Name: GitHub Copilot for usage-based budgets
Uploaded: 2026-05-06T15:30:55.282Z
Duration: 52 min 36 s
Description: GitHub Copilot for usage-based budgets

Transcript for "GitHub Copilot for usage-based budgets": Hello there. There's a new system. I'm just getting my head around it. Thank you all for joining. I think we'll just give people a couple of minutes for everybody to get in and then we'll kick things off. For those just joining, I'm just giving people till, probably about two minutes past, and then we will kick off. So thank you very much for joining. Just give us a few seconds. Okay then. I think that's probably enough time for everybody to get in. So I hope you're all having a lovely day. I'm Greg Shisorth, a solutions engineer at GitHub. Thank you very much for giving me your time today. I know the session's in for an hour. We'll see how we get through. It might be a bit less than that. Hopefully, I can give you a bit of time back today. It's a lovely sunny day here in The UK. I hope wherever you are, you've got something similar. What are we looking at today? It's a bit of a hot topic. This is the new usage based billing for Copilot. Specifically, this session is more designed around looking at some of the new budgets and spend controls that we've got. If we look at all the things we're going to cover today, feels like a lot and I'll be honest, there's a lot of detail in here. It might feel a little bit like we're sat in a math classroom at times, so I apologize for that. I'll try my best to make it as interesting as we can. Some of the things we're going to cover, how Copilot billing is evolving and some of the changes that are coming on. We've got some really cool news around a new auto mode improvement that I want to share to help drive a bit of efficiency. We'll look at how, there's a new shared usage consumption model and that actually has quite a big impact on how usage based billing works. We'll dive into that. We'll then look at the budgets, how you can control spend, and then we'll take a look at a bit of more of a deep dive at some example scenarios to give a bit more of a real feel to some of the content we're covering today. Another point, there should be a Q and A panel somewhere you can see, I think it's on the right hand side of your screen. As we go along, please any questions, pop them in there and I've got a couple of my wonderful colleagues on the call in the background who are there to answer those questions. And first off, I know if I was watching this, one of my first questions would be will there be a recording I can watch later? Absolutely yes there'll be a recording With that being said, what's happening and what's changing? Like I say, we're calling this usage based billing. Really the aim here is to change the Copilot billing model to a way that better aligns to the compute and resource usage of the various Copilot features under the hood. Starting on June 1, the premium request units, so that PRUs that you'll know that we use today. They're being replaced by a new currency that we're calling GitHub AI credits. These credits consumed on a token based usage. If you're not familiar, concept of tokens, it's a way that AI can measure, it's almost like English language, machine version of sentences and words. Really, I would just think of them as like counting the number of words used in AI. It's a little bit different, but it's about there. The new GitHub AI credits are consumed based on a measure of token usage within those AI features you're using. That includes input tokens, output tokens, and a concept of cache tokens as well. It's important to note similarly to how it is today, those token usages are also scaled based on the model you're using. Smaller, more efficient models are just as today, cheaper to use than larger, more advanced, more computationally intensive models, which are a bit more expensive. In terms of the timeline, April 27, the announcement went out. You should have all received that and I'm guessing part of that's the reason why you're on today. Throughout May, we're launching a billing preview. You'll start to see that popping up in your GitHub billing page. Then like I said, this all goes live on June 1. The timelines are fairly tight on this. A little bit of a look at the reasoning behind why this change is happening. Like I mentioned, this is all around moving the pricing to better reflect token based pricing, which really is the industry default now for AI. We found that increasingly the request based pricing hasn't been well aligned to the computing costs and particularly that's increased as we brought on more and more advanced Copilot features and you've got requests that can take multiple steps in a genetic ways and potentially run for minutes. Potentially, I've seen people even manage to get hours out of a single request. It's quite cool. But essentially, that was becoming more and more separated from the realities of the compute costs. It also provides us more flexibility to bring in new models and bring in new features. This whole concept of the new AI credit currency helps us with that. Then a really cool bit, which really is nice to have, this allows us to also bring in larger context window sizes. Up until now, some of the context sizes haven't been exactly the same as what you would get from the model providers directly, That's been down to partly down to the pricing model. Because we're now matching the same token usage that the model providers are doing, we can pass on those same complex window sizes to you as well. Fundamentally, all this change is about changing copilot usage to better reflect your actual AI usage. A little bit of the nuts and bolts under the hood. We looked at how you've got tokens. That's how AI consumption is measured. This counts both tokens in the prompt, your response, and again, that concept of cache tokens. It should be day to day, you probably won't worry too much about that, but at least as an engineer, I find it quite interesting to have a look and see what's going on there. We're introducing two new concepts in the billing side and this is included usage and additional usage. With your copilot plans, you get an amount of credits with each plan, and I'll show you those numbers in a minute. Those numbers all go into the included usage pool and that's pulled across all users. We then have the additional usage, which is once the included usage pool has been exhausted, you can then, if you want to, continue to use like a page you go model on the additional usage and continue using above and beyond the credit value you get. It's worth mentioning as well. We'll go through the budgets, but a key point here is this is all measured in US dollars and one AI credit is 1¢. $1 is a 100 AI credits or a 100¢. I mentioned we'll look at the how many tokens you get or your AI credits you get with your plans. The pricing is not changing. Your per month pricing is the same, but now you get a matching dollar amount of AI credits for your monthly plan. Remember again, those credits go into a pooled model and we'll dive into that a little bit later. Worth mentioning as well to help smooth over the change here, we've got a promotional offering. This is for current customers. We're bumping up the included usage value as you can see here. Copilot business, dollars 30, Copilot enterprise, you're getting $70 off AI credits value and that's through June, July, and August. A few things that are staying the same. I think it's just worth calling out. Like I said, the base plan pricing is the same as it is today. Also what we're calling the core developer experience, which is code completions and next edit suggestions in your IDE. They are included in all plans and won't be consuming AI credits. Now, a few things that it's important to keep in mind. Today, there are some models that have a zero cost associated with them, which means that if your users have used up all of their PRUs, they can fall back to that zero cost model. Under the new model, those are no longer available. All usage with the two exceptions I mentioned on previous slide, all usage consumes AI credits. The other important thing to keep in mind, code review. Now that will also use actions minutes. Up until now, we haven't been counting the actions minutes when it comes to billing, but because code review runs on top of actions, those will also now be counted as just standard action minutes that are consumed. I've spoken a lot about how we're now measuring usage and so now usage becomes far more important and top of mind. I wanted to give a brief rundown of the things that we're looking at that can impact that usage. I think probably one of the biggest ones here is the model that you choose. Like I said before, some models are small, not very computationally intensive and consume AI credits at a slower rate. Some models like Opus, much bigger models, much more computationally intensive, and they will consume credits on a faster multiplier. Choosing the model can have a big impact on the usage that you see. Prompt size, again, that's directly related to the input tokens going into the model. The bigger the prompt and that also includes context that you provide to the agent, so how many different files you're asking to include in that context. That will increase the prompt size. Response size, again, counted in the similar ways. Prompt size, that's response size. The longer it is, the more tokens you're using. Context and cache, again, similar. The more context you're putting in, the more expensive, more usage you're getting there. The agentic workflows, if you've got a long process that's going off doing its own thing, making changes, doing a multi step process, each one of those steps is consuming and producing tokens, which again is translated through to your AI credit spend. Then finally, MCP skills, using those skills are all again, agentic workflow steps and consume credits. That's just a rundown of when I talk about usage, what do I actually mean? Now, some really cool stuff, auto mode. Up until now, auto mode has really been focused on just looking at how you can sending your requests to models based on model load at that time. Really, it's been focused on making sure that you get a really good experience. You're prompting a model that's currently available and not experiencing your epic load and falling over a little bit. What we're bringing in now is more intelligent routing. We're looking at having auto mode, look at the prompts you're giving it or the task you're giving it, understand the complexity of that task. If it's a simple task, that will be routed to an appropriately sized model, so a smaller, more efficient model. If it's a really complex task, then that's going to one of the more advanced models. At least for me, this is really, really cool. When I'm using Copilot, I'm hoping my boss isn't listening today. I just go for the biggest, best model I can. I don't want to think about, you know, am I choosing an efficient model? And frankly, I don't really understand the models, you know, well enough to make a well informed decision on is this small model good enough for the usage I'm trying to put through it or do I need a better model? Auto mode is now going to do that for me. I can just click auto mode, forget about it, and have my requests automatically optimized to whichever model makes sense for my task. We're targeting this to come out June 1 as well. As with all roadmap items and everything I talk about here, always subject to change. But that's what we're targeting. For me, this is really quite exciting. Moving on to budgets. Like I said, this is where it's going to become a little bit feeling like you're back set in a school maths lesson, but we'll see how we go. The first really key concept here is the new pooling system. We mentioned you get per user, per license gets a dollar value of AI credits with that license. Those credits are now all put into a single enterprise pool and this enterprise pool is across all your users. If you've got 10 users on Copilot business, that's 19 of value they're getting each. In the pool, 10 users, dollars 19 each, that's $190 of credits to be consumed across the enterprise. Now the really key thing here, and you'll see this as we go through some of the budgeting options, All users, when using Copilot, draw credits from this pool first. That's really key. The idea that until that pool is empty, all of that usage is coming from the pool. Once the pool is empty, that's when we're moving into additional spend. That additional spend so that you're only going into that when you've used all credits from every user. What's really nice about that is in any organization, you'll have some users that are using Copilot really heavily, getting a lot of value out of it, and there's some users that will be using in a more light fashion. That means that under the current model, the users that aren't using Copilot that intensively, their credits at the end of the month are just gone. They haven't been used, it's gone. With the new pooling, those credits can be utilized by your more heavy users. Hopefully, basically you can spread out the usage draw on your credits. And hopefully you'll see a much more efficient use of those credits, particularly for those people who aren't using as much of their Copilot license. Now, how do you start controlling this? It's worth mentioning, so as we go through this, I talk about budgets and part of this is spend control. Part of this is more around like fair use and making sure that there isn't a single user that is using all of the pool. The budget layers that we'll go through, now I'll go into each one of these as we move through. But you've got the enterprise level and the cost center level and then the user level. The enterprise and cost center level, they only impact additional usage. Those only come into play as budgets when the included pool of credits has gone. The user level, that covers both pool usage and additional usage. They just cover all of the usage per user. How do these budgets actually work? As I mentioned earlier, these are all dollar value budgets. The enterprise level, this is a budget that goes across your whole enterprise. Really, I think of this as a top level additional spend control. If you set this to zero, no additional spend can happen at all. Once the pool is empty, all users are blocked, Then you have set it to anything above zero and then that's how much more additional spend can incur before users are blocked. If you think about the total number of credits available to all of your users, it's the included pool plus the enterprise additional spend budget that we're looking at. Put that into numbers. If you've got 100 users with Copilot Enterprise, they're each getting $39 of AI credits. That pool is 39 times a 100, dollars 3,900. You've then got the additional spend. If you set that to say $2,000 the total spend is those two numbers summed together. You've got $5,900 of additional spend. The key thing here, enterprise budget is additional spend across the whole organization or enterprise. If we drill down a level, you've got the cost center budget. This is, again, very similar to the enterprise budget. This is the maximum spend that a single cost center can incur or maximum additional spend a cost center can incur. Again, this only kicks in once the pooled credits are all gone. There's an additional caveat with this one that you can actually set these budgets up. The settings actually in the enterprise level, but you can set it up so that the cost center level can ignore the enterprise budget. There are certain circumstances where that can be really helpful. It starts to make things quite complicated to think about because you're starting to override other budgets. I want to call that out. But I think for the purpose of this session and most users, you probably won't need to use that setting, but it's there if you need it. That's the cost center. We move through now looking at the user level and there's a few different options you've got here. The user level budgets, these are controlling the total spend of each user. This is purely about usage now, all usage. The one thing you're going to want to set up is a universal default level budget, and this is new. I think of this as the fairness control. This is a single budget that you set up that covers or applies to every user. This is a usage limit you put it in and really I see this as preventing any single users from doing messes of usage and starting to hit your enterprise level budgets and blocking other users. Think of it as just making sure to think like everyone's behaving themselves, but I need to find a better and more positive way of saying that The second one is the individual user level budgets This is basically an override for the universal level budgets and this is on a per user basis. Let's say you've got your universal default set for everybody, but now you've got a set of power users that know what they're doing. They're really awesome with AI and you want them to be able to spend more credits each. This is really what the individual budget is for and you set these on a per user basis. If I've got a 100 power users and I want to give them all a bigger budget, I need to set a 100 individual budgets, one for each of those users. If you've got loads of users you want to do this for, there's an API available. We're also looking at how we can make this a bit easier and we have a planned enhancement. We're still working on exactly what this might look like, but I think we're aiming for basically the ability to assign an individual use level budget to every user that's within a cost center or group just to make the manual tasks a bit easier here. There's an important call out as well. I know this won't apply to most of our customers, but some of our really big customers in GitHub, you only have a maximum number of 10,000 budgets that you can apply. And these are the budgets we're looking at here are just the same budgets, so they come out of that same 10,000 limit. Bear that in mind if you're setting loads and nodes of individual use level budgets, you might start to eat into that limit. Just something to bear in mind. That's your enterprise level, cost center level and user level budgets. If it's like me, when the first time I was looking at this, I was like, okay, well, that's a lot to get my head around. How does this work in practice? I pulled together a bit of a flow that might help. When each of these budgets kicks in, really it's like a veto, an override for all of the budgets. The key rule here, lowest remaining headroom wins. If we run through this, there's two phases that you need to have in your head. It's what happens when there's still credits in the pool and what happens when the pool is empty. We have a user, they go and use Copilot, the budgets are then checked to see if they've still got credit budget left. The first step is, do they have headroom left in their pool? If they do, then the enterprise budget and cost center budget are not relevant. They only come into effect when there's no room left in that pool. Headroom left in the pool? Yes. Happy days. I'm going to go check my user budget. That's the default or the individual budget if I've been granted one personally. If it's all good, approved. I can use Copilot. Let's go back to the beginning. User uses Copilot. In this instance, there is no headroom left in the pool. All the credits have been used up either by me or other users. It's gone. Now we start looking at the additional spend controls. We go and check enterprise budget. Any in that? No, I'm blocked. If we've still got enterprise budget, I can continue. Is there any headroom in the cost center budget? Again, as long as we're good, I can continue. If not, blocked. And again, we will then go and check the user budget. Hopefully, all is good. I can continue using Copilot. Really, the point I want to get here is just that to make it clear how these budgets all work together in tandem. And they you know, we try to make them as flexible as we can so you can set this up to do whatever makes sense for you. Okay. Now fair usage. This is a really interesting one to to talk through, and I think depending on sort of the views of your organization and the policies you've got in place, you'll approach this in a different way. But fundamentally, each one of these budgets are all just about putting maximum limits on AI credit usage for either users or different groups. The key here is there is no reserve or partition for the AI credits. No one place is going on. Like if I have a cost center and I've got a cost center budget of $1,000 that's not guaranteeing that I have $1,000 to spend in my cost center. I could still be getting blocked by, say, an enterprise budget. Then also, and I'll dig into this in more detail later because it's a bit weird to get your head around, but how you use the pool, the included pool also impacts how much usage you might get in a cost center. We'll dig into that in the scenarios. We've looked at what tools and what levers you've got available. Here's just a little bit on if I were in your shoes and I was looking at, okay, well, what do I want to set my budgets to? Just a few things, a few thoughts that I might be looking at. And and really it's like, do I know what my power users are using and who they are? How much are we currently using to get an idea of what would use might make sense. Like, in an ideal world, I want to prevent sort of irresponsible or malicious use of my AI credits. But beyond that, I want my developers to be able to use the tools, you know, in the way that makes sense for them. So getting an idea of, like, what what's there and what's available is really important. We've got a tool for that as well. Again, same thing, but on the cost center level, often certain elements of a large organization will be placing different levels of focus on AI usage and spend. And then the other one that I think is going to become increasingly important is how you approach enablement. In AI, generally, I see that we've had it's been really exciting. It's been really awesome to watch AI progress. But we're moving into a phase where actually optimization of AI is also going to become important. Bringing enablement to your developers around that can turn be important. We are looking at bringing out some more resources around optimization as well. Watch this space, you should see those coming out. The other key thing that I think is going to get really interesting here is, you've probably seen this already, but the budgets, they're visible to the users so they can see how they're progressing against their user budgets. And I think you'll see that driving behaviour as well. How that drives behavior, I think is something that's worth bearing in mind when you're looking at that enablement. Again, a big win here is auto mode. The new auto mode is a really set and forget optimization that hopefully will be pretty useful. I think a breather, that's a lot of content. Grab a glass of water. I hope you're staying with me here. I know it's a lot to go through, but there we go. Next up, I want to give you an idea of what this will look like for you and what the timeline looks like. One day before is what I called this, but this is today. What you've got, you've got your existing PRU budgets and that's there to control your additional spend. Those budgets are there today and available for you to use. Now what we look at, if we move through to June 1, what's going to happen? Those PRU budgets that you've got are going to be transitioned automatically to cover AI credits. If you think that fundamentally these are just dollar value limits on additional spend. They will then apply to your AI credits as well. This isn't something that like on June 1 you really have to rush and put in place. You can look at doing this beforehand and I'm sure the vast majority of you have already looked at putting some budgets in place already, so they will still apply. You'll have the same AI usage dashboard. This will give you a view now into the full entitlements across all of your licenses and the usage towards those pools as well so you can see whether that pool, how the usage of that pool is going, which will help guide where you should be putting your budgets as well towards the end of the month. We're also updating the billing UI dashboard to include the AI credits and the ability to drill down on per model and per user. But the key point here is those user level budgets we spoke about are new. So June 1, they won't be there. So bear that in mind as well. The way I'd be looking to approach this, June 1 and maybe a little bit of this beforehand as well. Look, start with visibility. You get an idea for how much you're using, who's using, however much Copilot, get an idea of who those power users might be, get an idea at a budgeting level, what you want your enterprise guardrails to be and cost center budgets, Then go for that default user level budget. Put that across everybody. Probably want to set it fairly high to start with so you're not upsetting people. But again, that's now your fair usage control. Go and set those individual budgets for your power users or wherever it makes sense. Then look, this is going to be an iterative process of seeing how these budgets work for your organization and changing them as time moves on. We've seen a lot of customers are adopting AI and copilot at a really rapid rate, which is fantastic to see, But that means that this might be fastly, it's not a word, a quickly changing scenario. The other key points, I started with saying, right, the first thing you want to do is start with visibility. The question naturally goes, how on earth do I start looking at what the usage is? We're building a dashboard that will let you get a good idea of what the change from PRU billing to usage based billing will look like. This isn't quite out yet, but it's in the final stages. I would expect this to be coming out fairly imminently. The way this will work, you'll see this in your standard billing page. You'll be able to go and download an export file, CSV, of your usage for a billing period. You can then go and put that into this tool. Again, you'll see this in your billing dashboard. Then in this tool, you can go and drill down into like effectively, it's this say this is what happened last month. How do those costs reflect and that usage? How does that reflect with the new billing system? You can also drill down into your users, the models you are using, where in your organisation the usage is. It's a really neat tool. We appreciate that this is a change and we want to be there to support you with this. This is one of the ways that we're going to make this as straightforward as we can. Okay. That's a whole lot of theory. What I want to do now is take you through a couple of scenarios to look at how these budgets work really in practice. Let me build that slide out. I've got a couple of scenarios and they're all fairly similar. The setup is I've got 10 users. They're all on Copilot business. Each one of those gets $19 each and that goes to my shared pool. My shared pool, dollars 190. I've also set an enterprise budget at $310 My total potential AI credit spend, including shared pool and enterprise budget is $500 As you'll see, that really handily divides by 10 to $50 each. I've set a default universal user level budget at $50 In a perfect scenario, in this example, everybody has used their $50 of usage. They've hit their universal level budget and we've also exhausted the shared pool and we've hit the enterprise budget. In an absolutely perfect world, the last second at the end of the month before this all resets, everybody's used all of their AI credits and they've all used them equally. Now you might be saying, well, you don't really want every user to be blocked. I would agree with you. You probably want to have some of these budgets a bit higher if this is the usage you're seeing. But just for the purpose of example, it just makes it simple to say everybody's used all of their use level budget and everyone's had the same usage. Skip forward. I'm now wanting to look at what happens when I put a cost center in there. This might be a cost center of just power users, it might be a cost center that just wants to limit the spend in a different way. Whatever it is, I've got my cost center and I've inventively called it cost center A. Cost center A now has a budget of $160 Again, in this example, we're halfway through the month. Everybody's used $19 That $19 is the $19 from their Copilot business license that went into the pool. Everyone's used the same amount, but that means the shared pool is exhausted. Gone. Which means that any additional usage is coming out of my additional spend budget. That's the cost center budget and the enterprise budget. As we saw in the previous example, the enterprise budget, everybody can use $50 each and when they do that, they hit the enterprise budget. That sort of all fits. But in this example, we've got the cost center budget. If we look at that, so everyone has $31 left to use, which gives the cost center budget 31 times five usage. What's that? A $155 The maximum amount at this scenario that the cost center can use, dollars 155. That means we're not going to hit the overage spend or the additional spend of $160 for that cost center. In this example, that cost center budget isn't doing anything. Now the reason why this is important is because we're now going to look at what happens when the shared pool is not used equally. I mentioned earlier that that can have an impact on how these budgets work. If we move forward, let me build this slide out. In this example, the cost center budget is going to start coming into effect. Where are we here? We've used the shared pool. Everyone's used an average of $19 so the shared pool is gone. But here, the users that are outside the cost center have used more than their fair share of the shared pool. They've all used $25 each. In the cost center, everyone's used $13 each. Really what this means is, so that shared pool has been allocated in a way that some skews towards the users that are not in the cost center. Why this comes into play is now because the shared pool is exhausted, additional usage comes out of additional spend. But now each user in the cost center still has $37 of usage to use. 37 times five, dollars 185. Now that's more than the 160. That means that the cost center budget is now being hit. The only difference between this scenario and the previous scenario is how the bigger bite of the pie cost center A have had from the shared pool. I really wanted to pull this out because it's a bit of a tricky one. At least I found it a bit weird to get my head around to start with when I'm looking at what I should be putting my budgets to. It's just understanding that that shared pool is always used first, always consumed first, and depending on who uses that much first, who has the most AI usage to start within the month before you get into additional spend, that can impact how these budgets work. That's the scenarios. I hope that cleared it up a little bit. At least gave you a feel for what you might be able to do with these budgets. The next thing I want to look at is very quickly how you might want to approach this from a cost allocation or like a chargeback model. If you've got multiple business units and you're wanting to charge them for their AI copilot usage at the end of the month. Because of this, the pools model where you don't have an amount of that pools credits ring fenced for your organization or your cost center. You don't want to be relying on the cost center budgets to be doing accurate chargeback. I would approach this would be at the end of the month, go download the export of your Copilot usage and use that as a basis to perform your chargebacks. Fairly straightforward. Or at least it is fairly straightforward for me to say that actually doing it, you have to plug it into all kinds of systems. I see how this can work. But the point is go grab that CSV export and base the chargebacks on that. We're getting towards the end now. Again, thank you very much for bearing with me here. Hope it's been useful. If I was looking at what I will be doing as next steps, I'd be looking at, go review your usage patterns, see how you're using AI at the moment, try and get a feel for how that usage is growing over time as well so you can start to think about how much, sort of headroom, wiggle room you need in any budgets you might be putting in. Think about up to the point now, this is a bit of an admin task that requires a bit of thoughts put into it. Who's going to be monitoring that and managing it? Who's going to make those decisions around budgets? Who needs to guard rails? Who are your power users you want to unlock and go be free to do whatever they want? All those things, have a think about that. Then again, going back to that enablement, if you want to start looking at optimization of AI, go look at reconsider auto mode. I suggest talking to developers about that when it comes out. Again, that should be June 1. I'm going to look at how AI and token usage works and understand what makes a complex prompt that can consume a lot of tokens, what makes a simpler one, and how you might manage that. Again, these are just some suggestions. It'd be interesting to see what you folk think. Finally, I just want to do a call out. I'm going to be running this webinar again. I know this has been very popular. We've got a huge number of you on this webinar, which is lovely to see. But for any of your colleagues or if you think this might be useful, please do let them know. We're running this again May, that's British UK time and that's the link if you want to grab a screenshot or something of that. That's the same link you would have used to sign up for this webinar. Feel free to socialise that with anyone you think makes sense as well. Now that's everything I've got for you today. Again, thank you very much for your time. I hope that proved useful. I appreciate it's a lot of new concepts and feels a bit like, at least to me, it feels like a maths lesson. But that's great. Just a question to Colin who's been in the background managing the Q and A with Anton. Are there any questions that came up that might be worth trying to run through in the last ten minutes or so? Is there any anything that's points that have cropped up? I can see you're talking, but I can't hear anyone, which is interesting. Hang on. This is me trying to use the system. They're backstage, so I can hear them, but I had them muted. Sorry, Colin, for the purpose of me. Is there anything that, is worth going over? That's good. That's good. I I I Sorry, I appreciate Colin for your benefit. Only I can hear you. Apologies for that, I'm learning this new system. Colin told me massive amount of questions. Really great to see, so I appreciate the interactivity. I think they've covered a lot of the topics. Question I saw pop up. There will be a recording of this shared. We'll send out an email afterwards. The other point that's worth just mentioning is please do reach out to your GitHub account teams. They'll go talk to them. They'll present it. This is a big change. We're very much here to support you in any way we can. Cool. Okay. I think with that then, unless there's anything else, we can wrap up early and I can give you ten minutes back. Lovely. Okay. Oh, I know. I can see a few questions coming in. I will hang around here and see if I can answer questions as we go. Thank you very much for your time, folks. So I think will the slides be shared as well? The slides won't be shared, just the recording. K. I can see a question about auto mode discount. It's a good question. Currently there is an auto mode discount of 10%. We will be continuing an auto mode discount. The percentage that will be going forward with that is yet to be confirmed. That's another one of watch this space. We'll see where that one goes. Lovely. I think we will leave that there and I appreciate that there's probably a lot of questions that still need some answers. We'll go through those and see if we can come back async in a way. There's a very high number of people in this webinar, so please do bear with us. I hope you have a lovely rest of your afternoon. Thank you for your time.