[MUSIC] [MUSIC] All right, welcome everybody. Day three, first talk on the Bits and Bombers stage. How's everybody doing? [APPLAUSE] Yes, still a bit tired, but we have nice weather today, not too much sun, so you can fully enjoy the talk by Joseph Devo Geis. He will talk briefly for about 25 minutes on the impact of software on energy and resource consumption and give us an overview of the environmental harm and the potentials for savings there and how free software in special is well positioned to address the issue. And in the end, we will have a Q&A session. And, yeah, if you have any questions, you can raise your hands, and we will get a mic to you. So give Joseph a warm round of applause again and enjoy the talk. [APPLAUSE] Yeah, thank you. Thank you very much. Thank you to the organizers at Bits and Bomber. This has been great so far. So the title of my talk today is Free Software, Software Design for the Environment. And this is part of a larger KDE eco-initiative working on sustainability issues in KDE and free software. Before I get started, though, I like to make my talks a little bit interactive. Feel free, if when I'm talking you have a question that's burning and you want to ask it right away, you can just raise your hand, and either the mic will come to you or I can repeat it for the audience. Before we get started, though, I wanted to ask you, what are some words and phrases that you associate with free software? And I'll repeat here for the people who are online. Great. Great, yeah, good. Empowerment. Empowerment. Powerful. Say it again. Powerful. Powerful. Customizable. Customizable. Collaboration. Collaboration. Collaboration. Great. Sustainable. Sustainable. So most people don't say sustainable, and I hope by the end of this talk all of you, next time someone asks you that question, will say sustainable. I have many links in the slides. If you want to download them, they're available at our repository. I will also return to the slide if you remind me at the end of the talk. Just wait a second. Great. So to get started, just to frame the issue that we're talking about. So this is a post on Macedon from Ed Hawkins. He's a climate scientist at the University of Reading, in which he has a graph. I don't expect you can really see the graph, but I hope you can see that all but one have an upward slope to it. Okay? You can probably guess what these are measurements of. This is the global temperature, CO2 emissions, sea levels. There's only one that's going down. That's the ice thickness. And he writes in the post, "Changes are emerging across the climate system. Everywhere we look, the climate is changing rapidly. These changes are not normal. They're caused by the burning of fossil fuels." Fossil fuels are also being burned in the ICT sector. So information and communications technology is right now an area that's being researched to figure out how much energy is consumed and how much the ICT sector contributes to CO2 emissions. This is a graph, again, all having upward slopes, which is a meta study looking at various research papers trying to estimate what the CO2 emissions are of the ICT sector. One thing I want to point out here is that the estimates vary. You can see some grow exponentially, some grow linearly. However, they're all growing. And if you take it, sort of try to average over what all of the studies seem to show, it appears that ICT sector contributes between 2 and 4 percent of CO2 emissions globally. To put that into perspective, that's roughly the equivalent of the global aviation industry, which is estimated around 2.5 percent. This analysis for the ICT sector includes everything from production, transportation, end-of-life treatment of devices, to energy consumption of the Internet, to energy consumption driven by the software you're running at home or in your office. Also includes Bitcoin, machine learning, training, etc. Right now, the current trajectory, if you look here as well as here, if things don't change, it's going to continue growing. And some estimates put that up to about 30 percent of global CO2 emissions by 2050. So it's growing. It's increasing rapidly. One thing I want to point out here, and I like this quote from this, it's a tech brief from the Association of Computing Machinery, computing can help mitigate climate change. How? By substituting other activities that we're doing. But it must first cease contributing to it, because it's growing in some cases faster than it's substituting. I'm going to focus today in this talk on the different ways that software is driving energy consumption specifically. Of course, energy consumption does contribute to CO2, depending on what the power grid mix is. I'm not going to talk too much about estimates of CO2. I'm going to talk specifically about power consumption, as well as production costs. I'm going to do that in two different ways. I'm going to talk about efficiency, energy efficiency, in terms of achieving the same result when doing the same task, but consuming less energy, as well as conservation, that is reducing the waste that's driven by software. And let's look at some examples of this. First, in terms of efficiency. This is a comparison of two word processors. It's from the Umweltbundesamt. It's some research done by the Umweltkampus Birkenfeld, in which they compared various software products doing the same thing, and then looking at their energy consumption differences. Word processor one, the green bar plot, is a free and open source software product. They don't mention which one. The blue one is a proprietary one. Again, they don't mention which one. And as you can see here, to do the exact same thing, to achieve the same result, word processor two consumes about four times the energy. Okay, for an individual, this may not make much of a difference. But you can't think of software's energy consumption on an individual basis. Word processors are in every office, in every university, on everyone's computer. People are using it globally all the time. If you think about these numbers when you scale it up, they add up very quickly. And to give an example of that, this is from a talk from Hasso Plattner Institute, Detlef Thoms, in which he does a back-of-the-envelope calculation, in which you reduce a process by one CPU second. Okay, it's a very minimal change. That's about the equivalent of a savings of 10 watt seconds. And if you multiply that by 1.5 million users, now that's not very many users, and you assume that there are 20 such reductions a day, so this reduction in the CPU usage happens 20 times a day, and you multiply that by 230 working days a year, you end up with 19 megawatt hours of savings. Now, if you're like me, that doesn't mean much to you. To put that into perspective, if you take a modern electric vehicle and drive it back and forth from Shanghai six times, that's about the equivalent of 19 megawatt hours. Okay, so these are very simple changes that result in big savings once you scale it up. And the thing about software is that you're not only changing on one computer, but you're changing every computer that gets that update, right? And when you think about it on a global level, these savings really add up very quickly. If I can convince just 500 developers-- and I've been giving similar talks now for about a year and a half, so this is a very realistic goal-- 500 developers to try to achieve just 10 of those efficiency gains, multiply that by 19 megawatt hour savings, and you're talking about the energy consumption of 30,000 two-person households in one year. So scaling it up, you get to really big efficiency gains. Okay, now conservation. Reducing the waste. And when we're talking about waste, let's start where the waste lands in the landfill. And this is a graph--infographic representing the e-waste from 2016. Already one year before that, Akim Steiner from the UN commented that there's a tsunami of e-waste rolling out over the world. In 2015, there was 44.7 million tons of e-waste produced. That's about the equivalent of 4,500 Eiffel Towers. If you were to stack those Eiffel Towers, it would be about 17 times Mount Everest high. Okay, this is in one year, and it's growing. In 2019, the World Economic Forum said that e-waste is the fastest growing waste stream in the world. In 2022, 60 million tons of e-waste, a 33% increase. Of that e-waste, about 20% is estimated to be recycled. And in what ends up in the landfill, although e-waste accounts for about 2% of the total waste, it's 70% of the toxic waste. What does software have to do with e-waste, right? So, e-waste is hardware. Software can drive e-waste. Some of you may have had this warning message that your device is no longer supported. If your software is tied to that device, then that device becomes unusable, at least not without serious disadvantages to you as a user. The software might become more and more bloated as more features are added, and the energy that's consumed or the processing power that's needed by the hardware is so much that older hardware can no longer support that software. As a result, what you have is devices that are now produced and transported unnecessarily, and functioning devices end up as e-waste in the landfill. This is particularly problematic, thinking about it in terms of CO2. This is a graph from a book from Lange and Santorius called "Smarta Kuhn-U-Neveld". This is data from Apple's own sustainability reports from an iPhone in which they estimate how much CO2 is produced per device and then what it's attributable to. They find, Apple reports, that the production, end-of-life treatment and transportation accounts for about 80% of the CO2 emissions of a device. So, sending a device into the landfill before its time is a huge waste of resources. So, to me, they're both the same thing, right? I mean, if the hardware doesn't allow the software, it's because the software can't function on that hardware. So, what could the hardware vendors where to open or provide support? Right, okay, so the question is, so things like bootloaders being proprietary and then not allowing devices to then run other software, right? That's a design decision. That's maybe not directly a software in the sense of... It is a software decision because that's proprietary. You can't then access the device. So, I would say, I mean, software and hardware, those are the...what you're describing, so the question was about when the hardware is so locked down that you can't install other software on it. And that's a situation where you have a device that's tied to one particular software product, right? But that's a design decision and it's still related to software. And I would say we should all be fighting for the right to repair, to include software, including things like opening up bootloaders so that you can install other software products on it. So, yeah, production and end-of-life treatment and transportation are going to contribute to much of the CO2 emissions of a device. But it's not only CO2 emissions. The end-of-life treatment also contributes to polluting air, soil, et cetera, often in countries in the global south, both where the devices are...the metals are mined, often under horrible conditions, and then they return there to then end up in treatment centers or in landfills. This here is in Ghana. So, I mentioned at the beginning that there are cases where digitization can substitute other activities, and one of the questions there is is, what is the relative harm difference between the digital product and the non-digital one? This is, again, from Lange and Centorious. It's a comparison of e-readers in terms of their environmental harm and books, paper books. Paper book production also contributes to environmental harm in terms of polluting water and soil with the chemicals used to create the paper. And the production of an e-reader contributes also to environmental harm in terms of the different materials that go into it, often toxic. Per device, 300 liters of water are required, and one estimate has CO2 emissions at 170 kilograms per device. And Lange and Centorious estimate that if you read about 30 to 60 paper books or more, at that point, the environmental harm, it's worth getting the e-reader from what you're contributing. This is one estimate. These are very hard estimates to make. The point being that these are things that you have to consider. Substitution is not in itself always better. In the case of e-readers, statistics about how many books people read per year in comparison to how long devices are supported. So one Wikipedia article had the start date and end date for production of e-readers as 1.5 years on average. The average US-American reads 60% read less than five books per year. So to get to the number of 30 to 60, it's going to require using that device for a long time, which likely won't be the case given the short-term lifespan of these devices. So one of the things I want to point out there in those parts is software design in terms of the energy consumption, as well as its support for hardware, can influence the environmental harm of the ICT sector. And this was the motivation for eco-certifying software. Many people ask, well, what does software have to do with eco-certification? That's what I'm going to talk about now. The Blue Angel released the award criteria for software in 2020. The award criteria are aligned with FOSS values of transparency and user autonomy. And there are three main categories to the award criteria-- resource and energy efficiency, potential hardware operating life, and user autonomy. Just to get an overview before I focus on some of the user autonomy characteristics, which are inherent to software license with a free software license. Before we get there, though, let's look at some of the different criteria-- resource and energy efficiency--that's related to the energy consumption of software. The criteria don't have a threshold, but rather a transparency requirement so that users are aware of how much energy this software consumes when running typical tasks. Potential hardware operating life-- the requirements are that it runs on hardware at least five years old. I think this is a little too low, especially given that many Linux distributions run on hardware much older than that. And the user autonomy requirements include things like uninstallability, modularity, which influences the sort of software feature creep and bloat, continuity of support, documentation, so that there's long-term reuse and repairability of software, transparency so that users can decide which software products to use so you're not locked into an inefficient program, as well as offline capability and freedom from advertising, both of which consume energy. Just some background, the Blue Angel Oculus from KDE is the first Blue Angel certified software. Currently, it's the only one. I hope to convince at least some people that it's worth at least meeting the award criteria if you're interested in the topic or considering applying for ECO certification. So here, getting to the characteristics about free and open source software and how it relates to energy efficiency and resource consumption. So looking at the criteria, it's hard to single out a single one when it comes to the topics of efficiency and conservation. They're all interrelated. But here, I want to focus on a couple of them together in three different parts. So first, I want to look at the energy consumption and hardware performance related to uninstallability and modularity, and perhaps related to offline capability and freedom from advertising. So we take it for granted in the free and open source software community that you can uninstall software. This is something that is obvious to us, and many of us probably have never thought about that in terms of efficiency gains. But unwanted software components that remain on a system contribute to longer boot-up times, longer shutdown times, perhaps extra processes running when you're using that software, et cetera. All of which, if you go back to the one CPU second reduction, all of which adds up when you multiply that over millions of users. So uninstallability is a way to create a more efficient system. Again, going back to this graph comparing the two different word processors, if you want to take a deeper look at what the energy consumption looks like over time, here's a plot, again, from that same measurement, looking at the word processor, the open source word processor, which is on the bottom plot in comparison to the proprietary word processor, which is on the top plot. And what you can see here is doing the exact same thing. They have very different energy profiles. The x-axis is showing over time while running different tasks, things like opening a document, changing the layout, and then saving the document. And I want to focus here on what happens after that blue line. That's the point at which the document is saved and the software goes idle. And what you can see here is that the proprietary software continues doing things in the background. It does it quite a bit, actually. And in comparison, the free and open source software product goes idle as you would expect. The report doesn't say what exactly is happening in this time period. There is also measurements in this approach where they're looking at network traffic, writing to disk, etc. That data is not in the report, so I don't know what's happening here. It might be telemetry. I think that's a reasonable guess. If it's proprietary software, you can't turn this off necessarily unless the company decides that you're allowed to. If this is free and open source software, developers can then take a look, see what it's doing, and then decide to remove this functionality if users don't want it. All of this is contributing to that energy profile of these two word programs. So, again, yeah, just that this adds up when you start thinking about it, scaling it up across millions or billions of users. Okay, the next topic, so potential hardware operating life, as well as continuity of support, which is what determines how long you can use that device and not have disadvantages opening yourself up to malware and things like this, known malware, documentation, how long can that software be supported over time, will depend to some extent on how well it's documented, transparency that users can decide which software they want to use and then move to different software products if they want to. And this will all influence how long you can continue to use a particular piece of hardware. This is an image from a 2009 Macbook. This end of life was in 2019 where it was no longer supported by Apple. And so the last version of MacOS that was supported was 1010. This is from a KDE contributor who was asking around if KDE users were using hardware that was no longer supported by the original vendors. This is now currently still in use today, running Kabuntu 22.04 with all of the security updates, et cetera, and is apparently working just fine. I know I have this experience. I'm running also older hardware at home on my desktop computer, which is no longer supported running Linux. This device would otherwise end up in the closet or perhaps worst case where it doesn't end up being treated or it goes to end of life treatment center and all of that CO2 that is associated with the production has now been wasted. Okay, and the last section that I want to talk about in terms of the user autonomy criteria is the offline capability and freedom from advertising. This is a report from a research and consultancy group which does policy reports for the EU in which they looked at what is called unwanted data use. And they define unwanted data use as the data for data mining and advertising that users can't shut off. Given a statistic that in the EU, about 60% of users would shut it off if they could. So they call this unwanted data use. And they try to estimate what the energy consumption and CO2 emissions are associated with that unwanted data use. This graph here is just showing that there are several components to energy consumption. I've been focusing on desktop software. That's most of what KDE does. But of course, if you're doing anything that's distributed, client server or similar type systems, you're also going to be consuming energy in the network and in data centers. And this unwanted data use is at best attributable to about 3 million metric tons of CO2, according to this report, at worst 8 million metric tons. And this would be, according to this report, equal to the carbon footprint of about 370,000 to 950,000 EU citizens. This is for something people don't want. So the blue angel recognizes that if users can turn it off when they want, you remove this energy consumption from the ICT contributions to energy consumption and CO2 emissions. Of course, these things are all intertwined. Things like advertising and data mining, et cetera, might also make it hard to run a browser on an old piece of hardware. In fact, I find one of the biggest problems when upcycling hardware is opening a browser the minute you try to do anything on the internet. Currently, website design is so energy consuming that it can break your system. That's related to Roberta's great talk yesterday. So these things are, of course, interrelated. Offline capability for your advertising will influence how long you can run hardware, et cetera. So we wanted to document our experience applying for the blue angel eco-certification so that other free and open source software developers can also do this. We released a handbook in February this year, Applying the Blue Angel Criteria to Free Software. You're welcome to download it from our website, eco.kde.org. The book goes over what I'm calling the three steps to eco-certification, measuring your software. So to do that, you need to define what typical usage of that software would look like, emulate it using a script, which is all documented in this handbook, and then run it several times, run some statistical analyses, which you can do with a tool provided by the Unveiled Compass called Oscar, the open source software consumption analysis in R. And once you have that, you document your user autonomy criteria that you're compatible with them and submit it. The book, though, is more than that. It's something that you can send to your friends if you're interested in this, send it to your marketing department if you work at a company, send to your project manager if you're on a team, in which it goes over various aspects of this topic. The first part is the why. Why does this matter? So related to energy consumption, resource consumption driven by software. What is the blue angel? What's its history? How is it recognized internationally as well as in Germany? And then how do you fulfill then the requirement for software? You can also read it online here. For those of you who are interested in this but don't have access to a power meter, power meters are... So the approach that the blue angel recommends and the one that we're using in our lab is an external power meter attached to a machine that then is emulating user behavior. Power meters are expensive. Really good power meters are really expensive. The off-the-shelf solutions, which are okay, they don't have great resolution. The one that is recommended, the Guta power meter, gets like five measurements per second, which you can then average and get an idea of what's happening. Better ones can get like a thousand measurements per second. However, you're going to be out of three to four hundred euros to buy one. If you're interested in just checking this out but don't have five hundred euros, you can invest ten euros and buy a power plug and hack it and turn it into a power meter that gets about one measurement for every 200 milliseconds. And that's also documented in this. Say it again. I don't know. Can we ask that at the end? Because I also wanted to open up... I usually like to do Q&A where also other people in the crowd, because I don't have a monopoly on this information at all. So if someone here might know the answer to that but I don't know it, ask it again in the Q&A. The lab setup looks like this. You have a system under test. That's the machine that you're testing. A power meter attached to it and then another computer that's then collecting the data and then used to evaluate it. We set up a lab in KDAB Berlin. This is a picture of one of the sprints. This is Arna Tarara from Green Coding Berlin who was very helpful with Folker Kausa, Nicholas Fela and several others in the KDAB community who helped us get this lab setup. The current setup is only accessible if someone is there in person. We're working on that. I'm going to come back to those slides. Right now in a project which should be done by next month in which we are making the lab remotely accessible so that a developer, and we want to open this up to all free software developers, this is not only a KDE thing, can design their usage scenario that they want to measure, upload it with their code and then it will automatically run in the lab in KDAB and then give you a report back so that all you have to do is define the usage scenario and then you get some energy consumption measurements related to your software. And of course all of this is KDE free software so anyone can take it, adapt it to their needs as they see fit. What you would need to do if you wanted to use a lab like this, you have to define a standard usage scenario that requires identifying typical tasks done by a user, perhaps identify particularly the high energy ones to get an idea of energy consumption related to that, create a flowchart of actions and then emulate it. There are many ways to emulate it. Right now KDE is focused on using Selenium, which is often used for web emulation, adapting that for desktop software. There were several talks at Academy just a few weeks ago which you can watch about this approach, but it's also useful for user emulation. Once you have that, then you upload it to the machine you want to test, you get the results back and then there's this tool, Oscar, where you can then upload the data and get a report back. The nice thing is that with this lab set up with a remote access, all of that's automated. Just a little bit more generally, KDE has a current goal of sustainable software. Every two years we vote on three unifying goals for the community. One of the things that we're working on is measuring our software, but there are several other things related to this, such as documentation of the ways that free software, in particular our software at KDE, supports sustainability. Maybe I'll end on this and then we can open up to a Q&A. When you look at the big picture, this can seem like small drops on hot stones. This is perhaps nothing compared to what it's required to train a large language model or the explosion of IoT devices, etc. This is a comic from XKCD, of course there is a comic about this, in which Q-ball says, "I'm trying to fix the problem of the world, can you help?" White hat says, "It's obvious you don't actually care, if you did, we'd be trying to fix this other bigger problem." Q-ball asks, "You want to help me fix that bigger problem?" White hat says, "No, for another reason I'll think of later." The bigger problem fallacy, saying there's something bigger that's a bigger problem than what you're trying to fix, is a fallacy, it doesn't mean the other problem is not a problem. Also, sometimes tackling the biggest problem is not the most effective approach. You have to think about what resources you have available to you. At KDE, we are desktop software developers, we can influence this. I hope I've convinced you that it does add up, once you start looking at the numbers. I'd also say climate change and combating climate change require changes in every single aspect of society. This is where we can contribute and where we can focus our energies. There are areas that maybe you can contribute to. I hope that I've convinced you it's worth working on, because most people here are involved in some sense with ICT, and you can make a change, and you can make a difference. I want to leave it on that note, and then go to Q&A. [applause] Thank you. Thank you so much, Joseph, for that insightful talk. By the way, before we start with the Q&A, just a reminder that if you found that talk so insightful as I did, and learned something new, maybe you can post that out to the Fediverse. The hashtag from this stage is #bitsandboimer. Now to the Q&A. We have our wonderful stage manager, Rico, who can come along and pass you a microphone, so everybody will hear you just fine. As I said earlier, I don't have a monopoly on information. If anyone here has an answer, please feel free to speak up. I feel like this is a wonderful crowd to crowdsource information gathering. Well, actually, it seems to me that this problem is an efficiency problem, more or less. We want to do a task with the fewest resources possible in the fastest way, of course, which is shouldn't we just continue to optimize our code, which is the way to do it, and to remove stuff we don't need. That's basically put it in a short way. Is it this correct, or are there other things also to it? So thinking about things like uninstallability, right? That's beyond just optimization. That's a design decision for how you then allow users to use that software. So I would say it's beyond just this one efficiency. I think that's absolutely important, but you have to think about it also in that holistic sense of can users -- okay, you might make the most efficient software, but can users uninstall it if they don't need it? That's a design decision. I don't think many people think of free software licensing as a design idea, but in many ways it determines how users can interact with your software. This is an example, one, of uninstalling something or only installing what you want. So it's more than just optimization, I would say. My question is that there is a sustainable supply chain law in the European Union and in Germany, Lieferkettengesetz, which if it would be taken serious, you could not import or sell any other phone here except the Fairphone, because other phones would fall under this law, which is prohibiting importing things where parts of it in the production process have violated human rights or environmental laws. So my first question is, is there anybody talking about this in this community, about this not fulfilling this law? My second question is if this Lieferkettengesetz, this supply chain law, could be applied also to software? Is there some idea that you just say it could not be sold if it doesn't comply some minimum standard of environmental responsibility? So I don't know anyone talking about that, the supply chain issues, maybe someone here does. I'll let the moderator decide. If someone has a comment on that, please feel free. Well, I wanted to ask a question that goes into the similar direction, because I was wondering, are there legislations being discussed about reducing the energy consumption of software and to the supply chain law? I personally thought that it's more about the human rights problem, mainly about hardware then, and that then this law would not apply to software, but that's only how... Can you pass the mic so that... So, yeah, so anyone else have a comment on that? Yeah, software in general and Lieferketten, we also have kind of Lieferketten supply chains, we have libraries, we have operating systems, we have whatever, and these things also need to be produced in some way, someone needs to sit down and code them. So it would make sense that it applies there. I'm not a lawyer, I cannot really answer this question, but it makes sense that we also look at our supply chains and how they are ecologically, that we are good or bad. For example, we could choose a framework that is more efficient, we could choose a framework that makes stuff easier for us, and of course, software can also be produced under conditions which are not compatible to human rights. You can force people to sit down and program stuff, and this code would be unethical to use in a way as programmers. Do we have time for another question? Yeah, so what do you have to say about the law of diminishing returns? For example, is there a way to know what's the point when we already reached those diminishing returns and we have to move on to something else to keep optimizing and doing that stuff? Is there some finger rule? I don't have a rule of thumb for when that is. I do think we still have a lot of low-hanging fruit, to be honest, before we get to the problem of diminishing returns. I think there's enough that we could do in terms of designing your website, following green-coding practices, being able to turn off advertising, continuing to support hardware, opening up bootloaders so you can install any software on any device that you bought. These are all things I think are pretty low-hanging fruit, which already make a big contribution before we're even at that point. That's my personal opinion on that. Thanks. What's your opinion on these bloat solutions like Docker or Snap and stuff who actually for a small application ship a whole operating system and thus bloat my disk so that I think next time I need a bigger SSD in my laptop because some applications take up to a couple of hundred gigabytes? Is there an estimation? What's the difference in CO2 footprint of half a terabyte and one terabyte? What's your opinion there? I don't have a particular opinion. I do think it's an empirical question that we have a lab we could maybe start testing to see if these differences in how you install, which approach you take to installing software, if that influences things like general energy consumption. I think we could look at that and see it. I can imagine that there are arguments that upgrades can result in more efficient software that might outweigh the disadvantages of installing it through a flat pack if there is a difference there, but I don't know. We have two more questions left. Unfortunately, then we have to close, but discussions can also unfold afterwards. Maybe you're around here for a while. I'm definitely around for a while. Great. Please. On the CO2 footprint of software, Docker Solutions and something else, look just into the power consumption of Hello World in C or in Python. It's 20 times more in Python to realize a Hello World program than in C or in some other languages. All the time in the back head when we talk about software or something like that, the big bloat solutions who do need Python and bigger compilers, Rust, and whatever has a high power consumption. We have to go in our mind if it's useful to make this all including big systems in software and hardware, or if we do not split it in small pieces, which can be done in a sampler in C and save us a lot of energy. Thank you. Thank you for that comment. Last question, but not least. I would like to highlight that besides efficiency, there's another thing that I would say is rather important and we talk too little about, which is sufficiency using less in general. For example, the supplies to social media or streaming services, being nudged into using those media for many hours a day, which is probably not very healthy for ourselves, but also if we didn't use the software for that long and the hardware as well, we might be happier in the first place and use less resources. Fully agree. Sufficiency is definitely a good question to ask. Thank you so much. I think these are good closing remarks. You want to add anything for the audience? I would love to talk with you some more. [Music]