What's Up with Tech?
Tech Transformation with Evan Kirstel: A podcast exploring the latest trends and innovations in the tech industry, and how businesses can leverage them for growth, diving into the world of B2B, discussing strategies, trends, and sharing insights from industry leaders!
With over three decades in telecom and IT, I've mastered the art of transforming social media into a dynamic platform for audience engagement, community building, and establishing thought leadership. My approach isn't about personal brand promotion but about delivering educational and informative content to cultivate a sustainable, long-term business presence. I am the leading content creator in areas like Enterprise AI, UCaaS, CPaaS, CCaaS, Cloud, Telecom, 5G and more!
What's Up with Tech?
How AI Transcription Is Rewriting Journalism And Media Workflows
Interested in being a guest? Email us at admin@evankirstel.com
A journalist screams, “I’m wasting four hours a day transcribing,” and a product is born. We sit down with Lasse Finderup, CEO of Good Tape to unpack how a newsroom pain point turned into a privacy-first platform used by millions—and why saying no to feature bloat matters more than chasing every shiny AI trick.
We trace the spin-out origin story and the “instant product-market fit” that came from building for colleagues who needed reliable, fast transcripts yesterday. Lasse explains the decision to never train on user data and to host models in-house, trading flashy add-ons for deep security, ISO-grade compliance, and trust. We explore global AI adoption gaps—from Denmark’s “I’ll just ChatGPT this” culture to regions where automated speech-to-text still feels like magic—and why context matters when you’re designing tools for journalists, podcasters, and creators handling sensitive sources.
From a tech perspective, we dive into an open-source stack centered on Whisper V3 Large and the heavy lifting around the model: optimization, infrastructure, and the real costs of self-hosting LLMs. Lasse lays out a sharp distinction between “record-everything meetings” tools and workflows where the transcript is the output itself. That sets the stage for Good Tape’s next big leap: an “artificial memory” that surfaces relevant past notes at the right moment, with user-controlled reminders that feel helpful, not invasive. We also touch on multilingual transcription’s surge across contact centers and newsrooms, market consolidation on the horizon, and founder advice: build for real needs, not just because AI makes it possible.
If you care about accuracy, confidentiality, and simple tools that get out of your way, this conversation will sharpen how you evaluate transcription tech and where the industry is heading. Subscribe, share with a teammate who fights transcripts, and leave a quick review to help more builders and storytellers find the show.
More at https://linktr.ee/EvanKirstel
Hey everybody, really excited today to unpack how AI-powered transcription is reshaping everything from journalism to media and the future of content creation. Lasse, how are you?
SPEAKER_01:I am good. Thank you so much.
SPEAKER_00:Thanks for being here. I'm well. Thanks so much. From Denmark, one of my favorite destinations. But let's start from the beginning. What's the origin story behind Good Tape? What problem were you trying to solve back in the day?
SPEAKER_01:Yeah, so Good Tape has a bit of an untraditional backstory. We're originally what's called a spin-out. So we were spun out of a media company called Sitman. Which is also why we sort of almost had what I would say call instant product market fit. Because we built something for the people that sat basically next to us, for the journalists. They were just basically screaming at everyone, saying, spending four hours every day just being a robot, transcribing, turning audio into text, it sucks. And whenever someone says something like that, it's like, okay, there's an issue here that we then made a pretty barebone solution to, like an internal folder where you could drag and drop. And then everyone went crazy and said it was like magic. And then it sort of just spread very uh organically when we we made like a live version that people could just use for free. And then we got, I think it was like 18,000 users just word of mouth in in a month and a half. Um and really saw, okay, something something's here. Um and then we went a little bit deeper and made a little bit nicer, made the UI a little bit more friendly rather than just like a very barebone uh piece of software. Uh and then instant product market fit and have just been growing ever since. And today we have about two and a half million users. Uh yeah, so doing well.
SPEAKER_00:Yeah, indeed. And when you look at the media landscape today, and I consider myself a little media company, uh almost uh like a two-man band, but the big picture, how is AI changing the way that journalists and podcasters and creators like myself are working, big picture-wise?
SPEAKER_01:I think I think if you had asked me two years ago, I would have said something different. Um because back then it was very much a how can AI eliminate manual work or boring manual work and and and do something for you as a journalist. And now it's more of how much can AI do, or can we add another layer? It's already at that stage. Um but I think there's still a lot of it, it depends on what market you're looking at. If if I'm looking at Denmark, AI adoption is just really high. Everyone sort of says, oh, I'll just chat GBT this or or whatever. But if we're if we're looking at at other markets like Indonesia or the Philippines uh or Taiwan, um there's just not as adoption. So just a speech to text being automated by AI seems like magic still. Um and it's just sometimes remembering that AI is everywhere in some countries, but in not in not in all countries. Um so just really keeping that that thought in the back of our minds when we develop things. Um that we're not necessarily competing with ChatGPT everywhere uh to some places.
SPEAKER_00:Well said. And Good Tape is known for not training its models on user data. That's very European of you. Well done. Why was that such a core value right from the beginning?
SPEAKER_01:Uh because it's the first question everyone asks. Um and if you say you train on data, they just click the red button and say they don't even say goodbye. Uh so it's like if uh it's it's our it's our access ticket to just get in the door, is to say we don't touch your data. Um I think that's a big thing at the moment, and I think we're gonna be a bit clearer about it in the in the coming months of how many of our competitors train on data without telling you. It's a bit of a it's a topic very, very top of mind for me. Um that it's yeah, it it's a bit I I call it like it's modern pickpocketing. Um to just take a little bit of data without you knowing it, but it's cheap and it's fast, but you're paying something else.
SPEAKER_00:Um yeah. Yeah, really well said. So there's a lot of layers to that confidentiality, you know, is make or break in many industries, professions, journalism, but also in the enterprise corporate world. So how do you how do you ensure that sensitive content actually stays protected?
SPEAKER_01:So we've from the beginning, everything we we do is built in-house. So we don't have just like an API or an endpoint somewhere like like a lot of of the like the ones popping up do. Uh so everything is built and hosted in-house. And everything I'm I'm not afraid to say that we don't have as many features as some of our competitors do. Uh, but it's because we've allocated all of our resources to go and build the core part of it really well and really securely, and not just oh, now you can do summaries, and then oh, actually that's because it's an integration to ChatGPT, and then you upload your entire transcription. Um so everything is very securely, and then we'll be will be ISO certified hopefully before the end of the year, uh, which is the worst thing I've ever done in my life. Um But it's just really saying like compliance and and doing things being the nice guys is a pretty big selling point when everyone, a lot of the other guys are not being very nice. Um yeah.
SPEAKER_00:Yeah, say the least. And what what do you think most people misunderstand users? What do they misunderstand about things like ethics, accuracy, oversight? You mentioned when it comes to transcription. We just look at it, it's magical, it works, and we don't think, or many of us don't think more about it, but what are we not thinking about?
SPEAKER_01:Um a misconception that I really saw a lot in the beginning, but I'm also seeing it less and less, is that there's just this raw term of AI. So we heard a lot, especially in the beginning, where users would say, no, I'm actually not allowed to use Chat GPT in my work. And we're like, we're not chat GPT just because we're AI. So there's a lot of there's degrees of information uh to understand the like just for example, the that a lot of the enterprise houses have something where they say you can't use any generative AI. And then we're not generative, but the users don't really have that level of understanding the differences. Um so so having to get past that barrier of understanding that also requires that they trust you when you say no, we're not that, that they don't just think you're trying to sell them something. Um so really the knowledge part of AI, which will be it's improving as we speak. Uh, it's way different now than it was two years ago, but I still I still hear it quite often.
SPEAKER_00:Fantastic. And you have, I think you said almost three million users. Um, what do you think earned that level of trust so quickly beyond being nice guys from a nice country? Uh what what else?
SPEAKER_01:I I think it is our approach to things, and it's a very we we from the beginning decided that there's a cost to all features of simplicity. Like if you if you check out our our platform, it's so simple. Like it's very neat and and it feels very approachable and and risk-free to just start using it. Um and then also just because it's such a clear need. Like I don't want to spend four hours doing this if something can do it for me in one minute, and it's ten dollars. Uh it's it's a pretty easy sell. Um so it just spreads like wildfire if you're actually solving a problem.
SPEAKER_00:Well said, yeah, very good. And when you look at your competitors, and we won't get into specifics, but are there particular innovations or features, functionality you think that really set you apart that you're proud of?
SPEAKER_01:There will be. Um there is, of course, now. I think I'm really proud of us hosting everything ourselves and the accuracy of the core value we serve. But the roadmap is really taking us down something where we I just came out of a meeting now where it's it's almost scary the level of user feedback that we have to get to to really nail it, because it's like uh it's a thing that that it's not comparable to anything. Um it's sort of like we're without going too much into it, but we're gonna try and and and make like an artificial memory that will remind you of things that you might have forgotten in some of your previous work, um, which is very difficult because you're trying to emulate memory. And my memory is very different from your memory of maybe you're reminded once every 10 seconds of something, and maybe I'm tired and only want to be reminded once every hour. Um, so yeah, it's a it's a very, very interesting um direction to take, which will be much better than just transcription.
SPEAKER_00:And and multilingual transcription is exploding. You know a lot about that in Europe, but beyond that, I mean, global use cases, so many diverse industry needs for multilingual transcription. Uh in the US, it's in contact centers, at call centers, and you know, we have a lot of different languages here as well. Um how do you think about that opportunity?
SPEAKER_01:It makes my job both more difficult but also more fun. Um, because it really becomes a job of saying no to a lot of things. Um if there's a large uh telecom enterprise potential client that says we'll convert if you build this feature. But then if we decided no, we're gonna be for journalists, that will not be relevant at all. Then there's a cost again to implementing something. So it being such a generic technology that can that can sort of solve a very homogenous problem across a lot of verticals is sort of a a challenge in terms of really sticking to your strategy and and really saying we're gonna be something for these people. Um, because you can be something for a lot of people, but you just have to decide who you want to be something for. Um which is a it's on my table.
SPEAKER_00:Lots of trade-offs for sure. Yeah. Uh can you give us a peek into your tech stack, sort of behind the scenes? How do you uh does it work? Uh maybe which LLMs do you use, and uh what kind of infrastructure do you need for transcription on such a scale? Yeah.
SPEAKER_01:Yeah, I can give a bit of a sneak peek. Um so we are running open source models, uh, primarily WISPA V3 Large. And then what we've focused on is just optimizing everything that's around the model. Um, because we saw we actually saw a larger yield on that rather than investing resources into trying to fine-tune something. So a lot of companies pouring hundreds of millions of dollars into fine ESR models. Um But we've built we we we usually say it in-house like we focused on building the PlayStation, and then we're ready to switch out the PlayStation game and being the model. And really sort of relying on the open source community, and then at some point, of course, we'll give back. Um, but we've really come far from optimizing everything that's around the model. Um and then we host uh I think it's Lama B3 or something at the moment, um, to really also host everything ourselves. To host a large language model is quite an undertaking when you're not that big of a company. Um I think a lot of SaaS entrepreneurs or founders underestimate the cost of AI when you then say I'm a SaaS founder. They say, Oh, fucking your gross margin must be 99%. But but AI does actually have a cost, um, which was a little bit of a surprise uh when we came to to hosting a last language mala cells.
SPEAKER_00:Interesting. Yeah, I'm sure the economics are fascinating. Um so zooming out, where do you see the this whole industry headed over the next couple of years? Transcription that is. Uh, what are some of the big picture trends you're looking at? And maybe just a sneak peek into your future. Like what what ideas are you noodling on that you can share?
SPEAKER_01:Yeah. Um I think there's gonna be a big difference in who you're building for. Uh I see a lot of people building for the everyday user. Um, you have a lot of meeting integration agents that'll just transcribe and summarize everything. Um, where then we've really nailed down who we're we're something for and we're something for the people that record something to use it. Um I'm interviewing someone to use it. You're we're talking here because I what I'm saying now will be part of the output. It's part of what value I'm trying to create, which is very different to oh, I just want a summary of my meeting because I maybe want to remember it at some point. Um, so there's from my point of view, there's two very clear distinctions of what is it and what is the need that you're building to solve for. Um and then I think structurally, I think we're gonna see a lot of roll-ups. I think there's gonna be a lot of money. Um, there's so many popping up everywhere. So I think at some point it'll just be a part uh it'll just be who raises the most money and goes out and buys the rest, and which is gonna be interesting to see. And I think it's gonna be soon. I think it's gonna be within two, three years.
SPEAKER_00:Yeah, fascinating. Um there are lots of entrepreneurs like yourself diving into this marketplace now with amazing apps and services, and um, it's sort of the Wild West uh out here in the US and around the world. Any advice for founders and entrepreneurs who would love almost three million users like like yourself? How any what what advice would you would you give them if they're just kind of starting out now?
SPEAKER_01:I think uh a mistake that we're also making often, uh, but we're catching ourselves in making it, is that with AI you can build so many cool things, but just because you can build it doesn't mean that there's a need for it. Um so we're getting caught up often and getting very excited about all the things we can do, to the extent that sometimes we're forgetting of why we would do it. Um and I see a lot of AI startups pop up where it's like, wow, okay, you can do that. Why? Um it's cool. And I can see that the technology is incredible and it'll be even more incredible in a year, but but why? Who would who would buy it and solving? Um so just the the idea of technology, the pace of technology has sort of overtaken the pace of problems, if that makes sense. That it's just there's so many things it can do, but it doesn't mean that there are equally more problems to solve, um, which is a trap I see a lot of people fall into.
SPEAKER_00:Uh so yeah, that's a lot of minefields out there. Um I'm heading into CES in the next few weeks, which is you know the big consumer pow. And so far I've seen at least four or five pitches for transcription devices, you know, wearables and tablets and uh pins and accessories, all kinds of things. I mean, do you think there's a future beyond transcription in just using our phones and and laptops? Are you looking at devices or is this just kind of a uh you know, a diversion from them?
SPEAKER_01:I think it's interesting because I'm I've sort of allowed myself to believe in the future that every meeting in all companies will be transcribed at some point. Um if I were to do a Black Mirror episode, it would be that all all tables in all meeting rooms had a microphone in the middle. Um for now we've built the we've built a recorder app because everyone has a phone anyway. No need to develop a new uh piece of hardware that I then have to drag around. So it's just like for us, it's just I I hold down the the power action button on my phone and then it starts recording and transcribes. And I just put it on the middle of the table when I have a meeting. Very easy. So I don't see a need to develop hardware. Everyone has the hardware needed anyway. Um so just build an app and then of course do meeting integrations and all that.
SPEAKER_00:Yeah, fun. Yeah, it sounds like that's that's for sure. Although Facebook just bought a a wearable device, so who knows what they'll do with that for transcription. Uh, but you know, moving forward, what are you you know excited about as we we head into the new year? Any any travel, any any plans, any news on the horizon we should uh listen out for?
SPEAKER_01:Um I would say the product announcement, I hope it's gonna be Q1, probably end of Q1, where we'll we'll do a beta of what we call artificial memory for now. It's name work in progress. Um we'll do that hopefully end of Q1, which will be everyone gets goosebumps when we talk about it in-house at least. Um really weird.
SPEAKER_00:Getting goosebumps now, just the name. I hope you trademark that because that's amazing. I'll go down. Whatever it is I need in artificial memory. Yeah, right?
SPEAKER_01:Okay. We have the industry internally where there's at least one colleague who's like, I'm gonna record everything my wife said. So keep changing their minds.
SPEAKER_00:Okay, and vice versa. Let's let's let's be uh equal opportunity offenders. You know, wives can record every everything that has been. Oh, yeah. Well, of course. You're you're very egalitarian in Denmark, I know. But thanks so much for joining. Really fun, interesting conversation. Thank you. Seems like it's still early days. So much more opportunity for you and this industry.
SPEAKER_01:Yeah, definitely. And I want to keep hold of that feeling and it's new early days.
SPEAKER_00:Thanks so much, and thanks everyone for listening and and watching this episode. Also check out our tech uhimpact.pv show now on Bloomberg and Fox Business. Thanks, everyone. Thanks for asking. Thank you. Bye bye.