtfulabout who we're bringing in and, and why making sure that they're a strong fit.That's always like one of the hardest pieces to do. We've, the, the 13 that wehave now, it's like we've kind of made the core nucleus but we're fullydistributed. Right, right.

Yeah. So most of the team members I'venever met in person until next week. Yeah. We we're gonna do a team offsiteflying everyone in from, from Brazil and Seattle and Stockholm and New Yorkand, and San Diego. I mean, we're, we're super spread out. Yeah. Flying intoMyorca for three weeks of just us working from one big house.

And, and really to, from, from myperspective, you really wanna build empathy, especially in that core unit thatyou have empathy for each other and the work that we're doing, but also empathyfor customers and get people's different perspectives. So that's what it's allabout and it is a big challenge.

Right? Yeah. Distributed is if, if youwere to ask me, Hey, it's 2024, you're gonna start a company are you gonna doit distributed or are you gonna all sit in the same place? I would say 10 timesout of 10 in the same place. Yeah. Right. Yeah. Distributed work when you'retrying to go from zero to one is so difficult.

Yeah. My co-founder, Fabian and I, he'sliving in New York and I'm in in Stockholm for the past three years have beenflying to each other for, yeah, one week every month. And that was where we'regetting our best work done. It was great cuz those other three weeks, it wasjust, Focused work mode.

Sure. But all of the brainstorming andcreativity, a lot of that comes out of just the in-person work, which iscritical for when you're trying to scale something from, from nothing tocustomers are excited or now they're paying customers, now they're expanding.Right. But, but now that we've reached kind of that threshold after, a long,hard grind.

We're much more well situated now forbeing a distributor team and getting talent from around the world.  

Julian: Yeah, and, and ifeverything goes well, what's the long term vision for Supernormal?  

Colin: So we're the, theoverall goal is that the future of work is one in which these really powerfuldevices that sit on our, on our desks, we're only using like 2% of them.

Right? Yeah. Like, especially like, myMacBook Pro with, with its neural engine, there's so much capacity. In it toincrease my time, save to help me execute the work that I'm trying to do toimprove my tone and how I communicate. Yeah, like that's one of those piecesthat people miss a lot, which is as.

ChatGPT starts running our, ourworkforce and, and help us. It's actually really gonna make work a lot morepleasant. Right? Yeah. Like getting content from that system or, assistedcontent is amazing, right? Yeah. It's usually in a really nice, kind, helpfultone. Yeah. Well-written that, I might have to spend an extra 10 minutes on ifI were to do it myself.

Right? So, so for us, just gettingstarted there, Thousands of companies and millions of people that can. 3, 4, 5hours a week right now of their work assisted and done for them. And, and yeah,I'm excited for the future and, and see how that scales across organizationsand different verticals and in different markets.

Julian: Yeah, I love this nextsection I call my founder faq. So I'm gonna ask you some rapid fire questionsand we'll see where we get. Cool. First question is thinking about, the amountof time saved and amount of capabilities you have with these tools, what do yousee just currently in terms of outcomes and, and what do you see in terms ofthe long term success for individuals who use tools like this being that they.

Capture a lot more conversations, butthen really consolidate the information in such a easy, simple way usingsomething like Supernormal. What, what have you seen outcome wise, any, anynumbers kind of speak to you in terms of whether it's increased productivityor, or increased understanding of, of their meetings.

What have you seen that's particularlyexciting about what your technology's been able to help people do and and, andwhat are some future predictions that, that you think would be exciting forother tools and other things you can inc.

Colin: So right now theproductivity gains are 30 to 40% on a given workflow that you're doing.

Wow. Right? Yeah. And it's like, that'sbig. That's really, really big. Yeah. I expect that before the end of the yearthat we'll be closer to 60 to 70% on some of the common workflows that you,that you have to go through. It's like, we have an email thread like schedulingthe Behind, Company, Lines.

Right. It was like, yep. Like email.Great. Like calendar invite should be done for you or suggest. All right. Likeyour agenda for it should just be pulled together. There's so much that can bedone as long as, you have the right context to, to feed into these systems and,and building that context is everything.

Yeah. So I think, I think we'll actuallystart seeing 70 to 75% productivity gains, right? Yeah. Which is really justlike when you really kinda like dial into it, it's just time. Right. Yeah. Soit's like, can you not quite get to the four hour work week? But yeah, what'sthat gonna look? Right.

Yeah. Where you effectively did what youwould do in two or three days within like a six hour span. It's, that's gonnabe where, where the puck is going, so to speak.

Julian: Yeah. One thing, onething that always kind of is interesting because I think. Previously in thelast, I don't know, it was five years, was kind of popularized around the ideaof consent for things that track information or, or consolidate informationfrom live inputs, whether it's conversations.

I think there, there was a lot ofrecorded meetings, had the. Disclaimer right before, just to alert people thatthere was a recording going on, how do you build around consent and, and hasthat been changing as more tools are becoming hyper integrated and, and presentduring, in real time? During the things that you're doing with thisconversation, the recordings and things like that, do you, is that a hugeconsideration that you have to, to think about and, and what are the ways thatyou're able.

I don't wanna not work around it, but,but, build around the, the compliance per se of consent and things likethat.  

Colin: Yeah, yeah. Trust iseverything, right? Yeah. It's like you need to not only keep data private andsecure, but the, the trust of the experience as well. Yeah. Recording video andaudio, it has legal implications, right?

Yeah. Like there needs to be two-wayconsent depending on what jurisdiction you're in. Yeah. Taking notes is, is aninteresting one because it's considered a derivative of the captions that arealready there. Right, right. So everyone's kind of already consented into,well, yeah, there's captions and, oh, do you mind if I, I automatically takenotes for this meeting?

We, we find that customers don't reallyhave an issue with. Providing notes for everyone in the meeting. It also kindof changed the way we thought about the product because when we first rolledout, it was your meeting notes. Right? As opposed to, you know what, thesenotes belong to the people in the meeting.

Yeah. So like, if you're in the meeting,you should probably have access to them. And so that small tweaker change hasactually made us grow extremely fast. Right. Right. 90, 95% of the meetingattendees open up the notes afterwards, and it could be 40 people in thismeeting. Right. So, so that's one aspect of it, and it's, it's really importantand it's, and it's one that I'm, I'm really lucky, my, my co-founder Fabianhaving run core product for, for GitHub has such a powerful mind for not onlylike scoping privacy and what does that look like, but how do we give peoplethe right control?

And the capacity to say, Hey, you knowwhat, like I want to introduce notes in this environment. Is everyone okay withthat?  

Julian: Yeah. That's amazingthat you're able to, I didn't even think about the, the, not only the abilityto expand and get us other customers using, Supernormal, but the ability togive people, everyone in the meeting notes and there's not one person that.

Gate keeps that information, that'sincredible to hear about. And honestly, such a hack. I'm assuming those 90% ofpeople, there's gonna be a huge conversion rate on that because it's simple,it's easy, you, you integrate it within whatever you have, and you're able toencapsulate, the notes in a way that's easily distributed.

That's brilliant in terms of how it'sgonna gain interaction with other customers. I, I, I can't, I can't not stickon that point, but I'm, I'll move on because I'm geeking out on, on what thatkind of entails. More personally as a founder, what's particularly hard aboutyour job?  

Colin: Hmm. There are soamazing co-founder there's one of those truths that happens, which is like, youboth aren't gonna be a hundred percent all the time, right?

Yeah. So one of us might only be 40% or50%. We might be having personal things going on in our lives, might just be inone of those really tough stretches of work and finding the, like the, theshared resilience and the stubbornness to just keep going, right? Yeah. Rightbefore we got to this inflection point with this product exploding, I think wehad a, like a month and a half of runway left.

Right? Wow. Like, yeah, we were down tothe wire and really just working super, super hard and it was, it was likeworking in a, in a dark cave almost. For, for like those last six monthsleading up to this massive inflection. Yeah. So it's, it's it's hard and, andthose founders that are getting down to like, okay, we only have two monthsrunway left.

I was like, that's plenty of time. Yeah.Right. Yeah. And of course there's some survivor bias in that, but being ableto just continue regardless is, is really important. And having, like ifyou're, if you're a founder or if you're a co-founder having that trust inyour, your co-founder that it's like, Hey, look like you're not, you're not ahundred percent.

That's fine. I've got this. I'll carryit. I'll pitch in wherever I can. Yeah. That sort of relationship is one ofthose kind of key pieces.  

Julian: Yeah. Yeah. 100%. And,thinking about, for, for other founders out there what, what's something thatyou know now but you wish you knew earlier in your, in your career as anentrepreneur?

Colin: Hmm. Being technical,even if, even if you're not an engineer, being technical is. So incrediblyimportant when I look at the founders around me that have succeeded or aresucceeding in scaling up to series A and series B and and beyond, if they'regoing the BC path or they don't have to because they're now customer funded.

Almost all of them have some level oftechnical expertise that they're bringing to the table. That's reallyimportant, right? Yeah. So if you're, if you're building out a, a tech company,a software company and having a non-technical founder at, at the. It's, it's aeven harder challenge Yeah. In a very tricky market right now, right?

Yeah. So, For me, it was just get, getinto it in like 20 15, 20 16 of just I want to build things, so teach myselfand, and off I go. But the most important aspect of all of that is if you aregoing down that path, find a mentor. Sure. Right. My mentor ended up being myco-founder. Right. Because he was just so, such a powerful engineer and, and sothoughtful, but also willing to spend the time and do code review and just helpme work through problems.

Yeah. We had a. I had a side projectthat was scaling really, really quickly in 2016, just because our first kind ofclient off of products was the spn and we were a social. Analytics tool. So onday one, the servers just all fell over, right? Because I didn't build it theright way, like it was a mess, right?

And then well, sure enough, people willcome help save a sinking ship and and you'll learn a lot along the way.  

Julian: Yeah, it's incredibleto hear and, and I I love this next question cuz it, it kind of plays in thesame vein, but I love how founders extract knowledge out of anything that theyingest. Whether it was early in your career or now, what books or people haveinfluenced you the most?

Colin: Mm. There's a smallcollection of founders and tech people that have formed a, a bit of a small,like little society here in Stockholm. 10 people dinners once or twice a month.And the, the really, like the, the signal of membership is like, how curiousare you? Yeah. Right. Like it's curious people only who are just consumingcontent day in day.

Seeing something and be like, oh, I wantto try that. I'm gonna code that. Right. And and that's been the biggest sourceYeah. Of just being able to connect with these people and say, Hey, like, Ifound my little like culture of, of and, and like niche of people. Yeah. Superimportant, right? Yeah. Because then the, the sharing of knowledge, we comefrom all different sides of tech, whether it's a head of AI for a batterycompany or someone who sold a company for 500 million last year.

Like just a random mix. Yeah. And.Super, super helpful because every single person in, in that group is justwildly curious.  

Julian: Yeah. Yeah. Thecommunity aspect is, is huge. It's, I mean, founders talk about it all thetime. Finding not only mentors, but people who are, not competitors or, ormaybe even not in the same field, but are equally in the pursuit of somethingthat they're passionate about or ambitious, or see a problem that they wanna.

Because they get it, be because they're,they're so, they're so entrenched in a similar either philosophy or, havesimilar habits of work and, and my, my business partner always tells me it'slike, find people that inspire you to be better. Who, who You're a little bitmaybe even co not, not conscious of what is it intimidated by, it's like,because that that cohesive relationship and, and that, that, playfulcompetition is, is so, so impactful in, in a positive.

Last little bit is, is I definitelywanna make sure we, we covered everything and before we give a chance for, foryou to give us your plugs and where to find you and all that, is there anythingI didn't ask you that I should have or anything that you wanted to cover thatwe didn't? Anything that we missed at all?

Colin: Yeah. One, one piecethat I find really important in, in founders and also in culture is this inherentsense of competitiveness, right? Yeah. Coming out of Boston, which is a sportscity, right? It a family where my father was an athletics coach, a runningcoach for Olympians, all the way down through high school athletes.

I was the youngest of four. Like just ahyper-competitive environment that moving to Sweden and Stockholm inparticular, it's like, you can compete, but you have to hide it. You have to,like, it just has to be there. Right? It's kinda like the, the, the mantra oflike, oh, you did a, a really great workout at the gym.

You just don't tell anybody. Like, youjust do it, right. It's just a stoic mode. But that competitiveness is, is oneof those pieces. I think doesn't get recognized or even found, uh mm-hmm.Enough. So it's it's one of those keys.

Julian: Yeah. Yeah. A hundredpercent. And Colin, it's been such a pleasure chatting with you last little bitis where can we find you?

Where can we support you, not only as afounder, but Supernormal as a product? Give us your websites, your LinkedIns,your Twitters, anywhere and everywhere that we can be a fan and supporter andmaybe even a user of, of what you're doing.

Colin: Yeah, if you if youhappen to be in Google Meet or Zoom or teams calls anytime Supernormal.com isthere for you, right?

So the, the dream of never having totake notes again became a ra, a reality in, October of last year. Andregardless if, if you're in in sales or you're in recruiting, Anything, right?Yeah. Marketing, design, customer success. We're in meetings all day long,right? Like this happens and this is sticking in the culture and like we'restill kind of fighting the zoom fatigue, but it's there.

Yeah. Right? So even if you're only inthree meetings a day, or four meetings a day just the fact that you're notgonna have to spend 10 minutes either right after the meeting, writing outnotes, or at the end of the day or four days later, trying to remember whatthat second meeting on that day was all about.

Yeah. Supernormal is just getting startedand making all of that just cohesive and in one place for you.  

Julian: Incredible Coman. So,such a pleasure. Not only learning about your background and your experience,building products, building interesting products that people will, will useand, and be retained in, in kind of that customer engagement philosophy.

But also what you're working on isSupernormal. Helping give back time and give back focus. And honestly, if Ithink about it, have a better meeting experience, not having to worry so muchabout critical information because that's taken care of you. You work on theflow of the conversation and build those relat.

I'm excited to see how much better thetechnology gets and and implemented in my team, and I'm hopefully the audiencewill do so as well. But thank you so much for being on the show today. I hopeyou enjoyed yourself and I appreciate you being on Behind, Company Lines.  

Colin: Thanks Julian.Appreciate it.  

Julian: Of course.

‍