Notion übernimmt

Transcript

00:00:00Welcome to ThinkDifferent, ThinkAI, the podcast by Mark and Jens.

00:00:07Two technology enthusiasts who not only talk about artificial intelligence but live it.

00:00:14Here you will find clear classifications, genuine practical insights, and a fresh perspective on what is possible.

00:00:20Understandable, critical, and always with a wink.

00:00:24Food for thought, for smiles, and above all for discussion.

00:00:33Welcome to a new episode of Psyngtiffin Psyng AI.

00:00:37Jens would love to be here today, but he has somehow made it a tradition,

00:00:43that whenever we have a guest to discuss Notion,

00:00:47Jens somehow is absent. This time he's not on vacation or anything,

00:00:51this time he is sick. Wishing you a speedy recovery, Jens.

00:00:53Jens and hello Dirk, great to have you back. I'm very pleased that we can discuss today about

00:00:59Talking about Notion. And you won't believe how happy I am, dear Mark, that I get to talk about

00:01:04Notion. Yes, yes. Well, it actually is the case that in the meantime I've been seeing more

00:01:10people around me who are dealing with Notion or are getting into Notion,

00:01:18who were kind of inquiring in their companies, who among you knows Notion, uses Notion

00:01:25and depending on the company you'll find more or less, but everywhere you find people who are into Notion.

00:01:32And I already mentioned in the last episode, you know, Notion, it's a bit like back then

00:01:39went by, until then, and you always said, use Notion, you always said, use Mac,

00:01:44Yes, I also said back then, it makes sense with Mac, and somehow with Nautchen too.

00:01:48That was version 3 here, right, when they started to actually bring in proper AI.

00:01:53And then we also talked in the podcast about what Nautchen brings and does for you.

00:01:59And just to jump into cold water.

00:02:03Nautchen has done something again.

00:02:05And then we said, let's have a spontaneous conversation.

00:02:08What did Nautchen do?

00:02:11Well, they actually did something that one wouldn’t have expected for a long time. They really opened up to developers, launching a Notion Developer Platform just a few days ago, in a style reminiscent of Apple with pre-recorded keynote-like presentations, like Apple’s reboot, like Apple’s unfortunate things, like Apple’s reboot, we both don’t like that, we want to see something live, but now it is like that.

00:02:35And that’s how they did it.

00:02:37And Ivan Sao, the head, started off and said, yes, we now have a real

00:02:41developer platform and the most important tool there is the Worker, so a, technically

00:02:47you can explain that better later, but a program part running on the Notion platform,

00:02:53which can access Notion natively, on all sorts of things via the

00:02:58API and can access Notion in return.

00:03:00And that’s already a really big step.

00:03:03because it was already published and tested in the Alpha, in this ambassador program, but

00:03:08I don’t have much knowledge about CLI programming and that's why I didn’t actually do it

00:03:13until it reached this point.

00:03:15And now I’m totally fired up about it.

00:03:18What I appreciated about the E-Note or the video recording, pre-recording, ah, whatever.

00:03:25What I really appreciated was actually this atmosphere that you created

00:03:30dark room, wooden chair, I sit down, speaking with a calmness and conviction,

00:03:37not calmness in the form of sleeping pills, but a calmness in being aware

00:03:42of what you’re showing. One is really interested, the Wellerpaum, was aware,

00:03:47that you are basically doing something new, very unexcited, but in what you showed,

00:03:53everything seems so, I would say I’m now a beginner compared to many others in

00:04:01Notion, but it seems a bit okay, we’re now also showing everyone who has Notion,

00:04:06there’s this developer community, this CLI, I don’t know, what kind of tools and you had

00:04:13already mentioned the topic of Worker in your mouth and CLI, so CLI is the opportunity,

00:04:20that we can intervene with things via the command line, Command Line Interface. We

00:04:27already know this from code, there’s a CLI, we know it from Codex, there’s a CLI. The

00:04:33fewest people in everyday life grapple with a terminal window or in a, what’s it called

00:04:41in Windows, CMD, command, ah, I don’t know, people who see black screens with a blinking

00:04:46cursor know what I mean and those who don’t know what a blinking cursor

00:04:50on a black screen is. Welcome to the new world. You presented something

00:04:55and I also watched that video after you sent me the link again,

00:05:01as preparation for today and I’ll try to summarize in my words what

00:05:06they showed us there. Because they addressed something, and I had a completely different view

00:05:10on it, that I always wondered how

00:05:14Notion will deal with these problems and I want to briefly expand, when you

00:05:18look at other hashers like GitHub Co-Pilot and so on, so Microsoft, you see that the

00:05:25pricing is going through the roof. They have announced new pricing models, prices will rise,

00:05:29will rise, will rise, will rise. And not because new models are coming,

00:05:33but simply because they charge differently. There are somehow no more subscription models

00:05:37and stuff for enterprise customers. And now Notion also addresses the

00:05:43business market. That’s very strong. And I’ve always wondered how they

00:05:47will maintain that with the pricing, because, they’re not small, but the costs are also

00:05:53a challenge for them. How do we remain attractive in terms of price? And I think what you just presented with this Worker

00:05:59also pays for that. In my perception, the

00:06:04Worker is the opportunity to write program code and provide it to an agent

00:06:10to work with, or also to provide it as a juice for developers.

00:06:14In Notion, it doesn’t really distinguish whether a tool is used by an

00:06:19agent or already by a human. And you can basically, it’s called

00:06:24TypeScript, programming language, totally irrelevant, yeah, it’s just a programming language,

00:06:28in the times of AI, it’s also totally irrelevant which programming language it is. Greetings to

00:06:34the developers who work for me, they always get on the roof when they

00:06:37encounter a strange situation. But to make a long story short. You are programming there basically with the help of AI

00:06:44small code snippets. Because it’s program code, you don’t need token costs, no

00:06:50AI. It’s reliable. It’s called deterministic. And these code snippets

00:06:57you then let run at Notion. And that doesn’t sound so boring

00:07:06but it’s not so boring because you can indeed work with this agents platform,

00:07:13yes, that's what they presented before, that you can have self-executing, self-supporting

00:07:19through, through whatever you want, status changes in a dataset or time or what

00:07:24have you, with these agents you can access these workers and now you can

00:07:28build an entire universe. You can create a worker that accesses your locally

00:07:32installed Mestral or you can create a worker that

00:07:37as you explained to me this weekend, I don't know, retrieves emails somewhere

00:07:41and so forth.

00:07:43So everything you might have done with Nacht n or make or saviour, you can

00:07:47do with this, but in this, let's say, environment completely under your control

00:07:54and not with a platform and so on.

00:07:58And you just have it, sure, you have to use Notion for that, but you want

00:08:01to, you have it in your system. And I find that quite impressive because

00:08:06they are of course opening up with this and the business model, and you mentioned

00:08:10it briefly, selling their own tokens, just like we all do now,

00:08:14because with the agents they also want to sell credits and they do sell them. It's

00:08:18not subscription-based anymore, but you pay per credits, so you have to pay

00:08:2310 dollars for 1000 credits and then they are really, then you pay new money again.

00:08:27And although that is their business model and they say great, we are now selling

00:08:32no software, we sell work, so we sell work in the form of tokens and so on,

00:08:38they still open up, and I think it shows that they still have quite a lot

00:08:43planned and are not going to stand still at this point,

00:08:46but are perhaps looking for short-term success to sell even more tokens,

00:08:51block it, but in return you can attract a completely different community.

00:08:55And one must not forget, I might be in a small AI nerd bubble right now,

00:09:01you in your Notion world, but nonetheless, what you just said is that they are opening up in themselves,

00:09:08is something where everyone thinks, who maybe can’t open Notion,

00:09:11why? What’s so special about it? It is actually something special. Notion has

00:09:15a really great way of storing data, with cell pages, databases, and all that.

00:09:20seen and everything is kind of the same, and a point why I still believe,

00:09:24why a company like Microsoft can never get it right, because Notion has done it fundamentally right

00:09:30in that regard. I think they have also paid a bit of tuition

00:09:34in the past, in terms of performance or maybe they got a bit ahead of themselves.

00:09:38And now they have the opportunity, the chance, if I take that,

00:09:42can be said, through AI, its good data storage and its good interaction capability

00:09:49to enrich it, making it much easier than the others have,

00:09:54to integrate agentic systems, to let agentic systems work and

00:09:58now also again with the Worker regarding opening. The Workers allow me not just to say,

00:10:05I'm looking to see if, I don't know, an email in Gmail, you just gave that example,

00:10:10is there, but I can really interact and in the sense of, I'm checking,

00:10:15I'm logging in, I'm checking if there are emails, I can check if specific emails are there, I can write emails.

00:10:21I could also check another database, I could use other tools, yes, so what we have in the AI world with MCP connectivity, you can put all that in there too, and the funny thing is, even if I spoke about a strange TypeScript.

00:10:36Thanks to AI, it doesn't matter to us what it is, because ultimately you can practically wish from the AI what the Worker should do.

00:10:44That means you have used AI and from then on again a deterministic system.

00:10:49I find it really fascinating that they built that adhesive tape, that duct tape there.

00:10:57It always sounds so unprofessional, duct tape, but it is really a special feature that is also simple.

00:11:05What did you build with the worker? What was it that we looked at there?

00:11:09So I built it myself, this, well I then,

00:11:13a bit awkwardly with Cloud Code in,

00:11:15well not in the CLI, but in this, with this app, from Cloud MacOS App,

00:11:21Mistral,

00:11:22connected it so that it runs locally on my,

00:11:24on my MacBook,

00:11:265M5 Pro,

00:11:28and not really, what's great is it can, but it was cool to see that it

00:11:32works just without internet,

00:11:33there's AI.

00:11:34It only has training data, and it doesn't have the fastest version yet, but

00:11:37a very good version that is already running, and I wanted to use that with Notion, wanted

00:11:41to find out if that works, and then I built this worker with Lord.

00:11:44But what I found a bit more realistic was the question I asked you,

00:11:48namely that I regularly check a specific Gmail account, and therein we then

00:11:54come to the first mode that this worker can take.

00:11:58It's a really supported, well actually proprietary mode, namely the

00:12:02so-called Sync.

00:12:03And I didn't even realize that this is such a mode, then you wrote

00:12:05me somehow how I do it in which order. And then I

00:12:09actually did that in this CLI, in this terminal I entered everything.

00:12:13That worked super well. Then I used the Cloud Code on Terminal and not on Desktop.

00:12:19And everything worked out. And now I have a worker that regularly checks, in this case

00:12:25every 15 minutes, a Gmail inbox for what comes in. And the crucial point now is,

00:12:30the worker creates the database initially, Notion creates the database,

00:12:37where things are supposed to come in and then defines what it is, in this case there are 5, 6 sync fields and

00:12:44they also have a special title, because they are basically synced from an external

00:12:49tool, so to speak, and then you can create some more yourself,

00:12:53as you know, but this is a new, new type of database field, namely

00:12:58a special type, and you can extend that as much as you want. Now you can make the worker look up something in another database and what

00:13:07links and so on. I haven't done that yet, but... and let's say with instructions, that I also needed, because I'm not, as I said, a

00:13:14real developer, rather I count myself as someone who just does sort of half-developer stuff. That was good?

00:13:21So the instructions you made for me were good, but I needed them too. Of course, they were super tailored to

00:13:27my case. So I could just copy-paste it. I didn't have to

00:13:31understand what I was copying-pasting. And I was just glad that it came from you and I

00:13:36thought to myself, okay, you wouldn't build something stupid for me. And well, that's not

00:13:39a security-wise, that's not a recommendable approach now. But if the

00:13:45cousin does it, then it must work somehow. In fact, for

00:13:49non-developers, I find it not to be such an easygoing process with this CLI thing. You have to

00:13:56understand all the nomenclature and so on, you can fumble through, it's

00:14:01not super ideal yet, but it is indeed possible even if not quite

00:14:05perfectly, and finally, Notion is also trying now in many small videos and on the developer platform to get the

00:14:11original Notion people on board because many won't go this way because

00:14:16they see this weird, strange cursor thing and say no, I'd rather do it myself.

00:14:20then I'd rather go back to night n so that's already a small hurdle but

00:14:24But of course, you have completely different options.

00:14:27Because I will tell you what I did with the workout.

00:14:29Maybe two things, because as you just mentioned,

00:14:32first, N8N greetings to SAP, who somehow

00:14:35has announced a partnership cooperation, participation,

00:14:38something along the lines of

00:14:40future processes of SAP,

00:14:43will be supported by N8N, which made me think.

00:14:47Yes, okay. Honestly, for N8N

00:14:50I personally wouldn't do much more,

00:14:52it's good for SAP, but the hammer shall be placed there.

00:14:57No, with N8n there is also something like Make, you mentioned that you

00:15:01use it, Make.

00:15:02And they also announced a CLI now, or rather published one.

00:15:06From that side, it seems like it is already quite something.

00:15:10When I think about when I started with IT, we worked with text screens.

00:15:16Then we celebrated that we got a UI, be it Geos on MS-DOS or Windows.

00:15:24And now, what are we doing?

00:15:26Now suddenly we’re all stuck back in text windows in the terminal.

00:15:29And I actually need to say honestly,

00:15:32all the people I know from Apple and so on, I'm sorry,

00:15:35I'm spending more time in the text window than somehow in the graphical user interface

00:15:39and I actually find it not so bad.

00:15:41Let’s see if something pretty comes along again soon,

00:15:43but the developer conferences are coming up soon.

00:15:45Now I said, let me tell you what I can bring to the table as a worker.

00:15:50Not Mistral, but Hacking-Field.

00:15:55No, Hacking-Face, Hacking-Face, I always mix them up.

00:15:56Hacking-Face, Hacking-Face, you can, in simpler terms, access AI models and run them locally.

00:16:04For example, image generators, video generators, voice cloning.

00:16:08And I built a worker that creates images for me.

00:16:13I built a worker that creates podcasts from texts in Notion. And you have

00:16:23your stuff in Notion and then you hand the worker to the agent

00:16:28and then the agent uses the worker to create an audio file with your voice and

00:16:34everything locally. And it knows nothing without any cloud. And that is already, that's quite nice,

00:16:40Yes, so I can also run an M5 MacBook, yes, we are the friends of Notion and that

00:16:45Apple, we have already glorified that, and then you stand there and think, okay,

00:16:51everything might take a bit longer than with 11 Labs or Nano Banana, but I

00:16:57would say, for the fact that it costs me neither tokens, no, it costs me no tokens and

00:17:04it’s basically done in the background, it's quite fascinating, if

00:17:10you think about the fact that, okay, I couldn't do this with Notion before. I could do it

00:17:14before also with other AI systems, yes, I mean, my third favorite hobby or

00:17:19so is skills in cloud code and so. But it was impressive to see and then

00:17:25I also thought, if you are someone who has your company in Notion, who,

00:17:33I would say, in creative unprofessionals. Who could that be, yeah. Who could that be,

00:17:38I don't know. And then you connect that and basically, I don't want to say that you can automate everything with it or not, yeah, but you have these tools available for people who might use them with an eye for creativity and aesthetics and then receive these modules, these artifacts, these results in the system, in their operating system, in their, whatever you want to call it,

00:18:08offered, and that's just the tip of the iceberg. Yeah, that's just the tip of the iceberg.

00:18:14Just to briefly understand that. So Hackingface is a provider that does,

00:18:21that any of these whole models like... You can subscribe there.

00:18:25No, no, no. Those are proprietary models. Those are completely proprietary models,

00:18:29that you can consume at Hackingface. There are also options for subscription and free and

00:18:35commercially free and commercially for payment. There are a lot of different things. The funny thing

00:18:40is that you also get models there. For example, I recently downloaded a model

00:18:45from Microsoft, where you can throw audio files in, and it then does

00:18:51person recognition and transcription on the audio file. And you just let it do that on

00:18:56run locally from your box, which is of course much cheaper than if you say,

00:19:01at home against cloud models, and of course now with something like Worker then

00:19:06is also simply to integrate. So this hacking face thing that generates the podcast for me,

00:19:13that took about half an hour and then it was up and running, that's pretty funny.

00:19:18Then I can normally use, as the agent now mediates between Notion and

00:19:23the Worker and the agent, I call it up and then I have this tool. And

00:19:28the tool runs on my computer, and if I want to deploy it now, that is, if

00:19:34it should be used by other people, then we have to think about which computer

00:19:38it runs on.

00:19:39Exactly, then it has to run on another one, it has to, as they say, it should

00:19:43run on another computer, which is quite relevant for my workflows now.

00:19:48that's fine with me.

00:19:50I might not have an M5 like everyone else.

00:19:54Please.

00:19:55no matter if you were Pro Max or not, you do need a bit of RAM, you do need a bit

00:20:01not CPU now, but graphics card or AI accelerator like in Apple's processors.

00:20:07I don't know, if you get a DJI Spark from Nvidia, like a box that

00:20:13you can put in the office and address from your television, that should

00:20:17actually work wonderfully too, because the worker is just told,

00:20:21there is the stuff, address the model and then the model is there. So that

00:20:26shouldn't be a problem, it won't be a problem, I just haven't

00:20:31done it yet. And these models that Hacking Face provides are

00:20:38indeed models that are also part of others, like Ulama or, so that's

00:20:44somewhat similar or what's special about it, because otherwise I haven't

00:20:47understood the difference yet. Ulama is a language model, so you can

00:20:50plug in language models to do text chat. And the models you get at Hacking Face

00:20:55are not specialized in generating AI in the form of, I say, write

00:21:01me a poem about the Eiffel Tower, but multimodal models, generate me videos,

00:21:06analyze videos for me, okay. That's basically the same, one is just

00:21:11for text-based stories. Yeah, okay. So I think I understand

00:21:15it now. And that means you have now connected this Hacking Face and can

00:21:19now somehow use all sorts of models in the worker.

00:21:23You can tell which models from Hacking Face are available with which

00:21:28runtime environments those models work. Not everything works on one

00:21:34Mac, some things really need an Nvidia, some only run on

00:21:38Nvidia not on the Mac, some only on the Mac, not on Nvidia, some everywhere.

00:21:42Some models are bad, some are good, it's a huge selection, you have to

00:21:46still look a bit at what is currently, I would say, currently relevant. Basically, I would

00:21:51say that the photo, video, and audio models are very close to what you might like to get out of it,

00:21:58so Voice Cloning, it needs 15 seconds of my lovely voice and can imitate me very well

00:22:05in different languages. And that's already, so I have to see what the model

00:22:13was called. Hacking Face is just the delivery form and the model itself

00:22:18we need to bring up. But it's definitely worth looking into because, yes,

00:22:23I can only repeat myself. It costs you no tokens, you have no

00:22:26data privacy issues, because it just runs on your device.

00:22:29You just need a box with a bit of

00:22:32RAM and a bit of muscle that can execute these models, and I admit

00:22:38I can tell it's running because, for me, the fans can definitely get loud.

00:22:41What you’re running can be quite audible.

00:22:48Well, and I mean, what’s important for our target group as a company is many medium-sized

00:22:57Companies in Germany are of course also reliant on the fact that there are also non-American

00:23:03models that need to work. So this and that, so they have relatively restrictive

00:23:08requirements, as they have set for themselves or must adhere to, especially when you consider that the suppliers might also be for

00:23:17High-tech companies and so on. And they say, okay, please, what is my path now if I can't use all these great American models?

00:23:26And for that, it's obviously great. Especially, it's great that you can, in the enterprise program, already also in Frankfurt.

00:23:31hosted, and then you somehow installed it somewhere on your computer, whatever infrastructure is needed.

00:23:39Then you have a model that might create images or whatever you're currently saying.

00:23:43And suddenly you have a solution that just a few days ago seemed unthinkable, exactly this scenario and many, many others.

00:23:51And this naturally makes it more open and turns it even more into this AI platform a bit.

00:23:58If you look at it from a user perspective, then you can actually say,

00:24:02okay, if someone sets it up for you a bit, if you don't feel like doing it yourself,

00:24:07then a company like us or some other company takes over that, then you can.

00:24:11you there directly this whole, you have direct access to this whole wonderful AI world,

00:24:16so without even bothering to engage with it a little bit.

00:24:19And if you then have the desire to use a locally installed graphic tool,

00:24:25that is, a model for graphic creation, then it suddenly works. That means,

00:24:30it is so, although technically speaking it may only be a small step, but the

00:24:34Options are now virtually limitless and before you were at the end. You have to

00:24:38say that, yes, if you want to work with Norrish and AI, then you will always have to use these

00:24:42American models, otherwise it just doesn't work. They also always natively complement

00:24:47more, even small models. I believe Diepsi can now also include some version

00:24:52And they will continue to do so because the token costs are otherwise always

00:24:55just really high.

00:24:56But the fact that it works right now, that you have built that, makes a big

00:25:01difference.

00:25:02And no matter where I go, all the companies, the medium-sized companies,

00:25:07they want a solution for AI, especially in marketing, where we are currently active, they are

00:25:13always also looking for alternatives and not just saying, okay, cloud

00:25:17install code or cloud install done, although that's somehow the

00:25:20You hold what is currently available, but they are either not allowed or it's too expensive for them, or whatever for the management.

00:25:26In this respect, that's really quite intense.

00:25:28And you were there.

00:25:30Hacking Face.

00:25:31Now I've finally got it out right.

00:25:32What you have is really that you say to him, you give the person the opportunity.

00:25:37It is, although possibly an American model, which then runs with you.

00:25:42Maybe it's worth mentioning again in terms of nuance.

00:25:45True.

00:25:45But you can really do video to text, text to video, audio to text, text to audio,

00:25:50rise of recognition in images, understanding of images, image segmentation, image editing.

00:25:56Sharing, I send an image and you get it back in Photoshop layers.

00:26:02Calculating what basically is there when you pull the layer away, so that there's not a gap,

00:26:08because Dirk is currently looking at the camera and I pull Dirk's layer away, then behind it,

00:26:13is what he thinks should be factored in. So there are really cool things in there.

00:26:19And indeed, because you said it, Claude Cote here would be, let's say, the benchmark.

00:26:25It's like a constantly overflowing circle. So there's Croc and then there's Entropic and

00:26:31then they don't give anything to each other and everyone releases a new model. Maybe at

00:26:36this point just a hint, you also mentioned Mistral earlier. You can also

00:26:41install models locally on your computer whether it's Mistral, whether it's Quen, whether

00:26:48yes exactly exactly and you can also link them again with clotcode, so

00:26:55against this CLI from Entropic you can go and say okay we take now

00:27:00not the opost models from entropic usa, we also don't take the entropic

00:27:06models that might be hosted at amazon bedrock in the eu 200 frankfurt

00:27:11but I’m working against a model that is in my basement or on

00:27:16my notebook and I'm coding with that, but I still have this CLI, code available and

00:27:23just recently you can even Co-Work, so the app you were just talking about, the

00:27:30Mac app, I think it’s called Co-Work 3P, and you can configure it like this.

00:27:37that also works against other models. That you say, okay, I now have my Quen locally,

00:27:44then there is something called a proxy, which isn’t explained in three seconds, but there are

00:27:50guides online, and it is also supported by Entropic. They explain it in their

00:27:55corresponding documents, and then you can run the thing against local models yourself.

00:28:00However, of course, not everything will work exactly as you might expect it to,

00:28:05maybe has the Entropic models in the cloud. But you can also see the manufacturers here,

00:28:10I don’t want to say they are topping up their own business model, but just like Notion says,

00:28:17well, we do want to sell work, but here is an alternative cheaper solution,

00:28:22then suddenly Entropic comes along and says, you find our CLI good,

00:28:26you find our application good, yes, you can also use something else behind it. That is

00:28:30a bit like they all like to hurt each other a little, I think,

00:28:34for fun.

00:28:35Yes, and does one understand the market?

00:28:37Exactly, and when you think about what they all do, I mean, there are those who really

00:28:42invest and invest in models and come out with models like Entropic or

00:28:48Google and so on, and then there are the others who use it, like Apple will

00:28:52also do now, they essentially just use something.

00:28:55What Notion does. And I think what I find particularly good about the way Notion works, even though

00:29:02it's still far from the capacity of Claude, is that it is AI agnostic,

00:29:10he says. So it doesn't matter which model ultimately prevails,

00:29:16all the infrastructure. Where do you have your prompts, how do you build your skills and your

00:29:21factories and how do you put it all together? You have that running, and at some point tomorrow

00:29:26Gemini comes along with then what is indeed Talos, and then you simply switch over, or

00:29:32you run these workers locally. This means the likelihood,

00:29:37that you invest in this abstraction, that it is worth it,

00:29:43in this abstraction platform or level, just like we are doing

00:29:49with Notion, I believe has a high chance of being successful because everything remains the same for you

00:29:54and behind it, some other agent is working away. That also brings us to

00:29:58Topic, what actually happens? This is yet another worker part they brought with them,

00:30:04which unfortunately isn't testable for ambassadors yet, at least not for me,

00:30:08but it's announced that you can somehow, you need to explain this to me too,

00:30:14use agents hosted by Claude, for example. So, you can explicitly say in

00:30:20this worker, okay, this shouldn't be a Notion AI agent, but

00:30:27some other agent that's hosted somewhere, and that should now do this work. And

00:30:33I thought, when that was mentioned, yeah, okay,

00:30:36then there’s no downside to having everything in Notion, in the whole context,

00:30:41all the content, the entire operating system, literally, an operating system,

00:30:47just like it is for us, because I can just wait a bit and then it will be

00:30:52time to access these external agents. I need to briefly mention, I don't know,

00:30:56that you saw in the video, you have a Kanban board and then you make a status change

00:31:00from, let's say, human in the loop, I'll say, release, right, approval and then comes

00:31:06a small animation showing some external worker with a little icon and a

00:31:11small movement-animation working on that task. And when it's done, there's

00:31:17a checkmark. That means you see on your to-do manager, so my colleague is working

00:31:22now, the Claude agent, somewhere on this. And that's really well done. So

00:31:28I found that very impressive. Entropic announced what you described some

00:31:35time ago and mentioned at the announcement that Notion is one of the

00:31:41first to support this and you can use Entropic's cloud infrastructure,

00:31:48where AI agents built by you run. So this isn't now,

00:31:56that you're programming something in a deterministic sense, like the workers themselves,

00:32:01where you're writing TypeScript and it really makes one plus one equal two,

00:32:06but you can build things together with skills and prompts and MCPs,

00:32:12which then react to events depending on what is needed, working in a meaningful way.

00:32:23And the advantage is, they are really small encapsulated, internally focused, which means

00:32:31small, small values.

00:32:32They are encapsulated areas of expertise that do what you want them to do at the time,

00:32:38that are always ready to go.

00:32:42At Entropic, there are the usual token prices for the computation plus a very

00:32:49small opulus that you have to pay on top. Entropic also operates this infrastructure

00:32:55with special sandboxes, so that the thing doesn’t break out, that it doesn't go crazy

00:33:03and cause you 5 million euros in costs, even though it should just read the weather,

00:33:08so that it doesn’t, because you give access, wipe your inbox. There are such cases

00:33:12had deleted a production database that we mentioned earlier,

00:33:18and then said, it wasn't me. No, no, it must have been you.

00:33:22I wasn't it. That things like this do not happen. And the beauty of it is simply and

00:33:28straightforward, you don't have to ask yourself the question, just like we did with the image- and

00:33:34sound and video agents. Where do I place my server? How do I handle the registration?

00:33:40How do I ensure that the thing is maintained? How do I ensure that the thing is secure?

00:33:45Instead, you get, in that sense, yes, remote-running agents that do not execute program code,

00:33:53but execute prompts for you, because you trigger them.

00:33:59As far as I understand it, you always have to trigger them.

00:34:05I haven't heard that they have their own timing interval,

00:34:10like every hour, but you basically have to send a trigger every hour,

00:34:14so that the thing wakes up and says, hello, I’m doing something. But that would probably be

00:34:18the next cool addition, when I think about it, you build some mail retrievals

00:34:25on one hand using workers and agents in Notion, and on the other side, you say,

00:34:31so, I have noted a few things, and my research skill, which is supposed to extract details

00:34:39like in Plexity, just rebuilt, could then, possibly, do fact-checking

00:34:45and quality control and preparation. You can also tell such agents how to

00:34:50respond, in which file format you would like it, like formatting, for example,

00:34:55always a headline and a subtitle and please everything separated by commas and no

00:35:00idea what else. And I would claim that would round off the picture really well,

00:35:07because this way you can also integrate the non-deterministic world automatically.

00:35:12But what I don't understand is, I can use an Opus 4.7 from Notion and

00:35:19let's assume the output is not a file, but the output is some text,

00:35:24so something, text work, images, so, this can be natively Notion. What is the advantage,

00:35:30then, of using a cloud agent in the cloud? So, what can it do differently, if

00:35:39you just leave out that part? You know what I mean? So, why should I have an

00:35:44external agent? I think that’s great at first because I think the world uses

00:35:49cloud and also Notion, but I understood that you can combine it now. But

00:35:53why should I do that if I can choose? Is there any advantage,

00:35:57using Cloud 4.0, the Opus or whatever, natively from the cloud rather than

00:36:04from Notion.

00:36:05You can delegate work steps and it rarely has to be done on your computer.

00:36:09So if I don't know, I’m doing software and now I'm checking software on

00:36:13GitHub, for example, for code management and I want to document it.

00:36:18Evaluate that for me, create an onboarding document or a what has been

00:36:23done document or something like that, then I can do that with GitHub Actions.

00:36:26So, GitHub actions execute, it can be static code, it can be the

00:36:32request from AI, and if it’s a public GitHub project, it costs me nothing, if

00:36:38it’s a private GitHub project, I have to pay for it. And such things can you

00:36:42let Entropic do. It basically does something with things that are online accessible

00:36:49without you having to have your computer on. That much point and simply, AI will still

00:36:57work when everyone around you is asleep. That if you, I don’t know, whether that’s

00:37:03now a case, yes, the customer presses on a feedback form here, I’ll send

00:37:09my issues away and the thing already evaluates it before they are there and sends

00:37:16the answer back. The agent in Notion would do that too if

00:37:21the researcher is. Yes, only that the agent exists in Notion and Entropic has also

00:37:26other clients besides Notion. Entropic also has users who not only...

00:37:32Not the other way around. So the possibility that has arisen since the

00:37:36development or announcement of the Notion Developer Platform, or the

00:37:41one that is coming soon, is that external agents, as you described,

00:37:44can be seamlessly integrated into Notion. I don't have to program anything,

00:37:49I can simply configure it in some way and then

00:37:52it works. And in this demo, I saw that I am in a task and this

00:37:57task maybe has some information from the customer on our

00:38:01homepage that still shows the old product and the old

00:38:07product image, please swap that out, and then it can somehow be passed

00:38:10to this external agent who would do it, he would check out the

00:38:14code from this homepage, change it and check it back in. I understand

00:38:18that. What I don’t understand is, well, Notion does that with the agents just

00:38:24the same way. The agents certainly differ from these personal agents, where

00:38:30you have this chatbot in the bottom right corner, simply by the fact that they

00:38:33are hosted by Notion and just always run. You can share them, if you

00:38:39have shared the agents, you can share them just like any page

00:38:42with 20 people or with a group, then everyone can use this agent.

00:38:47It's not just related to you.

00:38:49And I haven't understood this difference yet, why one should use the cloud agent

00:38:54and not the Notion agent.

00:38:58Yes, phew, exciting question.

00:39:03Whoever knows that is welcome to leave a comment, but as far as I understood

00:39:09it.

00:39:10The agent is part of the user interface and a function that runs in Notion and the Entropic

00:39:16Manage Agents, or whatever you call it, yes Manage Agents, is basically the underlying cloud

00:39:23infrastructure that is offered by Entropic. I'm not even sure if these

00:39:29agents from Notion do not even run there. How much you need that as a user,

00:39:38I couldn’t really tell you, you would probably have to spend some more time

00:39:44on it, I do see a certain gap, a certain gap there, but if I

00:39:49were to think about it, maybe you’ll prove me wrong, but this Notion agent, it

00:39:53does things like creating text, maintaining databases, analyzing Notion pages and a Managed

00:40:02agent at Entropic can work outside of Notion. It can address things outside of Notion.

00:40:10It can maintain permissions both outside of Notion and in Notion. So I believe,

00:40:20it becomes interesting simply when you say, okay, I have things,

00:40:24which I can’t manage with a worker and an agent in Notion alone,

00:40:29or maybe I don’t even have Notion in the intermediary since I say,

00:40:34I connect my email inbox to Spotify and any results only interest me later,

00:40:41but not during this initial call that you get further help there.

00:40:47And of course it is the case that if it’s about really writing code based on

00:40:53a trigger that comes from Notion, then it’s clear, in my opinion,

00:40:58Notion can’t do that.

00:40:59I don’t know, maybe I just haven’t understood it yet, but I don’t believe so.

00:41:02This means that if I then have this external agent, who knows my entire system,

00:41:07that somehow understands GitHub and then knows that at this and that point I need to

00:41:12check something out to do something, which Notion alone couldn't do.

00:41:15So there are definitely enough use cases when you expand the formats.

00:41:19But exactly.

00:41:20I was actually just thinking, is there really a difference if I use the

00:41:23same format.

00:41:25Probably there isn't, it's more about how it can interact with other systems

00:41:29differently.

00:41:30Claude for example or then also others.

00:41:33Yeah, okay.

00:41:34And the agents can really work for a long time.

00:41:37Yes, so these management agents can also handle really long tasks, well you have to

00:41:42pay for it, but they can map long workflows where I don't

00:41:47know where the time limit of classic Notion has been lifted, yes, from that side.

00:41:57And how would one, if I may ask briefly, just a little overview, how would one

00:42:01deploy these agents in Claude, so how, just like the horses, how does that work?

00:42:08Do I just tell the thing, yes, turn this skill into a managed

00:42:12The last time it was good, it's probably still the same today, but when I created my manage

00:42:19agents, you had in the, there's a cloud console, it's not the

00:42:24command line, but a website, where you can get API tokens, API access to the cloud

00:42:30and stuff like that, and there are also manage agents and you can piece together

00:42:36your manage agents through a Q&A game.

00:42:39Yes, you can say, what do I want? You can directly give it prompts for the skills or be guided through a question-answer mirror to this managed agent.

00:42:50And you can also upload results that you create in Cloud Code, in a sense, and say, here, do something with this, I've already made detailed thoughts about this, there’s something on the plate, I have, as I said, skills, I have no idea what.

00:43:03Then you run them there and also do something like cost control.

00:43:09It says okay, how often does the little creature do that, if it's actually running seven times in 20 and you didn't want that at all.

00:43:16And that works through this interface and as I said, you could upload and provide many things there and as far as I understand you can

00:43:25That is now also triggered from Cloud Code, directly triggered from the CLI.

00:43:30and that you somehow have to copy-paste into this web interface, and therefore it would be

00:43:38my initial assumption, maybe only a nice UI can be thought up,

00:43:44but if you wanted to use this today, I would almost bet that if you had an

00:43:50account, you could already use it today. Because I would say, you just need

00:43:57an API token, you need the accesses, you need, that’s all you need today,

00:44:01you can do the same thing with your Gmail. Only, it's not called Gmail, but

00:44:06Entwrappel. Yes, I believe that too. Well, I mean, to summarize a bit again,

00:44:11the vision behind it is to have a relatively easy-to-use

00:44:17platform like Notion, to perhaps make this AI era a bit simpler for

00:44:25individuals, for me. I believe that's actually their mission

00:44:31and I think they are succeeding. Sure, I said at the beginning, the CLI topic, there was

00:44:35I was still dependent on help, but basically, the

00:44:41path they are taking is, in my opinion, the important one, namely that basically

00:44:46everyone is enabled to build their own tool and that I no longer have to purchase or rent a

00:44:52big finished tool somewhere, but rather that

00:44:58I can piece it together step by step, I would say, with

00:45:02really simple means.

00:45:03It's really fun for someone like me, who isn't a real developer, because suddenly

00:45:08all these things work and you also can't forget the whole chain

00:45:13of requirement engineering and this and the specifications and whatever comes up

00:45:17all goes away when someone like me can just build it myself, much to the frustration at times of my

00:45:24people, who then somehow think that the weird old man is up to something again.

00:45:28made up.

00:45:29But I don't think it can be done any more efficiently than integrating all these things.

00:45:33And that's really, I have to say, that's really true.

00:45:36So I think I mentioned it last time as well, for me that's really

00:45:39a tool that captivates me just like Apple has continued to do with

00:45:44all its products. I’m curious to see if it remains that way.

00:45:47I'm curious how you will come out of your shell at the next WWDC now that you

00:45:54have this deal with Google regarding models, whether you will gain some momentum again

00:46:02and show a little something from this new world, because we both notice it now,

00:46:07whether we like the CLI or not? I started here by saying that in the past

00:46:13we were happy with beautiful user interfaces. Today, you voluntarily retreat into

00:46:19these command-number interfaces, so that people like you and me suddenly start coming up with ideas far removed from our

00:46:26professorship or far due to our age,

00:46:30because we simply, well, you can turn any idea into a skill, into a prototype, into

00:46:37something and talk about the results, instead of spending weeks trying to figure out

00:46:42whether the button is on the right or left, suddenly you have 20 versions in the room

00:46:46standing and can try them out and know much better, does this work, does it

00:46:51not that, what emotions do you have, no idea, what else, yes, I would like to

00:46:55not put a term into the world or use one that doesn’t come up in my daily

00:47:00work life, but still you notice that this, this playing,

00:47:06I also recently said, it’s like a text adventure. Back in the day, you played Ultimate

00:47:10Online or whatever, and you said open door, go through, watch out, zombie,

00:47:16kick it, run, yeah, grab the color bombs. So today, you actually write,

00:47:20I mean, that would be a really great function, let’s try it out and oh come on,

00:47:25I’ll press enter now and then go to bed and suddenly it’s three

00:47:28hours later and maybe you’ve achieved something but are still tired and have

00:47:34eye strain, she played the game of life with your professorship, also quite funny.

00:47:38Yes, AI as a text adventurer, my good friend David is a neurologist and needed a

00:47:45really secure billing and management tool for his new private practice,

00:47:52and he didn’t want to buy such an expensive tool and any 1000 certificates

00:47:57And then we said, well, okay, let’s build something with Markdown files as storage,

00:48:02a persistence layer, MacOS in X-Code, with cloud, of course, we talked about

00:48:08that before.

00:48:09I learned that from you with these Markdown files.

00:48:11You know, and then we built that in three hours, he has his own

00:48:15computer, he has no internet, no Wi-Fi, nothing, everything is down,

00:48:19old even, well, not entirely old, but sort of.

00:48:22And on it runs this software and I set it up for him, cloud

00:48:26this 20-euro version, whatever, and he’s doing exactly what you’re saying,

00:48:31namely at the end of the day he clicks three times again and then he’s back to three hours not

00:48:34slept, because he now has the doctor's letter in there and these X-rays and that

00:48:39automatic billing, and he just has to keep saying what he wants. And that is

00:48:43just somehow done. And then you have to know the infrastructure. And that is for

00:48:47non-developers always the biggest hurdle, the infrastructure. And if that is clarified,

00:48:51then it’s just amazing. And fundamentally, if we think about what we’ve

00:48:56talked about today, Notion and Ducktape and what word should I have put in my mouth.

00:49:02They started as, what, a notebook, note-taking option, note solution, and then somehow they got the hang

00:49:10of supporting databases and you name it, business processes

00:49:14and now suddenly, I mean suddenly, they are developing into something,

00:49:20this whole complexity of routine tasks, of maybe creative

00:49:28topics of whatever, is being taken away, because now you have sat down and done something with

00:49:34the CLI, you don’t have to have everyone do that.

00:49:37It’s enough if two or three people do it, who provide the workers, the agents

00:49:40and then suddenly everyone can work with this new possibility,

00:49:44never have to see anything of it, or maybe the job description changes at that point.

00:49:49Another topic. But a tool that you’ve had for years, which people are used to bringing

00:49:56updates again and again and then maybe something new, suddenly fundamentally changes

00:50:01the way you work with it. I don’t think two or three years ago you would have thought

00:50:04that Notion was capable of something like that. And I think that’s...

00:50:08Yes, really often. Yes. Sorry for interrupting. I interrupted you.

00:50:13But especially when I think about this fast-paced time you have, I mean, today

00:50:17something is cool and in three months the next one comes around the corner and how they

00:50:20challenges and how they pull each other through the ring with new features, possibilities,

00:50:26paradigms understood. I mean, today, yes, you said yourself, models are good and

00:50:31they are exchanged and the runtime they are in, the tap is, it is basically

00:50:36the new gold, yes, how it stores knowledge, how it imprints the model, how it takes the abstraction

00:50:42from you. And who knows what it will be in half a year, but I somehow

00:50:45feel Notion is a very quiet but a very modern companion of this change.

00:50:53So, do you remember what you wanted to say? Yes, I think the, the wiring is going to be,

00:51:01I wanted to say that, you didn't know what you were doing three years ago. I knew

00:51:03three years ago I wanted to have everything in one system. I have always wanted that

00:51:07in a way. And I would rather have an integrated system

00:51:11like Apple than a decentralized one, where every program has its justification and then

00:51:17the best tool for the job. I always found that personally and for my company stupid,

00:51:21because I wanted it to be integrated and I didn't know yet,

00:51:24that this would become so important one day. And today, it is indeed the case that because

00:51:27everything is really included, I have the best possible context and the thing always knows

00:51:31everything already. And I can do financial analysis, bookkeeping now always brings up the things

00:51:36because I don't feel like connecting it with Datif. That works too, but

00:51:41I haven't. And it just brings them up, then the agentless one looks at it,

00:51:46checks for any booking errors and so on. But I can also use all this information

00:51:50in another context. So the funny thing, and that's what I wanted to

00:51:54get at, was, I just wrote on LinkedIn that we accidentally invented a marketing operating system

00:52:00because we didn't have this great strategic plan. And then

00:52:04when AI is ready in 2026, we will have this great tool, rather it just happened that way.

00:52:09coincidentally turned out this way and now we have this and now we can really tell our

00:52:13customers, look, we have all of this in there, the strategy,

00:52:18the corporate design, the playbooks, all that stuff and of course we can pass on a part

00:52:23of this advantage to you. So we don’t have to keep all the AI advantage just for ourselves

00:52:29we’ll pass it on to you because we know that if you have the tools

00:52:33for the gold rush, you also sell quite well, it’s a pretty good business model,

00:52:37you don’t always have to mine yourself. And that’s what people want now. And that can

00:52:40all be done with one platform. I can invite the iZone Ocean. And I already have

00:52:44the first pilot customers doing this. And they can then use AI without having to

00:52:49deal with it themselves. Simply as a service from us. We don’t just

00:52:53do, you know, we don’t just do all the things we talked about.

00:52:57have, but for the customer it is completely, they don't have to do anything. Just like a

00:53:02CMS used to be, they can just sit down, write a ticket and then something happens.

00:53:06And either that's done by an agent or a person or both, and that is really quite

00:53:10impressive.

00:53:11And that has nothing to do with the fact that we have a certain degree of kinship.

00:53:15But I have also, and I won't name names, had dealings with various other companies

00:53:21like that.

00:53:22Also, I say, well, what does it mean to deal with, right?

00:53:24So external companies that offer their services to us, and when you look at that,

00:53:30you deal with KÖI. And how they are changing, I mean, I didn't learn this

00:53:35overnight either, but also how they are changing, and you certainly are too

00:53:40you have to carry your packages, but because this is in Notion,

00:53:47the manufacturer Notion offers you a solution that doesn't just give it to you,

00:53:54but makes it easier for you to reuse the knowledge. It is already there,

00:54:00to access, to ask the agent to extract things, to ask the agent,

00:54:05to do things without you having to manage PowerPoint for Bird Axel, different directory structures,

00:54:11countless, I don't know, whatever, yes, file formats, Axels that crash out of the manuals

00:54:17and work with macros and preferably also with color coding to highlight the uniqueness of numbers,

00:54:23what might all be possible, but what I want to point out, the basis, the

00:54:27foundation, the groundwork is already there and Notion is also evolving in that direction,

00:54:34they are integrating it, we talked earlier about the type of data processing and there are

00:54:40enough companies that may have a great consulting offer and smart people,

00:54:44they all have, yes, I mean, we are all those who think working with smart

00:54:48people everywhere, I would sign that, but they may have due to

00:54:52the tools they have used. It used to be Lotus Notes. Now it’s mostly

00:54:57a lot of Microsoft stuff. Even if it’s Apple stuff, well, you know,

00:55:01but I don't know. Yes, but with Notion, I think you have a good foundation, at least

00:55:06also in this changing time with N8n and Make and then came Codex and Cloud and Crock

00:55:14and I don't know what. Yes, it provides a base on which you can build to offer

00:55:20more value to customers, employees. And I actually think that’s really cool. Hats off,

00:55:27yes, it's really a great thing. And that also keeps me in a positive mood. In this respect, I think,

00:55:35okay, they have Markdown at their core. They have AI integration at their core. They have this openness now

00:55:41with the worker and so on to the outside, that even if in three months or in two years or in a

00:55:48yes, a cool new trend comes along, they have the opportunity because of the

00:55:52good data storage and the openness to at least go along, if not even to hold your ground.

00:55:58And I actually think that's getting overlooked a lot, because most people hear, oh,

00:56:02Entropica has shipped another update. Yes, oh, they want adult chat and

00:56:08no idea what it was, it’s all just, bleh. So, and Google, ah, now they’ve, now Google

00:56:12has crushed everyone again and no idea what it was,

00:56:15Coop-pilot was never good, it’s more like cold, parentheses steroids.

00:56:19You never hear anything from Notion.

00:56:21But if you engage with Notion, it’s awesome.

00:56:25Yes, it always was like that.

00:56:27When Apple reached a point where they were already good but nobody knew them,

00:56:32because all the things were set in motion.

00:56:36I believe that this comes at a decent pace, because it is already a large pace.

00:56:41They are now a thousand people, that’s not much,

00:56:43But for them, of course, it’s an insane growth; they somehow came from, no

00:56:48idea, from 300 or so and 1,000 sounds ridiculously small for what they achieve in

00:56:53output.

00:56:54But they are one of those new companies that don’t need that much headcount to

00:56:58create this output.

00:56:59And if you watch all those videos on how they work with their own tool

00:57:03and of course they have 1,000 agents running there and doing everything with it

00:57:07and every Slack channel has an agent from Notion that first answers it already

00:57:12and listens a bit, and if someone wants to know, I’ve also built 2 or 3 now,

00:57:16then he says for new people, for example, yes, answering questions in Notion in Slack based on

00:57:23the context in Notion. Well, what I wanted to mention briefly is that you can really say,

00:57:28and I find that to be the big task now, especially when you have it in Notion,

00:57:33you can as an employee relatively cleanly and in small steps enter this AI world,

00:57:41because then you somehow do a prompt and then that's what we call

00:57:46now a skill and then you can just say add somewhere and then you provide the skill name,

00:57:50then something happens and suddenly, you need a completely new interface,

00:57:54some new system, you have to copy-paste into a chatbot or or or, it's just there

00:57:59in the dose you want and it just becomes increasingly visible to you that

00:58:04our project is precisely to bring all the people on board. I don’t want to come to the end

00:58:11without saying one more sentence about that. Because we also repeat this regularly in the podcast,

00:58:16the technology is there. Just because technology is developing quickly doesn’t mean that

00:58:21the technology as we have it today or as it was three months ago is bad.

00:58:25It can do a lot of things and it doesn’t fail because the technology was bad or has been bad

00:58:30for some time, but rather that you also have to take people on the journey,

00:58:34because the majority still hasn’t heard anything about AI, except maybe in the news. Those,

00:58:39who have heard something about AI, may have partially used it

00:58:43or think they have, but essentially they’ve never used it more

00:58:47like Google.

00:58:48After some input, an answer came, thanks, goodbye.

00:58:50And very few pay for it, even fewer interact with skills and stuff or get lost

00:58:55on the CLI and then use, I’d say, AI in its pure form and from that

00:59:00side, it's about how to take people along without overwhelming them on the one hand.

00:59:06Because too full windows, too many buttons, typical IT stuff, you have to click there, why do you click

00:59:13here?

00:59:14No, you do that and then please go into the submenu and then enter that, now

00:59:16just go ahead.

00:59:17Yes, so it’s like this, everyone in their Jaguar might be a little impatient with people

00:59:24from someone else’s Jaguar, that’s the point of explaining.

00:59:27And from that perspective, what you said is of course also a point,

00:59:31if you are used to working with scores, which surely for the person who

00:59:35is switching from Nauschen-Kentenumstieg, you can introduce it homeopathically. And that’s surely not to be underestimated

00:59:46and also shows, if that perhaps was never the point of why one dealt with Nauschen

00:59:50that it wasn’t so stupid after all that you use it. Switching is,

00:59:56when you are on something different, is quite hard. You have to get people on board,

01:00:00you have to pull in your data. So the bigger you are and the more people you have,

01:00:04I believe, the harder it is to handle such a different user paradigm.

01:00:15But I found it great again that we talked about Notion. It will surely

01:00:24happen again. I have this very strange feeling that Notion will eventually have an update

01:00:28soon. At the latest then I would allow myself to invite you back. Maybe

01:00:32I’d like Jens to be there too, it’s probably a full excuse that he’s crashing.

01:00:36Yes, maybe it’s just an excuse, but I would also like to meet him.

01:00:41So, greetings to you, as you say so beautifully.

01:00:43And it was very nice to be at your place again.

01:00:45We are at one hour and two minutes.

01:00:47I think this is the target time.

01:00:48Yes, something like that.

01:00:50That works plus minus.

01:00:52That fits very nicely.

01:00:53Yes, thank you very much for the invitation.

01:00:55And for the explanation in advance.

01:00:57So I am now a proud owner of already two Workers.

01:01:00The absolute hammer.

01:01:02If you liked it, Schinte, leave a like, a comment, tell your friends,

01:01:05no matter if they work with or without grades, that you give our podcast a go and stay tuned,

01:01:10to see who our guest is next time.

01:01:11Until then.

01:01:12Ciao.

01:01:13Bye.

01:01:14Welcome to Think Different, Think AI, the podcast by Mark and Jens.

01:01:21Two technology-loving minds who not only talk about artificial intelligence,

01:01:26but live it.

01:01:27Here you will find clear classifications, real practical insights, and a fresh look at what is possible.

01:01:34Understandable, critical, and always with a wink.

01:01:38AI to ponder, to smile, and above all to discuss.