# AI-Driven Product Strategy and Operational Scaling

**Podcast:** Stories Connecting Dots with Markus Andrezak
**Published:** 2026-07-02

## Transcript

Hello everyone here again at Stories Connecting Dots, today with my guest Lars Giere.
Lars ken I since I don't know how many years, but we never worked together.
We worked together together, we knew each other well.
Hello Lars, welcome to you.
Maybe you first tell me who you are and what you do Lars makes to Lars?
Hey Marcus, just a simple question to the beginning.
Yes, exactly.
What I mean, I can tell you a little bit about what my Werdegang is.
I've actually always really hard to solve problems with computers.
So I was very nice.
I started to program with 12, and then with 15 I built my first small company.
There was so Content Management and...
Shop-Systeme für andere gebaut und Community-Websites für Computer.
Moment, Moment, Moment.
Das geht mir schon zu weit.
Mit 15, erzähl mehr.
Naja, es war halt einfach so, ich komme aus einer kleinen Stadt, die Aspeine bei Hannover.
Oh ja.
Man hat halt nicht so viel zu tun, wenn man jung ist in so einer Stadt.
Und genau, da habe ich halt mit einem Freund zusammen angefangen, halt Sachen zu entwickeln.
Und dann wussten wir einfach nicht, was wir machen sollen.
And then someone said, then make things for others.
Maybe they give you even money for them.
And that was so.
And then we were both computer games and my brother also.
And then we built community websites for computer games.
And that were the old days as Online-Werbung, so as Print-Werbung.
That's how Double Click came just so on, but was also paid for print-calling.
And it was then quite a bit easier than just another job as a student to search for it and still a little bit of money to have.
And that's what we then just did.
Yeah, cool.
But you probably have already mentioned that you have already mentioned that you have already already the first break with.
That with the advertising, it's going to be different.
Yeah, that's not working anymore.
I think 2001 is that.
2000, 2001 is that then much harder.
And everyone realized that it works somehow differently from the dynamic.
And an impression online is perhaps not so much worth as an impression in a print medium.
And then everything turned out very, very strongly.
And then it wasn't really easy to take a moment to take a moment with it.
But yeah, I had that first already with it.
Also ich habe dich absichtlich unterbrochen, weil ich glaube, das sind so Dinge, wenn man das früh in der Karriere erlebt hat, dass sozusagen Sachen nicht taken for granted sein können, dann macht das was mit einem.
Ich behaupte ja, weil wir mit diesem riesen Suchmaschinenzeug in den 90ern rumgefummelt haben und so und wir gar nicht wussten, was das wird.
Und dann hat das funktioniert, aber wieder nicht so, wie wir wollten, aber es war doch cool und so.
That's what it's so very important.
When 10 years later in our job comes, it's like, oh, everything goes always up, there's no crisis, that's all relatively reasonable, that's always going to go forward, and the time goes in one direction, and the success goes in one direction, then it's the next time super weird, because it can also go down, people will get this new tool, what can that be?
I don't know, I must have to try it.
That's just a complete...
and other experiences that you can do now, what can be a bit of a overwhelming FOMO and so, and I think that people like you the first thing to do, in an way, it was not existential.
It was nice and so.
You could see the world as a play.
What you probably do now do, we'll see later.
But that's just like, I think, when you're in a world born and you're like, oh, it's the Rezepts and the Marty Cagan-Bewers, I make that a little bit and I make this trio and everything will be good and everything will be good.
Why does my chef do that not?
It's just like a different world, I think.
Yeah, exactly.
Anyway, also mit 15 warst du so ein Entrepreneur, hast dir die Hörner abgestoßen und wie ging es dann weiter?
Genau, habe dann studiert, auch sehr solide dann eher, also Mathe und Informatik mit Physik zusammen.
Mein Kombi.
Ja, es ist halt das, was mir liegt.
Also ich habe erst Informatik nur gestartet und habe dann festgestellt, dass mir Mathe mehr Spaß macht tatsächlich.
It was for me was for the opposite.
I started with Maths because I wanted to be theoretical and I was supposed to be a good Mathematiker.
Then it was so painfully, I said, then you take Informatik to.
Yeah, it's super exciting.
Yeah, exactly.
In the way, it was exactly the opposite.
That's really funny.
And then I was just a little bit in the research.
That was nothing for me.
That made me not so much fun, because everything was...
I was very much more than I was thinking about it.
And then I went back to the online world, I was really out of it.
The last time I built things, there was no Google Analytics.
We were actually for the Kunden, their websites were all over-the-sites, and then all of a sudden was all of the professionalism.
And I realized that I can all the basics.
and understand how it all works together.
You must always always build everything.
There was no infrastructure.
You must always put everything together.
I realized everything is completely professional.
I sought a job as a product manager at ImmoScout.
I built the entire Mobile-Sachen on the time.
It was very much time I had to do it.
It was very much fun.
Then to mobile.de, where you were also.
That was our first contact point.
And then the last real job I had, as I like to call it, was the eBay Classified Group.
Then, where mobile.de had to go, where I had the global development of new business areas.
And there I then actually...
First for Germany and then global, I built a business model.
It was about auto-calling and financing.
And that was I did.
Then I founded Felmo.
For those who don't know, it's a mobile health service.
That's what we solved was that many people have really difficult to get with their children, just by cats, to come to the hospital.
And instead of coming to a T-Arts home.
That's also available.
Also, it works well.
It works well.
The customers love the product and service.
And we have done a lot of things that were done with the human medicine.
That means that we had a whole patient-like process that was digital.
And then we had to see what the doctor did.
Because in the normal way, he had a lot of support in the health care.
And he had a lot of support in the health care.
I'm there since two and a half years, a little more now, two, three quarters, I'm at Filmoraus and I've done a little bit of a conversation and now since a half year I'm with Clarice working and I'm again working on it.
Yeah, it's a super, super, super, crazy journey.
Also, what do I find that?
That's on the one hand, you know, offensichtlich irgendwie auf der Command Line irgendwie...
irgendwie weiß, was man da tun kann.
Auf der anderen Seite kümmerst du dich oder hast dich gekümmert um abstrakte Geschäftsmodelle.
Lustigerweise finde ich, sind alle Themen, die du bearbeitest, zumindest in Deutschland, irgendwie auch hochgradig emotional.
Also Autos sind ja hier emotional und nicht einfach ein Ding, das man rumschiebt und so.
Tiere sind ja der...
And that's the Wohlbefinden of the things that are the way to do an emotional emotion.
That's interesting.
One of the two things you do, on the other hand, the command.
Probably a learning topic.
You have learned Informatics, you know how to do code and send it to you.
And the same thing is Mathematics and physics mega abstract.
An der Uni.
Das kann so der Schüler nicht nachvollziehen, wo man eher so Regeln lernt oder so, aber an der Uni kann es ja selbst für die Guten erstmal ein essentieller Schock werden, auf welcher Abstraktionsebene man arbeitet.
Keine Ahnung, im ersten Semester hat man so einen DIN A4 Zettel mit drei Aufgaben.
Der DIN A4 Zettel ist ein Viertel voll geschrieben und man sitzt die Woche nächtelang dran und sagt, was wollen die Schweine überhaupt von mir?
Yeah, absolutely.
I think first of all, this language to learn.
I think that's the biggest thing for the most.
If you have taken this one, and this abstraction, so, in the mathematics, then it's better.
But that's a very hard step.
I'm absolutely right.
And it's not easy.
It's not easy, as you said.
I can't believe that it was easy.
The first two Semesters are really bitter, and then it's fun to make it.
I don't know anyone, who the first semester was easy.
And just the ones who always with a good performance course from the school were always shocked, because it was not so easy.
so ein Resilienztraining eigentlich in komplexe Dinge aushalten ist am Ende des Tages.
Und ich glaube, das sieht man so ein bisschen in deinem Weg, diese Kombination von sehr konkret auf der Command Line zu abstrakt aushalten und die Verbindung dazwischen suchen.
Hast du eine Erklärung für das immer zu emotionale Thema, das ich behaupte?
Lass mich mal überlegen.
Ich glaube, ich bin halt immer hingezogen zu Themen, and in the areas where you can really move something.
What works for me in every constellation is when things really change.
I'm not a good manager, but I'm in the most part always very good at the time when it really changes.
In the two years, in which I had advisory things done, I helped me also in strukturelles changes in the companies.
That makes me also fun.
Good org structures for Skaliering and Wachstum, which I learned by FEMO very much.
We came in a half years to over 300 employees.
And there has to be a few things learned.
And we have also super much wrong.
And that makes me also fun, what is not just emotional.
I think it must always...
this topic, that you can see that you can move or move.
And that is what it is, I think, even more than that.
That's probably even more than that.
Maybe not.
Maybe there is something in me, what I didn't understand.
But it's more about what I know about myself, that I can move things.
I have only a hypothesis and we have never talked about it.
I think that emotional things are just that being able to see and see and have an impact.
A topic that emotional not beset is, is not and so.
But the Kunden-Rickmeldung is absolutely clear in topics that emotional beset are.
When you build a business model, it's albany abstract and everything is lagging metrics, but when you work on a topic where people are crying, they are in the same way and so is the fall high emotional so that you can directly from the customer's own way of what you can do.
What you can do is, I don't know, Chrome to put it.
except in the crisis, the Strom is over the weekend in the cold.
You can't get in the regular way.
Everything, what Commodity-Themen are, the food is standardised, you get these Kunden-Äußerungen, because it doesn't exist anymore.
That was at Felmo very extreme.
The Resonance, the we got there, was really positive, what was also very important for a topic.
That is of course what makes fun, also without a question.
But so much stronger is it probably really this impact to have and that you can see, okay, you can change something here, so to better.
We have looked at us very unemotional things at the time.
I with my co-founder Philip from FEMO, we have looked at us very unemotional things at the time.
And that's what we had done, if we had something good found.
And we had now found in the health care.
And that's good.
And that's of course nice.
That's of course nice.
That's of course nice.
If you have a topic with your own children, they find that of course great.
It's nice if you have something with animals than with solar energy.
There can be more with it.
At least to a certain age.
At a certain age of solar energy is perhaps even better.
I have to, oh, I have to say, I have to say, that everything is a bit of a pyramid.
You can probably look at the society, that things, which are not so much of a Marge generate, so that it is, and everything that is on Self-Actuation, where I belong to, there is a Marge in.
Mode, Emotionen, alles ist oben in der Pyramide und das ist, wo immer die Kohle und die Marge drinsteht, auch wenn das, wie gesagt, in seinen gesellschaftlichen Folgen nicht mehr schön ist.
Beispiel Pflege oder so.
Kinderbetreuung, keine Ahnung, kannst nehmen, was du willst.
Alles, was sozusagen, warum auch immer gesellschaftlich nicht hochemotional besetzt wird, obwohl es persönlich hochemotional besetzt ist, bleibt halt außen vor.
Ja, das stimmt.
Das ist auf jeden Fall deutlich schwieriger.
in those areas to do something, it's good to be supported.
Yes, exactly.
And then you're in the field of regulation, state-wide, support, so that there is something that goes on.
And so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and so, and You ask yourself what you do, you can't explain your children, you can't describe your impact.
You have a strong external factor that you can't influence yourself, which makes your business successful or not successful.
And that's always something that I always try to avoid, such a thing in terms of having.
Also konnten wir bei Felmo nicht ganz vermeiden, ehrlicherweise.
Da gab es tatsächlich Themen in der Lobby von Tierarztverbänden, die nicht ganz glücklich waren mit der Entwicklung, dass da jetzt so tatsächlich jemand so viel Tierarzt bündelt und halt nicht aus dem klassischen System kommt.
There were not only positive responses, but also negative responses and a little bit of fear, what that means.
I can say that I had my, as I say, my hammer was a dark year at my hammer.
But I have learned a lot.
And what I learned is that, as soon as you have been in your area, you have it very hard.
The naive idea was that you bring with my hammer a transparency on the market.
Wie entstehen Preise?
Ich bekomme mehrere Angebote, ich habe nur eins.
Die ganze Kram ist amateurhaft dargestellt.
Das ist natürlich genau, was nicht jeder Handwerkerverband gut findet, dass Transparenz auf dem Markt gibt.
Die Verbände sind nicht immer dafür, die Transparenz zu erhöhen.
Dann kommst du so naiv und sagst, ist doch gut für beide Seiten.
Nö, finden wir nicht.
Das wird, glaube ich, auch immer ganz spannend, was man sich denkt und was wirklich ist.
Ja, total.
Wir haben das auch gelernt.
Also wenn man jetzt sich so...
Because for the children, it's very exciting to see the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of the cost of And then we can take a lot of positive feedback from people.
And then we can take a lot of pressure on the positive feedback from people.
So we can take a lot of pressure on the other side of the patient.
And it would be more than the other way of getting the patient.
And it would be more than the other way of getting the patient.
And there has been tried to look at the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of the public side of was dann sozusagen ein Geschäftsmodell garantiert.
Also ich gebe mal ein Gegenbeispiel.
Meine Tochter reitet.
Niemand muss das wissen und so, aber sie reitet.
Und was ich beobachte, ist, dass es jede Dienstleistung, die es für Menschen gibt, gibt es für Pferde nur doppelt bis dreifach so teuer.
Also wenn du ein Pferdephysiotherapeut, keine Ahnung, wie die geregelt sind oder nicht, wenn Pferdephysiotherapeuten bestellen würdest, was wir nie tun, bist sofort beim...
three-fache Satin eines Menschenphysiotherapeutin.
Und alles, was sozusagen nicht explizit Tierarztleistung ist im Bereich wie Pferde, ist sofort absurd.
Also alles, was so richtig Pferdewellness sozusagen geht, da kannst du dreimal für Urlaub machen.
Total abgefahren.
Und in diesem Gewusel sozusagen das System zu finden.
Wo ist unsere Dienstleistung angesiedelt?
Wo hilft das wirklich?
Wo ist das demokratisierbar?
Wo ist das automatisierbar, prozessierbar und so?
Ich glaube, das muss ziemlich viel Gefummel gewesen sein.
Ja, wir sind da so ein bisschen, man ist ja immer so, dass man am Anfang so leicht naiv reingeht.
Das ist ja nicht das erste Mal.
Ich habe irgendwie, so als ich ganz jung war, ganz naiv Sachen aufgebaut.
eBay classifieds so ein bisschen weniger naiv Sachen aufgebaut und dann bei Felmo vielleicht noch ein bisschen weniger naiv, aber immer noch mit so einer gesunden Naivität, weil viele von den Problemen sieht man gar nicht und die kommen dann und dann muss man halt die in den Momenten, wo sie kommen, lösen.
Weil ich glaube, man hat sonst immer wieder diese Punkte, egal welches Geschäftsmodell man aufbaut.
Leute stellen sich das immer so vor, man hat so eine große Idee.
And then you do that and then you do that.
In reality there are thousands of small screws that have to be able to do it.
And they have to do it all together and work.
And 90% of those, who you do, work first not.
And then you have to fight very much, that you find the right solution.
For example, we have totally understood by the business model of Helmo, We have it on Scheduling and Dispatching, which means which is when and where?
It's clear that this is a topic, but we have, trotz of the mathematical backgrounds of my own, a bit of this complexity, how fast that explodes.
And as we were then with 20 Tierarztes in Berlin, we realized that you can't 30% Business-Effizienz rausholen, wenn du eben den Teil vernünftig löst und halt nicht nur so wie das Standardlösungen machen.
Und dann muss man da halt reingehen, weil ansonsten kriegst du so ein Modell nie profitabel.
Das ist so cool.
Allein das Beispiel finde ich so, so cool.
Ich stelle mir vor, wie Lars ein bisschen Number-Crunching macht und feststellt, sozusagen voller Überraschungen.
that exactly this Logistics is a necessary, but a necessary factor.
It was said, I must solve it, but suddenly, if you don't solve it, it breaks your business model.
And how you will all your energy in this Dispatching stecken, that's in real life, schlaflos.
Also schön, dass ich es entdeckt habe, aber Mist, jetzt muss ich es lösen.
Und da habe ich nie daran gedacht.
Aber dieses fucking Detail, das ist jetzt sozusagen was determiniert, ob wir sozusagen das durchstehen können oder nicht.
Ja, es gibt ja noch witzigere, kleinere Details.
Also das absurdeste kleine Detail.
Wir haben, so als wir mit...
and then we started to start a little bit, we were very interested in what it actually means to trust a doctor.
What does that make?
We were very rational and very first into intuition and thought, okay, good education, good education, good university, man can well with him.
The people were asked and in the first rational interviews, the we had, came out exactly what was was.
And then we were there and then we had prototypically scribbled, okay, how does that look at the whole thing?
And then they said, is this now more likely that you have to do this?
And they said, no, I don't know how that is.
And then it was first time, okay, that's completely what we learned from.
What does he mean?
Well, if he knows my dog and can't do it.
And then we were there deep in and asked, what does that mean, with dogs can't do it.
What is that for you?
My Tierarzt, the name of my Tier, I always greet him.
And we thought, okay, that we'll try now.
And we did that to a routine.
And the Tierarzt said, when they come here, first to greet him, best to give him Leberwurst in the hand, then give him the Besitzer.
A difference is like a day and night in the awareness.
The training standard of Tierarzt is high.
They can all that.
They are super-facchlich.
And this difference is really, is that someone who is in the way, on my Tier to go?
And how I understand if someone goes on my Tier.
That's just that.
And that's so much, much more important than this, this Kleinigkeit.
But in this whole service.
So cool.
So gleichzeitig gibt es ja nebenbei gerade so eine Produktmasterclass auf ein paar Ebenen.
Also erstens dieses genauer hinhören und wo ist das Signal und so.
Cool, macht keiner, sollte man, versteht man.
Entschuldigung, sag erst mal.
Und sozusagen auf der anderen Ebene ist es ja noch was, was so wenig Leute verstehen, weil sie es nicht erleben.
Also erstens, wie du sagst, es gibt diese Idee, ich habe das Geschäftsmodell, dann mache ich das irgendwie und das ist ein Rungeformel, dann wird das schon.
No, no, you set up a system.
It's not like Lars, the held, but Lars, the learning and listening to it.
There are situations where Lars is saying the truth is going to be explained.
But between the end of the day are always these rounds where you say, actually I don't know, I don't know, I can't do anything.
So bitter it is.
Auf einer Makroebene muss ich natürlich einen anderen Plan liefern, das es uns im Jahr noch gibt.
Aber zwischendrin gibt es so viel, was ich jeden Tag lernen muss.
Und ich brauche mich gar nicht hier wie ein Held hinstellen, sondern ich bin dieses lernende Ding.
Und ich brauche ganz viele Vektoren von außen, die mir all dieses Wissen überhaupt sozusagen auf den Teller liefern, dass ich irgendwas damit anfangen kann.
Und ich glaube...
What is so a lot of people learn these recipes and read a few Kagan books and think that's the role and I have to know everything I'm the one who is the one who is the one who works in the field, how crazy that system is that you must build up to this system where these totally weak signals come, the with a lot of work, with a lot of discussions with others to decisions later will be.
That is, I think, so hard to understand.
if you don't do it in your life.
I think that's so.
I was also in Product-Leadership-Rolles.
I think that's when you work in the existing companies, are the signals even more diffuser.
Fairerweise.
Also das wird ja nicht leichter, das wird ja schwerer, weil die Signale ja immer schwächer werden.
Weil es ist ja so, in den meisten Companies, in denen Leute arbeiten, die haben dann ja Product-Market-Fit zu dem Zeitpunkt.
Das heißt, das funktioniert ja erstmal grundsätzlich und man denkt ja jetzt nur darüber nach, wie man Dinge noch besser machen kann.
Und das ist natürlich das Signal, was man kriegt, noch schwächer.
Das heißt, man muss eigentlich noch differenzierter...
to listen and have a stronger skill.
And in reality is it always so, that's not always the first time, but it's always iterative.
I've never seen it ever when I did something.
That I think, okay, this is the target group, the I'm going to go with, I'm going to talk to you, I'm going to do interviews, I'm going to do something and that's it.
That's not a single time in my life.
Not a single time.
I'm often going to go out of the way, that it works so.
And then you're just happy with this one round, you're going to get that somehow, you're going to get that with and you're going to do something with it.
But that works in reality not so.
And I think there comes also a meta learning.
I speak now in my old Can-Bahn-Sprache.
Also on a normal Can-Bahn-Board, when you have cars...
Toyota, bla und so.
Da ist die Kanban-Karte ein bestelltes Auto.
Das ist da.
Das ist validiert.
Jemand will das.
Wir müssen es nur noch liefern.
In der Software ist es ganz anders.
Wenn du ein Ticket schreibst für einen Entwickler mit einer User-Story, ist das eine Idee, die du hast, dass Kunden etwas eventuell haben wollen.
Du weißt aber nicht, du kannst nicht 100% validieren.
Die Validierung ist ...
That thing is online and it will be used in the past.
And in the post you get the signals, that has worked.
That happens not often.
We know even that it's just in a fifth of all cases really works.
Trots all the rationalization we do.
But this difference to understand, that everything we have to do, always a wette and not a reality.
I think that is the Bescheidenness, which we have to do.
zwingt.
Wir wissen das noch gar nicht.
Obwohl wir so viel Research gemacht haben, obwohl wir alles Mögliche getan haben, um dieses Ding zu schreiben, ist es trotzdem eine Wette, die rausgeht.
Was der Kontext aber zwingt, ist, dass du diese Wette halt mit aller Selbstsicherheit verkaufst.
Ja, ja, das wird einen Uplift geben.
Was halt auch diese ganze Spannung...
in the industry.
The CFO must not understand that you have to understand the whole time about the Wets, if you are so sure there.
I think that explains a lot of what happens and how much this role of a PM actually is.
Yeah, 100%.
And I think you have these two extremes that can then build out, which I always saw in our own Org or other Orgs, where I supported.
You have to have this extreme, that people have no signal anymore because they have learned that it's just a bit of a thing.
That happens often.
That's the other extreme, that you don't trust any important steps to make.
Every little bit starts to test AB, but also to test no big things anymore.
Because then you get too much together and then you get into another kind of stuck mode, that you only do minimal iterations.
And I'm a friend of minimaliteration.
Don't get me wrong.
I think sometimes, you can have a lot of models with small changes in the middle of the big impact as with big.
But you have to balance that and make the big things even more trailing.
And there was...
then sometimes when you don't have confidence in it, because you don't understand that whole thing and didn't have time or didn't have support or even that knowledge, how you can have this security, also big bets to be able to get.
And that's, I think, always very important to build up, that people come in the way to be in the way to be, big bets to be.
I think the first thing that's what's going on is from leadership that's self-built.
Who are we?
Are we a innovative company, the big new things will do or are we really just doing it?
And it's all okay.
It's not worth it for me on all these levels.
If it's been 20 years and you don't want to do anything, everything is cool, it's not.
If you just want to do it, you want to do it for 200 people to have a job and you have to do it for, that nobody comes from the side, everything is in order.
You have to explain to your people what you want to do.
So a problem that it is, that all the innovating, customer centered, what else they want to be, and Elliot Kipchoge want to be in the way, although they actually, almost all the companies are just a bit and it's okay.
Then he says, we're doing a bit of things, but hire not people, in which you say, we need the coolest, innovative, creative...
a very grenzensprengend PM, the then comes and will make a standard job.
That's not.
Have a self-image that fits your company, regardless of which level you play.
There is no good and bad, it's only pass.
Show the people exactly that self-image and show them, which free room they have or not.
Then everything is good.
I think that's totally important to understand.
Yeah, 100%.
And at the same time it is, I think, important to be honest about being in which Defensive Plays in der Form überhaupt noch funktionieren, weil du ja einfach so gerade so eine unglaubliche Beschleunigungsmöglichkeit zumindest hast.
Also ich glaube, es ist noch nie so gefährlich gewesen, wie jetzt gerade rein defensive unterwegs zu sein.
Weil die Beschleunigung, die wir gerade haben, ist Wahnsinn einfach.
Ich habe schon immer gesagt, wer dich auslöschen wird, three people with a laptop and not a big one.
That's now the worse or the worse, as it was.
But with one of our main themes, you have a film called, you are going to be out.
In the meantime, you have to create a film.
You are going to be out.
In the meantime, you have to talk about 300 employees.
Probably are it more.
Irgendwann you have to do so many steps.
was jemand wie mir überhaupt nicht verständigt ist.
Oh, wir müssen ja Tierärzte anstellen.
Ich fände das immer noch gemütlich, solange die irgendwie da sind.
Ich verwalte das sozusagen.
Aber in dem Moment, wo man die alle anstellen muss, finde ich das krass.
Das hast du so mal eben verdaut und hast zu mir auch mal gesagt, ich rechne das ja aus und dann weiß ich ja, dass es geht.
Das halte ich schon für großes Selbstvertrauen, überhaupt großes Vertrauen in die Sache.
Mega cool.
Aber da hast du irgendwann aufgehört und das ist genau die Zeit.
Ihr habt auch traditionelle AI schon verwendet in Thelmo, aber dann so jetzt im letzten Jahr was passiert.
Du hast das eben angedeutet, das macht was mit dem Markt, mit der Dynamik.
Erzähl mal ein bisschen deine Reise, wie du sozusagen das, was wir jetzt AI nennen, die LLMs, wie du das entdeckt hast und was die ersten Spuren waren und was dann die späteren Entdeckungen waren für dich.
Okay, when we start with the discovery of this, I think that was the same thing as we all did.
When the chat GPT was the first available version of the first version of the first version of the first version, I think that the first one that showed me was Stefan, also Stefan Schubert-Peters, who was a part of the product that was leading.
from filmo and he came to me and showed me that and said hey Lars, look at this one and I said yes, yes, honestly, Stefan, I've seen so often looked at these things, they don't work well and he said to me, no, no, make it, it's just different, really, it's different and then I thought oh my god, what's happening here then?
I had to try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try and try There is something different than before.
And I thought first of all I had to go back to my Uni-Zeit.
Because with friends together we had so much philosophy, what happens in the moment, in which AI is intelligenter than you are.
And it's irrational, it's still to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get to get And that was then very acute.
which all the linear algebra used, where you know what's happening with the dung-sets, vector-rall and blah, where come the hallucination here, that can't be done today.
And I was so long for the time to be criticized at the technology, because things that it is me today is on my own theory, it doesn't get delivered.
That was my expectation and the technology was for me long against.
Now to the Entscheidung.
I think there are two strong words.
The one is the one to meet and the other is accountability.
So what I think from the medicine, that we can see before, LLMs, is that the bild-on-on-one and the analysis behind often better results than the doctor alone.
The doctor with the AI, that system brings but statistically gesehen, die besten Verfahren.
Also obwohl der Arzt sozusagen alleine schlechter ist, die Kombination von Arzt Reviewed, AI, ML, was auch immer Ergebnis scheint besser zu sein.
Accountability bleibt beim Arzt, obwohl wahrscheinlich sozusagen die Entscheidung, die die AI getroffen hat, fast immer übernommen wird.
Und ich glaube, Accountability ist fast so ein moralisches Thema.
Also wenn was schief geht, wen kann der Patient anschreien oder die?
And that's probably the same thing.
What is in the relationship with Entscheidungen and Accountability that you talked about?
I think that these differences, how can you talk and play and play, so important for everything, what is happening, because the words are not debattated.
So eine oberflächliche Diskussion von, klappt ja, klappt nicht, wenn es von dir kommt, ist es Slop, wenn es von mir kommt, ist es richtig, weil ich habe den Prozess gesehen und so.
Vielleicht mal ganz abstrakt, was ihr da mal so diskutiert habt.
Ja, also wir haben wirklich sehr, sehr viel darüber gesprochen, was hauptsächlich, wie gesagt, dieser Gedanke, was...
What happens with one and what role does it have, if you should not have a decision to make and it irrational, to make a decision.
And in comparison to you, I would say in some areas that is already so.
Also, even for me, if I look at it, I have always developed software, but I was never the best software developer in the world.
I just created things in the real world and that was always my focus.
And that's for me personally, it's irrational to me one type of code to look at, but I don't need to look at the code.
For me personally, and I think that's not for all areas and for everyone.
But for me personally, it's an that point irrational.
I put that in other areas.
For example, when I prepare my Steuererklärung.
That's not the same.
what is in many areas the role of being.
And I think that what you have mentioned with accountability is so important, because that is rather than in many cases a lawful construct.
And who has really the responsibility?
Rightly, when something is done, especially with medicine, a big issue.
We see cars here for us a big issue, in other countries a big issue.
And at the end we have seen it all, and I think that will be this time, that this technic development is not to hold on.
That means everything that everything that life for people makes easier, will we accept and do it and do it.
And I think this time is very important that this is in the right way.
I'm totally with it.
Aber ich glaube, das ist nicht aufhaltbar und das wird kommen und da wird es dann irgendwann diesen Moment geben, wo man dann tatsächlich überall sagen oder in ganz vielen Bereichen sagen muss, okay, man muss hier jetzt abgeben einfach.
Und ich glaube, das ist so das Spannende daran.
Was war denn, du hast es so schön gesagt, was macht das mit einem?
Also meine Theorie ist ja, dass wenn ich merke, oder so gesagt, wenn mir...
If I have no idea of cars, I'm not going to go.
I'm going to go in the car, I'm going to go in the car.
I'm going to go in the car, I'm going to go in the car.
I'm going to go in the car, I'm going to go in the car, I'm going to go in the car.
The car must be changed.
The only thing that I'm going to go to the car is to go to the car.
I'm going to go to the car.
I'm going to go to the car.
I'm going to go to the car.
The car will be better than I.
This analysis of the photo, which I also trust, that determines me zero.
So I say, yeah, it will be so.
Probably I'll make a test of a test, just to see if one is better than the other.
In the moment, where I'm as a person determines what's just the software developers' business.
Every night I have talked about what Lars wrote for a bullshit code or not.
What's for a cool code.
And how he describes the code, that describes me at the end of every coffee break.
And why I describe that and I talk about it, because it's what moves me, what makes me as a person.
And if this is the way away, because that can be done by someone else better, what is also only a subjective debate, is it really better or not?
But in the moment, where it's about my own self-esteem, my own self-image, is it in the moment, where my decision can be taken away from, objektiv, macht es mich ja fast zu so einer Art Opfer oder so.
Und ich glaube, das ist halt, was wir ganz viel sehen.
Und aus diesem ...
I think that my skills are now reduced to zero.
There comes a lot of things.
What that kind of thing makes society, is this distrust.
That can't be done.
We see many defensive movements everywhere.
They come not out of a objective, but rather emotional.
That can't be done.
That it can't be done.
What I'm sure has.
That's what I'm sure.
That's completely impossible.
Is it that on the emotional level, what has been influenced you?
No, that's been, that's been, that's been, that's been, that's been, that's been, that's been, that's been, that's been, that's been, you must have been, in so many mathematics studies and all the people, all the people, the abstract thinking, just very, very much love and in a certain form, maybe I must have been doing this a little bit, yeah, also that was for, in the discussion, the we then always had, kein schöner Gedanke, sondern ein sehr, sehr negativer Gedanke.
Weil das waren natürlich immer alles Leute, denen es sehr viel Spaß gemacht hat, abstrakt zu denken, die sich auch, und das tue ich bis heute, ich definiere mich stark über Denken.
Und auf einmal wird viel davon abgenommen und der Wert davon ist geringer und auch der Wert davon, und ich glaube, das ist ja bei Entwicklern so ähnlich, so tief in Themen abzutaufen und sich da drin mal zu verlieren.
And then really two hours there to be, deep thinking and then again to get on.
That was just a little bit.
And I had to fight with it, as I started to use AI.
It felt like it was a bit of a feeling.
It felt like it was a bit of a feeling.
It felt like it was a bit of a feeling.
Okay, it was a bit of a feeling.
Und ich bin jetzt nur noch derjenige, der irgendwie Sachen reinschiebt und überprüft, ob es richtig ist und Fehler korrigiert.
Und dann wird die harte Arbeit, die eigentlich so viel Spaß gemacht hat, die nimmt jetzt jemand anders mir weg.
Und ich mache nur noch irgendwie den Quatsch an den Rändern.
Also nicht ganz.
Und musste da erstmal auch meine Arbeitsweise wieder neu lernen.
I can have fun with it, if that makes sense.
In the phase I personally went through.
It was nothing, what we saw it then.
Except that it was more categorically thought, okay, that's not a cool day, if that comes.
Because then not to think about it, it's not cool.
It makes fun.
What I'm curious about the discussion is that we describe the output at the value of the value.
That you can't solve a value.
The value of the value of the value is not.
The value of the value is on another level.
That you have five years of this practice, makes you to a different person, the world can't meet.
And the value of the value of the value is not away.
It's not that...
Man geht der Beschäftigung nicht mehr nach.
Das ist verrückt.
Also der eigentliche Wert ist halt super versteckt, aber sagen, dass du ein anderer Problemlöser bist, anders, nicht besser, egal.
Ein anderer Problemlöser bist als andere, die Biologie oder gar nicht oder was auch immer studiert.
Der Wert bleibt halt da, verrückterweise.
Und das ist wahrscheinlich eher der Wert eines Studiums, bei manchen von der Doktorarbeit und so, als der Output, den man generiert.
Also es ist wirklich diese Persönlichkeitsveränderung, die den Wert darstellt.
Und das geht verrückterweise nicht weg.
Ja, auch das kann irgendwann sinnlos sein, wenn so eine Problemlösung von uns nicht mehr gewollt wird.
So alles cool.
Aber ich glaube, das ist noch was, was da ist.
Wie begegne ich der Welt als Mensch?
Ist davon determiniert, wie ich geübt habe.
Ja, total.
Und selbst jetzt in der...
In the actual time, I found it very exciting.
It started to have really fun.
Then it started to have a real change.
I found it a real change.
I realized that I can still have this think and this deep think.
But just like before.
So you have to find something new.
And you can then move from...
löst jetzt nur die Probleme, die ich vorher hatte, zieht man das Ganze eigentlich wieder an der Abstraktionsebene nach oben und baut dann sich auf einmal die Systeme, die die Probleme lösen können und genau dann diese Probleme an den Rändern halt wieder knacken und dann fängt es wieder an Spaß zu machen.
Und dann ist es wieder toll, weil dann hast du das tiefe Denken, bist wieder total im Driver Seat und hast aber eine viel, viel höhere Wertschöpfung als vorher.
And then it was funny, because it was for me exactly the same process.
I have everything, as I mentioned earlier, because in my theory, it can't be good.
You have to have problems, which you have to have.
And so, until I have to be so deep in mind, that I have to see absurd good results.
Where did they go?
That can't be done.
And so.
And then was really exactly this, and then it was really fun to have the machine built, which me solves problems.
And then I was in a world where I could not have a lot of criticism.
This, now is your thinking sloppy and you think...
I was just, I was just sitting on the whole day and think like a devil, like a wilder, about other things and so on.
But I can't understand this sloppy thinking, that it would be abstract.
The thing would be like, the thing is going to generate, the thing is going to generate.
I'm just saying that the only document is not interested in the machine, which is the document is.
And the crazy thing is, I think that it's also that, how to change the life of life.
The designer is not more the one who never had to be the pixel, but the designer is the one who will create the design system and will create the design system so, that the output of the machine will be better.
And now he's actually the designer, where he should be, which is the system of designs that are, which are consistent, consistent, the work they are going to do, but he has a different way of the machine.
And before he said, that's the mood board, that's all good.
No, it's not.
You have to do a machine, which mood boards are going to do.
And that's the right thing.
And yes, you probably need that mood board not more.
That's the way it's the way it's going to be.
And the funny thing is, if you can't design it, you can't build a machine, which is good designs.
What I always find is, at the end of my competence produce the slope.
If I build a machine, which is going to build a workshop and a strategy, everything is super.
I dream.
I can do it to the breakfast and it will be super.
In the moment, where I have been coding for 20 years, I have to go with the output.
And that's the moment where I start with the machine emotional to be like, yesterday you were so good and today you're so good.
Why?
Because that's just a mirror.
I'm now at the end of my competence and the machine fills my Not-Kompetence with what they know.
Now comes something out of the way, completely wrong and freckless.
And that's where it goes.
So, the designer, the machine has to design, the must describe what his process is, how good it looks like this.
If you can't do that, you don't have to wait.
That makes the machine never.
You have to create the room, you have to say, where she knows, what she knows, what she knows, fill it out, and then it works.
But that's not the value.
And that's the value, the first thing goes away.
Yes, and I think that's totally exciting.
wie sich so alle Rollen dann weiterentwickeln.
Wir sehen das ja so in Product Tech, irgendwie ist es so sehr stark da und überall ein Thema zumindest mal.
Und auch da gibt es ja jetzt noch nicht so das Blueprint, was überall gekopiert wird.
Ich glaube aber da, also was für mich zumindest eine sehr logische Entwicklung wäre, was wir jetzt so in unseren eigenen Entwicklungen die ganze Zeit sehen.
is that once the roles all in a certain way converge.
That means what do you need now?
You need to have the right Leitplank for your context, in which you can move, with which the whole system can be done.
And then it works for someone who doesn't have so much Ahnung.
So you can, as you said, we can be a designer.
If you have a design system, then you can as a non-designer job work.
If you are not a software engineer, you can develop, but architecture can be like, if a good architecture is in the Cloud MD files, or whatever, you can have a or two skills, which have the light planks, then I can develop.
If they are not there, then I have security issues in.
And then sits Ben, my Co-Founder, there and he's going to be on, that our data keys publicly exposed.
So that's the difference.
And when you're in the Lightplank, it works.
And when you see that, it's actually a lot of times, the context in which you are in your own context, in which you are in your own context, then these systems to build up, which work and scalierable to get.
There were a lot of discussions about the death of product management.
And I was not sure where it goes.
But now I'm sure that for the people who are in product, the best time comes, because I think that almost every the position in a company, in any form productized.
Because at the end the question is, what is actually a good leader in the future?
Is that the knowledge of the business, what is inside it?
Or is that the possibility of building systems, which scalable work?
And I think, everywhere, where you can get to the second time, are product managers, who embrace this and build this skill, have all their work with them, to make a fantastic job in this world.
And my observation, that I haven't discussed yet yet, I'm curious, what's your look at?
I'm just going to be so aggressive and emotional when I see people who tell me what's going on from a product management job.
Judgment and Taste.
I'm going to be insane when I hear it.
What is that?
What is judgment and taste?
Can I learn it?
Have I done it from birth?
And when is my judgment and taste better than yours?
Also I hold that for not just a danger.
And the funny thing for me in the Product Discipline is that we are working for 10, 15 years and say, when a boss comes with my opinion, we say, hey, hey, but data.
What are the data here?
We have to get data.
And now comes AI and can do it with data.
But judgment and taste can't be done.
That's our job.
What?
That doesn't work.
And how does the chef do not work on judgment and taste?
I don't like judgment and taste.
I don't like judgment and taste.
Let's do it.
What should that be?
And I think what we need in AI is why does development work so well?
Because they build loops on the way, they say, develop, but test against this, test against this.
And we don't test against anything, because we didn't have the discipline.
When is the PID correct?
Will he ever 100% correct?
No, the code is not 100% correct.
But the developers have over 60 years built tools on the one aspect, the other aspect.
Correctness test.
And we can say, if a PRD is the right thing as a other.
Is he a certain need of a customer?
Is that something analyzed or not?
Is that something that is a Lick in the market or not?
Where is the quote?
And there are we super sloppy as product people.
We've been living for a long time over rhetoric and that's called judgment and taste.
And I know the future better than Lars or so.
So I'm better than Lars.
Or so.
And Lars thinks that's the same thing about me.
But that's not the way I think in this world.
I think we have to learn to formalize what to formalize is.
That's a bit of a pain.
That's a bit of a pain.
That's a trock of looking out.
Ja, aber nur dann können die Agenten für uns arbeiten, kriegen Stopp-Kriterien, wann sie fertig sind.
Das ist der Trick in Softwareentwicklung, dass sie Stopp-Kriterien kriegen, wann ist es denn gut und ist es jetzt gut oder nicht.
Und ich glaube, genau die Kombination ist, glaube ich, unheimlich wichtig jetzt, wenn wir auch Maschinenbauer werden, sozusagen, als Produktmenschen.
Ja, das glaube ich auch.
Und das ist tatsächlich ein Thema.
with whom we are now in terms of Claris very much involved.
Maybe we can talk about what Claris does.
Claris does at the end synthetic research for product teams, that means it will simulate interviews, surveys, UX tests, and focus groups.
because we believe exactly this topic so strongly, that it is important for product teams to have this muscle, to get to a more knowledge, which is enough to make it more, so I would say.
because at the end of the end, what has built the product discipline in the way you said, what is compared to the tests that are in the development of tests, which are at the end of the tests.
Maybe surveys, but surveys are in the software development very seldom and also very long.
That is a tool that is actually too long for the development process in AI-Zeitalter.
And A-B-Tests have from my opinion, because the few companies have so much traffic on things, which are tested, if you make it right, that you get in a time, which is in AI-native development, to the results.
So you have that always so stark lagging and you get the results first then, if you're actually already 8 steps further, that doesn't fit anymore.
Plus, there's also a maximum change in the customer.
You can't all the things that you're interested in, you can't mess in a statistically relevant or significant number.
The last company, where I had to do something, we had a committee that tests out, so that we...
in all these tests, from all these tests, to get any signal?
When can which test start?
How many percent is played?
How interferes with other tests?
You can't even 20 tests on, I don't know, 200.000 customers.
That brings no significant results.
There is a little bit of a focus on it and that's right.
Absolutely.
Nee, genau.
Und das ist genau eins der ganz wichtigen Themen, wie man da tatsächlich schnellere Loops reinbauen kann.
Und das ist halt so was ganz, ganz Spannendes.
Ich komme gleich, mache ich mal den Zirkelschluss so zu Taste, weil, wie du sagst, das geht gerade überall rum.
Also, tatsächlich was wir sehen ist, dass wenn Leute dann so ein Produkt benutzen, das geht uns selbst auch so, aber wenn tatsächlich dann unsere Kunden, bei den Kunden das eingesetzt wird, ist, dass am Anfang erstmal gar nicht so gut verstanden wird, was kann ich damit jetzt überhaupt machen?
Und der häufigste Nutzungsmodus von unserem Produkt ist vielleicht das ganz Wichtigste über über MCP dann eingebettet in die normale AI-Umgebung.
Das macht einfach am meisten Sinn, dass es natürlich, dass das da immer noch ist.
Dann kann man tatsächlich innerhalb von Sekunden gute Interviews, Surveys oder UX-Tests starten.
Am Anfang ist erstmal nichts da und deswegen, und Leute wissen gar nicht, was man damit machen kann, weil es halt so neu ist.
Also das ist so total unerwartet, dass das jetzt einfach geht.
Und wenn man dann aber das, first found out, then it goes into so intensive use.
Because you can do it, what you wanted to do as a product manager.
In reality, at least every survey I made in my life, I went out and had more questions than before.
I thought, okay, it's exciting, but why is that now?
And that you can do it now.
You can always go in and go and go and nachhaken.
Every iteration you can do in your product specs, you can just check it out and get back the signal.
That means the whole process, to actually bring a product back to the point, that it has a higher probability of being successful, is so enormous and if that product teams then get with it.
then you can see that it's like a explosions-like.
And then you have 10, 20 of these discovery things pro Tag.
And that can't do anything else.
That doesn't work.
I think there are a few theoretical advantages we currently have.
The one is...
We always compare the Post-AI world with the golden pre-AI world.
We think that synthetic persons could not work, because it's not just people, they are just together and that's never the mensch's behavior.
What did we have before?
We had no persons in many companies.
We had a few people who were debating, what could be three people with the same perspective.
Or we had Personas, which also synthesized.
And I said, a great German publisher came to my office and said, the Personas are all like your WG-Mitbewohner.
Yeah, yeah, what's that with your Leser?
Ach so, yeah.
It's not so that we had to work with great Personas.
And also the nature of Personas is that they synthesized.
echten Dingen und so.
Warum?
Weil immer mit echten Leuten zu arbeiten zu kompliziert war.
Das ist das eine.
Das andere ist auch da, ich glaube, Leute, die nie gesehen haben, wie Context Engineering auf so einer Produktebene, nicht Engineering Ebene funktioniert, wie das funktioniert, dass immer der richtige Kontext in Subagenten geschickt wird.
Das klingt jetzt so kompliziert, ist es gar nicht.
Da haust du dir verschiedene Skripte durch nacheinander.
Egal.
And that can concert, orchestrate of these scripts and so on, I don't know, auseinandersetzen, debate, A gegen B werfen and so on.
What for Ergebnisse come out, how many of the theoretical possibilities are released through this process, that's irrelevant.
And what I can do with the people, the people can see that they can see that.
What I can also see is, when I...
I'm in the meantime on Strategie-Workshops, which I have the context, I have the context, I have the context, I have the context, I have the context, I have the context, I have the context, I have the context, I have the context, and after four rounds and a hour or so, I have a workshop and I make that so, that it's not a agent, the is hiding, but I see the results, that the report to me and say, oh, now comes a spannet, and so.
That's what is, and that's what it's like, where it's so bad.
I think I think that many people, if they see the results, these life strategy workshops are not so difficult.
We think that the strategy is so difficult and abstract.
And if you go down, there is a lot of social news there and a lot of roles and so on.
But the facts, the debate are debated.
are not so difficult.
It's not so that someone has a strategy workshop and then with a master strategy.
It's in iteration.
And regardless of what kind of problems I had in the early phase of my machines, they come always with the main direction, which is in real life.
Also, in good strategies.
Here, Lego, I don't know.
Or in existent companies, which now knacken, like, how does the HR world have been from AI, where the Kunden has to hate and not do not do.
It's always the same solution.
It's not so hard.
The human being said it, but we still have to talk about it.
We still have to talk about it.
We still have to talk about it.
Make the solution worth it.
Nothing from this work comes to work.
And I think this, look at this, how a workshop will simulate and how debates will be carried out and if they really are different than in real life.
And look at this result that comes out of social noise.
You will be able to be happy to be, how the Aufgabe is.
And there is backwards from the way that...
That to formalize what is correct and not correct what is really abgeleitet instead of fantasizing.
That is not so hard.
Abgesehen from the social aspect, which is a huge aspect of a workshop.
We align us, we're going to do it again, we're going to do it again.
And so the one likes it, the other has it.
That is already clear as you think.
And if you have that for yourself modulo menschlicher Wir-sind-anders und Eitelkeit und sowas, ist da doch eine recht hardcore funktionale Aufgabe dahinter, die erfüllt werden muss erstmal, die verblüffend wenig schwer ist zum Großteil.
Und nur dann schwer ist, wenn ich wieder keine Kontextdaten habe, keine Umgebungsdaten, wenn ich nicht weiß, wie der Markt ist, wenn ich nicht weiß, ob die Lücke im Markt existiert.
If I can't understand the conditions, the conditions I can't understand, then everything is difficult.
If I have everything, it's a surprise.
I think that's hard to tell.
I think that what you said there, that's really important.
How and in which way can you trust the things that are coming out of it?
And for what?
And that's a very important point.
What we see now with our product, with Clarice, often often is that in product management, you must then often prove that you are in a certain direction.
That means that the data are not only for oneself, but not only for oneself, but also for others.
And in that moment, they are very different.
But, at the same time, we have seen a lot in our research, as we have done this whole thing, that there are areas that are not perfect in which the things are not perfect perfect.
When you go from surveys out, it's always so, the Render are underrepresented, if you do nothing against it.
There can be a way to work, and there are methods for it, and that's then also in the grip to get, and that's not just a LLM.
That's a topic.
That's a LLM is first of all more stereotypical.
as that actually the case is the case.
That's a little bit of this irrationality of people.
That's a very important point.
All the actual opinion-based, that's also political simulation, is very, very difficult to achieve ad hoc results.
Because that's very, very strong.
And then there are still some issues like, where do I get my news?
How do I get my development?
How do I get that all?
How do I get that all?
That's a lot too diffuse and works well.
And that's two examples.
And one of the important things that I learned, what I always learned about, is that you can give people with, where it works and where it works not.
That's for me nowadays one of the main factors, which is one of the main factors, which is one of the important things that you can use as a system in addition to a LLM, is the decision about whether I can trust the things or not.
And then also, that's how we have to do it, because as I said, you have to do it sometimes, the easy way to do it, that's how we can do it in the one of the 100 cases where you have to do it, you have to do it with real people without having much more attention.
And that's the things we learned that you have to offer, so that you have to offer a system useful and useful and useful to use it and not only for yourself to use it, because it's always a very different question, but also outside the way to make it more useful.
Jason Mollick hat das irgendwie vor ein paar Tagen geschrieben.
Was er geschrieben hat, ist, dass die Cloud Code, Codex, die Harnesses super gut für Softwareentwicklung gemacht sind, weil das Ergebnis testbar ist.
Und ich glaube, in dieser Gedankenarbeit, in der wir sind, braucht es immer, und dafür sind sie nicht so gut gebaut, braucht es immer Nachvollziehbarkeit des Sensemaking, bis hin zum Beweis von mir aus.
Und das ist sozusagen, wo der I have sought for a while, when I build an agent, who works in the background and works with the results.
I couldn't use it.
I was like, I'm not kidding.
I'm not kidding.
I could use it.
But I can't use it.
That's weird.
I'm wrong.
Or are they wrong?
And I realized that the Claude Code, the me orchestrated, then also over skills, Metaskills or so.
Prozesse ablaufen lässt, die ich nachvollziehen kann, hilft mir viel mehr, weil ich es hinterher erklären muss.
Das war für mich so, hä?
Und dann, nee, ich bin gar nicht falsch gewickelt.
Das ist ja mein Job.
Ich muss ja Leuten erklären, warum das stimmt und so.
Und wo das herkommt.
Das muss der Entwickler nicht.
Der Entwickler muss nicht erklären.
Und übrigens, dann habe ich das so und so gemacht und dies und das.
Nee, das steht ja erstens in unseren Guardrails drin und das testen wir hinterher und deswegen ...
Is the code right?
When he's done, the test agent says, it's done.
That's right.
We have to explain this stuff.
So we have to explain this trace back.
Where does that come from?
Why should that be?
Why not right?
Why do you say this?
I think that's very important for the Not-it-it-it to understand, the these systems use.
So if you want it or not, you will have the accountability for the sense making.
And that's explaining.
I think that's a very important point.
I think that's why it's now for products, many of these skills libraries are not a fan of.
Because exactly what you said, it's not difficult to build skills.
The AI baut dir den Skill.
You must just say, how.
And that's not a lot of work.
But that's not a lot of work.
But that's what you should do with it.
What's happening?
What's happening?
What's happening?
What's happening?
What's happening?
What's happening?
What's happening?
Deswegen gar kein Freund von diesen Skill Libraries.
Die kann man sich einmal angucken, wie die die Struktur gemacht haben.
Da kann man was draus lernen.
Manchmal sind da clevere Ideen drin.
And that's about it.
Vielleicht mal nur als statistischer Data.
Null Werbung soll das dafür sein.
Ich gebe diese Kurse dafür, Produktmanager spezifisch mit Cloud Code zu arbeiten.
Und alle kommen erst mal mit der Idee, dass man sich solche Systeme klauen könnte und so.
Nach der ersten Woche guckt sich niemand irgendwelche Systeme an.
Nach der ersten Woche verstehen sie sofort so, Kontext sammeln, Endoutput überlegen, wie soll das sein, wie könnte das gut aussehen, das muss ich beschreiben können.
Den Prozess muss ich meistens schon gar nicht so gut beschreiben können.
Das findet das Ding schon irgendwie raus.
Wenn das Endresultat doof ist, kann ich das ein bisschen verbessern über Zeit und so.
Genau, und ich lasse das dann bauen.
Ich diktiere das den Dingen im schlimmsten Fall und dann wird das halt gebaut, die Skill.
Das ist wirklich nicht die Kernkompetenz.
Die Kernkompetenz ist, wenn das mein Ergebnis ist, was brauche ich dafür?
Welchen Kontext muss ich reinschicken?
Wie sieht so ein Endergebnis gut aus?
Und wenn das noch nicht reicht, dann sozusagen, wie kann ich das in kleinere Schritte geben, wo ich so einen Kontext spezifisch reinschicke?
Und das ist schon richtig, richtig advanced, wenn man auf die Ebene muss und so.
The first steps are already done.
And how that's a quantum leap to, I chat to a couple of things from a copy and paste from a couple of things from a document, that's what the people, I believe, have to experience and what the skills are, which they now need to build, um the machine to build.
Yeah, exactly.
And I think also, if it's so one Entwicklung there was, the I've seen as a look at, how it's been so far, how it's been so far, The language models are always intelligent and better.
So at the beginning, you needed very many Guardrails, so that it worked and they were working well.
And all of them had hundreds of MD files written, so that things were on the right side.
I have always realized that I have to re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re-re- can you go to the Abstraktionsebene again?
The thing where we are just a lot of experiment with experiment, is then iterative Loops.
That means, really so to go and see, in which areas can you do that so, if you build a system, that so Aufgaben are taken over, which are really repetitive.
For example, if you look at some kind of...
Contact-Requests by us rein and then directly answer to answer.
You can either set a CRM-System or you build a loop and the makes that better because the more Context has given and then you don't need a CRM-System to get, but that thing does it.
Or Security Reviews laufen by us so through and then the adjustments start.
Then the Pull-Request must be looked at.
I think that's the next step where you can go and look at it again.
That makes a change.
And the change will probably go again.
I don't know what's coming.
But that's the learning.
I had this.
The Abstraktionsebene was then up.
Models are always intelligent.
Guardrails are always going to take away.
And I think that this life for people for people in the industry is easier to look at it than for developers.
Because the developers have always the first impulse to look at it in detail.
And that's a huge psychological burden that I don't have to do it.
For people in the industry are often not so deep in the technique, it's easier to talk about the right level with the LLM to talk and say, no, I want to have in the document following chapter.
Can you do that?
Dann weiß die LLM, was gemeint ist.
Und früher musste man wirklich sagen, ja, so ein Kipper-Bekapel wäre dann so und so aufgebaut und dies und das.
Und da ist jetzt so viel Weltwissen, wir wissen, dass es kein Weltwissen ist und so, aber so viel Wissen im Hintergrund, dass wirklich ganz viel von diesen Details weg können.
Und das heißt, Skills bauen ist noch einfacher.
Und immer einfach anfangen.
Bau mir dieses Dokument und dann gucken, was fehlt.
Da kann man immer noch nachsteuern.
Aber erst kann man es richtig doof machen, so blöd das klingt.
Ja, total.
Wir sind total weg von Prompt-Engineering, das spielt keine Rolle mehr und Context-Engineering ist gerade überall.
Aber ich habe es auch ehrlicherweise noch nirgendwo richtig gut skalierbar gelöst gesehen, das ganze Kontext-Thema.
Ich weiß nicht, ob du das schon mal gesehen hast.
Also ich halte Vorträge dazu und kriege immer ganz viel Gegenwind.
I think people understand everything.
What I say is, it's totally simple.
I think you have a metaphor from a living room and your own room.
And in the moment, when something is divided, comes to a living room, you need to have a PR and then it's a gift.
And that's where it breaks for the most.
And then I say something like, it's cool if the company's strategy is in MD, your product strategy is in one and so.
Und wo es bricht in so einem vorauseilen Gehorsam ist, dass die Leute das verstehen, dass man es bräuchte, aber dann sowas sagen wie, du kannst ja dem Geschäftsführer nicht zumuten, dass er irgendwie, das ist jetzt wirklich Stereotyp gesprochen, im Kern ist das, was kommt, dass der jetzt in Git die Strategie eincheckt.
Schäuble hat ja auch E-Mails ausgedrückt, dass es mir wurscht, wie man es konkret regelt.
The context of the context is a problem.
That is what developers can do with Git Repos, also on their own.
That problem is that it is currently in the company nobody owned and then it's a very catastrophic knowledge management problem and knowledge management problem was eh not yet solved and then you're in this area.
But if you want to solve it and someone that's the problem owned or that pro...
that is all about the business.
Soll es geowned werden oder nicht?
Aber die Modelle wären einfach, wenn es jemand ownen will, glaube ich.
Ja, genau.
Oder alternativ, also worüber wir gerade viel nachdenken, ist auch so für uns selbst und experimentieren da gerade mit rum.
Also wir sind jetzt mittlerweile so weit, dass wir tatsächlich behaupten, dass eigentlich das LLM, also bei uns gerade Claude Code, obwohl wir da auch wieder...
You can't change it.
There's a lot of issues, but let's say Cloud Code is actually our operating system from the computer.
If I make my computer off-match, then I don't go into the browser or do I do things with programs.
I go into Cloud Code and make everything there.
That's my operating system.
If I'm honest, if the point is, is that what it's done, or is that my use?
If it's done, then it would make sense.
that you build that so, that the context always from the chats, which you have, automatically entsteht.
That means if a strategy is used to use, then it is recognized, read and then it is just updated and it must only one decide, at the beginning, what is the information that is all about using and for whom is that?
But I think it is very important that this This thing that this technic process has and this overgebaut, that that disappears, that that takes it out.
And for me there is no reason why it can't just happen.
We know what I mean?
Absolutely.
And I think the people are on the chair in the moment where they think...
Is that Obsidian, that is in the background triggered?
Or are those memory systems that I build?
The solution is there, the solution is not there.
The solution is on no other level.
And that is for me the perfect Internet story.
When did the Internet have its explosion?
Not as we had to go down and so, but as it was clear that the browser will be the new integration level.
And that had a technical...
so that the dynamic language can be in the browser.
And so that the integration works.
And so that the problem of statelessness is in HTML.
That was the technical background.
We know that the Clouds and Codecs are okay, but not the right interface.
We have a lot of tools to get with the MCPs.
We can see everything, everything, everything, everything.
Aber sie sind nicht die Interaktionsebene für ein persönliches und ein Firmenbetriebssystem.
Und ich bin mir sicher, dass das die Dinge sind, die gebaut werden und dass das die Dinge sind, die die Explosion bringen werden.
Also jede Technologie braucht immer so einen höherwertigen Wertcontainer, wie den Browser, in dem Dinge passieren.
Und wir haben den höherwertigen Wertcontainer nicht gefunden, dass Token irgendwas explodieren lassen, was dann sozusagen einen gesellschaftlichen Wert erzeugt.
In the moment where the technologies are there and idiots are safe, is that where it explodes and where everyone sees these advantages.
And the interaction will probably be quite sprachful.
So, make this, make that, help me here with, I'll do it there.
And a little bit of a formula, the connection to pull and so.
Maybe then visually, I don't know.
But that's, I think, the work that the industry actually is before.
Yeah, I think I'm very sure.
I think that whole topic, if you think about it, it's very elegant and beautiful.
And the question is actually, what is between today and the point, until we are there?
And what does it make until then?
And that's the bitter time of the time.
What?
Maybe it's a half year, but it's a horror.
But it's fun.
Or like in your math study.
That's what I'm doing.
That's how I'm dealing with things.
And that's for me actually the nice thing at the time.
And not so much.
So I'm not so much in the way.
I'm not so much in the way.
and so on.
I think I just saw that just on.
I thought, well, let's see what's next problem I'll solve.
I find it also very exciting, what's on us.
I'm on the other hand, I think, more than that, that it's so net positive.
Man kann jetzt nicht wegreden, dass das für Leute auch negativ ist.
Das ist bei jedem Technologieumbruch ja aber so.
Das ist nicht neu.
Das ist immer so gewesen.
Ruhrgebiet bis heute.
Also es ist nicht überwunden, dass sozusagen der Wechsel in die Serviceindustrie ist.
Ja, genau.
Nicht weg.
Ja, und das ist auch dabei, aber ich glaube ganz fest daran, dass das net positive ist.
Ja, und ich glaube, also was wir auch nicht wegdiskutieren können, ist, dass wir ein ungelöstes Energieproblem und so weiter haben, vollkommen ungelöste gesellschaftliche Probleme.
Und es fängt an mit sozusagen, dass die Technologie gebaut wurde aus Diebstahl an Intellectual Property und so.
Alles nicht schön.
Das Problem ist sozusagen, dass das Dooffinden alleine nicht hilft.
Das ist sozusagen mein philosophisches Problem oder ethisches Problem.
Die Technologie Dooffinden und...
Verteidigungswelle aufbauen, wird halt nicht helfen.
Es gibt ja den Vergleich zu den Luditen, die waren ja hochgradig technologisch begabt, haben trotzdem die Webstühle kaputt gemacht, weil sie als politisches Statement sozusagen, wir wollen das wirtschaftlich nicht, was hier passiert, obwohl wir die Technologie verstehen.
Das hat aber nichts gebracht.
intentional vielleicht super, aber es hat nichts gebracht, die Webschule kaputt zu machen.
Wenn du jetzt was klein haust, kommt es halt trotzdem wieder.
Deswegen glaube ich, das Einzige, was hilft, selbst wenn man hochgradiger Gegner der Technologie wäre, hilft nur Beschäftigung.
Also selbst wenn ich sie kaputt machen wollte, müsste ich ja im Detail verstehen, wie sie funktioniert und müsste Fehler vorführen können.
Das klappt halt nicht aus, das ist doof.
Das ist mein Problem damit, mit viel, was man so liest.
Why he should not do it, he should not let it in.
I think that's the way the people who come from this defensive position are not so much.
That's my problem.
That's my problem.
That's my problem.
I understand that.
I'm also in a generation where I would normally say, okay, I've learned a job, I'm very good, I'm very good, I'm very good and that's what I do now, so long until I'm going to work and not anymore can or have no more interest and can now a little bit ernten, what I saw before.
And that's what I'm just taking on.
That's clear, because you don't have this luxury, you're not going to take on it, what you've already done, but you have to deal with this topic.
You're going to have a problem, probably.
And I understand that people can frustrate that, if that's what I'm taking on.
And it's just alternatively.
I think in the next two or three years, you have to move it to the other side, what you're doing else, and do it again.
And that doesn't work.
That's the nice thing, Lars, is that there is someone who has been very old, who has already lived, who has already lived.
I have written a lot about it.
I have written a lot about it.
I have done a lot about it because the problem can't be new, which we are now.
The technology is new, but the problem must be old.
Heidegger wrote about it.
Who else?
Not that I have so much with him on the hook, but he calls it that concept of Gepard.
We are in a context of war, and about it has no choice.
Also auch die Helden, Boris Cheney, der sich jetzt dumm und dämlich verdient damit, ich weiß nicht, der schon immer ein total guter Softwareentwickler war, der kann nichts dafür, dass sozusagen in so einen für ihn positiven Kontext geworfen wurde und jemand anders, so bei meiner Branche Beratung geht es ja nicht so gut dadurch, also tatsächlich.
Und niemand sucht sich dieses Schicksal aus und niemand hat die Macht, dieses Schicksal für sich zu wählen.
What you can choose, is how you deal with the success.
You can choose all the positive options and try to get them all out of the way, without guaranteeing that you can't get them.
This American thing, if you give up and you don't sleep anymore, you're already successful.
That you can completely forget.
Everything you have in the middle of it, is to be able to get them to find positive options and them out of the way.
But if you do that not do it...
then you have not chosen what happens with you.
I think that is really emotional all of a very high-sounding.
But if we have the 20% ratio, we have to have, you have to try to have, this option to beherrschen, which in the game are left.
Yeah, exactly.
And I think now is also one of the points where you really start to start.
because it's all about all.
So, it's very important.
Cool.
What do you do with your options in the future?
Also, the things you want to tell, that you options use, use.
What's your life?
What's your life now?
Yeah.
In general, I'll do this with Ben and Clarice.
And now, I'm going to go with the family in the Urlaub.
That's so the concrete.
I'm like every family, every year.
Cool.
I write in the show notes, where you are going to find and where you are going to find Clarice.
So that people can understand that, who have Interest.
Ansonsten, mega cooles Gespräch.
Ich wusste gar nicht, wie weit wir gehen würden, aber wir sind weit gegangen.
Hat mir total viel Spaß gemacht.
Ebenso.
Und hoffentlich bald im echten Leben wieder.
Wir sind ja nicht so weit voneinander weg.
Genau, ja, das hoffe ich auch.
Das werden wir mit Sicherheit machen, Markus.
Waren viel zu gute Gespräche in letzter Zeit.
Ja, super.
Vielen Dank, dass du dir die Zeit genommen hast, Lars.
Mach's gut, Markus.
Tschüss.
Bis dann.