Open Source Friday: Spec Kit part 2

Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.

Original language: English

Original video: Open Source Friday: Spec Kit part 2 · Channel: GitHub · 1:01:31 · Watch on YouTube

Topics: AI agents, Task and workflow automation, Software engineering with AI, AI developer tools, Open source and software ecosystems

Summary

GitHub engineers Manfred and Nicole discuss Spec Kit’s evolution through its 1.0 release and the newer capabilities shaped by community feedback. These include bundles, custom workflow steps, idea assessment, bug-fixing flows, and convergence to check whether implementation matches a specification. Nicole demonstrates Copilot-specific plugins and canvas apps that provide guided visual interfaces, while the core Spec Kit remains coding-agent agnostic and usable from the command line. The conversation also covers customization, token-conscious workflows, and the value of assessing ideas and asking clarifying questions before implementation.

Key points

Spec Kit had reached version 1.2 at the time of the discussion, following its 1.0 release.
Bundles let teams group vetted extensions, presets, and workflows.
Workflow steps can be hosted and customized, alongside the core steps.
New workflows support idea assessment, bug fixing, and creating custom processes.
Convergence checks whether an implementation adheres to its spec and can add tasks to address gaps.
The core Spec Kit is coding-agent agnostic; the separate Spec Kit Copilot repository adds Copilot-specific integrations.
Copilot canvas apps provide visual interfaces for spec-driven development, idea assessment, bug fixing, and pipeline customization.
An experimental canvas designer can generate a canvas app from a configured pipeline and metadata.
The lean preset creates fewer artifacts—spec, plan, and tasks—to reduce the amount of generated material.
Community extensions and presets can adapt workflows to project or organizational requirements.

Chapters

Start	Chapter	Summary
01:17	Welcome and introductions	Sara introduces Open Source Friday, mentions GitHub Universe, and welcomes Spec Kit engineers Manfred and Nicole.
04:32	Spec Kit 1.0 and community-driven additions	Manfred describes Spec Kit’s growth and new bundles, workflow steps, idea assessment, and bug-fixing capabilities.
09:11	Convergence and MCP support	The guests explain convergence as a way to check implementation against the spec and discuss experimental MCP server support.
12:47	The Copilot integration	Manfred introduces a separate Copilot repository that adds natural-language ways to drive Spec Kit while keeping its core agent agnostic.
17:25	Using Spec Kit through Copilot skills	Nicole demonstrates initializing a project and creating its constitution by asking Copilot in natural language.
24:07	Keeping the core agent agnostic	The guests explain how Spec Kit scaffolds skills for different coding agents and how the Copilot integration layers on top.
26:44	Canvas apps for guided workflows	Nicole introduces Copilot canvases and demonstrates visual interfaces for spec-driven development, idea assessment, and bug fixing.
33:26	The wizard and customizable pipelines	The wizard lets users browse presets and extensions, assemble a workflow pipeline, and try it, including with a pirate-speak preset.
39:08	Experimental canvas generation	Nicole previews an experimental designer that can generate a customized canvas app from the pipeline configured in the wizard.
44:49	Copilot-specific features and token use	Manfred clarifies which features require Copilot, and the guests discuss the lean preset as a way to produce fewer artifacts.
50:32	Architecture and Spec Kit’s role	The discussion touches on Python’s place in the project and how agents use the generated specifications and other artifacts.
53:08	Idea assessment and clarifying questions	The guests describe how assessment workflows can surface missing context and help teams decide whether an idea fits a project.
56:15	Community customization and open source	Manfred and Sara discuss adapting processes with presets and sharing extensions, workflows, and other community contributions.
58:21	Closing remarks	Sara points viewers to the repositories, asks the guests about ice cream, and wraps up the episode.

Quotes

If you don’t like it, make it your own. — 10:13

The core of Spec it always has been agnostic. — 24:07

Spec driven development. — 48:26

it really has now become a process kit. — 56:15

Concepts and entities

Spec Kit, GitHub, GitHub Copilot, Spec Kit Copilot, Spec-driven development, Canvas apps, MCP, Manfred, Nicole Hougen, Sara Kaiser

Full transcript (original language)

01:17 [music]

01:59 >> Woo! >> Well, welcome, welcome everybody to Open Source Friday. Uh really excited. We have some really cool a really cool project today and some really awesome guests that I’m really I am definitely excited to talk about talk with. Um it’s also Friday, so happy Friday. Um but yeah, so quick uh point of note, GitHub Universe is coming up. Uh My name is Sara Kaiser. I work at GitHub.

02:30 Uh and yeah, you can check out um the our our kind of big conference where we’re announcing all kinds of new stuff um from everything from the github.com platform to Copilot, all of our tooling is going to be super fun and you can register for free online. So, um it’s a great way to kind of stay on top of what’s going on at GitHub. All righty. So, I am going to bring on

03:01 our friends for today. Um we will be talking so we uh Hey friends. Uh we talked about SpecKit on our previous episode with Manfred. Um and so today we’ve got uh Manfred who is a principal software engineer at GitHub as well as Nicole Hougen who’s also a principal software engineer [laughter] at GitHub uh to talk about uh not only the 1.0 release of SpecKit, but kind of

03:32 what’s changed since the last time we talked. So, um yeah, if you want to if there’s anything else fun you want to tell folks about yourself, Manfred. >> Well, uh you know, I like cats. I have a lot of cat shirts. Let me actually show you this one. So, uh so uh you know, anyway, um yeah, still loving SpecKit. Uh you know, yes, uh 1.0, but uh me let give Nicole introduce herself. >> Sure. Sure. Sure. Sure. Uh yeah, my name is Nicole. Um I work

04:02 remotely from Minnesota. Um so, I’m about 3 hours from Mall of America. So, just a little fun fact about myself. Um and like Manfred, I’ve been in the industry for a long time, and I’ve been helping out with Spek-It recently, which has been really fun and interesting. So, so happy to talk more about that today. >> That’s awesome. I didn’t know that. I am also from Minnesota. >> Are you? Okay. [laughter] Great. >> Originally from southern Minnesota, and then from the Twin Cities. So. Uh I definitely have been had had a couple birthdays at the Mall of America. >> Yeah. >> [laughter]

04:32 >> It’s a rite of passage. >> Indeed. [clears throat] Indeed. So, uh yeah, let’s um I’m going to pull up the repo so we can chat about this. But, Manfred, I’ll I’ll throw it to you. Um what uh what’s the what’s the latest and greatest on Spek-It? >> Well, as you already uh said, you know, um version 1.0 is out. Uh you know, we’re actually already at 1.2, but uh you know, at on the birthday date, I actually uh released 1.0 because I thought that would be fun, you know. Uh

05:04 we all we all turn an age. Well, Spek-It turned an age, too, you know. So, there we go. Um yeah, it’s been actually quite a wild ride. I mean, since May, uh you know, we have doubled our numbers in stars almost. Uh you know, we doubled definitely the number exte- of extensions that our community is delivering. Presets are in the same way. Um you know, yeah, the star count is going up. >> I saw that you had the plot at the bottom of the read me earlier. >> See that happening. Um

05:34 but, you know, there’s other things that obviously have happened that are of note. And one of the things that uh you know, we we’ve been trying with Spek-It uh is to see what the community actually needs and wants. Uh the biggest thing there is, you know, if you’re using it in a team, you know, the you know, we introduced the concept of presets for that so that you can actually layering your own stuff. But, you know, come you know, folks were like, “Okay, well, how do I bundle this now?” So, there you go. We

06:05 created bundles, you know, so that’s the one of the later editions so that you can actually bundle stuff and bundling allows you to, you know, create you know, uh a vetted set of things, vetted set of X-things, vetted set of presets or even workflows. Yeah, I do want to call out that >> Ah, okay. >> folk. We had done work for it. Now, we have five, so that one is finally getting off the ground as well. And then, previously, you know, during that

06:35 era that between, you know, May and now, we we’re not able to actually host workflow steps, which is the steps within a workflow. Now, we actually have that same hosting platform that all the other primitives has have as well. That makes it so that, you know, obviously can introduce your own steps cuz even though there are 13 core steps, you know, I don’t know what kind of other steps folks would need. So, you do have the ability to do that. Now, as Sarah’s right now showing, the other thing that happened is

07:07 basically saying, “You know what? There are things that happened before SDD and there are things that happen after SDD.” Um SDD is still the big still a very big part of Spec-it, but you know, we also wanted to say cater to the stage before and that is, you know, how do you assess an idea? So, there is a core extension that allows you to just say, “Hey, I want to assess this idea on the merit of how does it fit the project?” And so forth and so on. So, you have the

07:37 ability now to do that in same kind of guide. You know, is the is this idea good, and meaning go, is this an idea bad, meaning kill, or is does do I need more clarification? And that’s basically the end result of that process. But, you know, the the reason why it is there is obviously, you know, uh previously you would have to do a lot more machinery and a lot of work yourself. An LLM is typically capable of pulling all kinds of information together if you just give it access to it.

08:09 So, on the far end >> [laughter] >> Yeah. On the far end we obviously have bug fixing, you know, so bug fixing is not normally a whole SDD flow, uh you know, so there you want, you know, hey, uh assess the bug, fix the bug, and then test the bug, you know, test as in did the fix actually fix the bug. So, that’s basically that one. And then as you can see, number four is uh well, create your own process, you know, and that is where the capability list comes in that we, you know, so if uh Sarah scrolls down there you now you can see

08:41 that. There’s a lot of stuff in there that you can do. Uh you know, uh but that’s all geared towards what am I trying to accomplish, you know, cuz that’s the key thing really with Spec It Now, it’s more kit than spec, uh you know, because [laughter] we went the route of opening it up and we wanted to see what y’all would be doing with it. And so, having that said, >> One second, I saw people in chat asking uh for the first episode uh to to catch up. So, I’ve got a short

09:11 link for you here uh if you want to open this and go back and start from the beginning. >> [laughter] >> But, cuz today we’re we’re talking about what’s changed since then. But, uh yeah, you can check out the previous uh video there. All right, sorry. Go ahead. >> [laughter] >> The other thing that uh uh it’s not a new primitive, it’s uh actually a addition to the core workflow, and that was convergence. Uh looking at all the extensions out there, uh and you know, we concluded that there is there was something

09:43 missing after implement and that was basically saying, “Hey, implement implemented, but did it actually adhere to the spec?” If it didn’t, it convergent will actually create new tasks that implement needs to go do to converge onto the spec, which allows you then to, you know, make sure that whatever was specified is actually adhered to by your implementation. So, that was, you know, came out of obviously all the community extensions that were around that as well. And that doesn’t necessarily mean that some of the

10:13 community extensions are now obsolete in that sense. It’s just a different way of looking at the process. You know, converge does it a particular way. Some of the extensions in the community does it do it slightly different. And which is that’s really the strength of Spec it as well. It’s like, if you don’t like it, make it your own. >> I think that’s a good framework. >> There There is one other thing and that one is highly experimental at the moment. Um And it’s also not listed in the capabilities, that’s why.

10:44 And that is server support. So, the Spec it CLI can be turned into MCP mode. It will slowly actually gain the same [snorts] functionality so that you can then interact with it as part of an MCP server kind of thing. It’s not the core of what we intend to do here, but it is, you know, a way of being able to interact with the Spec it CLI using MCP. So, my anticipation there is not actually running the commands through that MCP server because, you

11:15 know, obviously the commands are still the prompts that the agent needs to execute. So, be aware of that. >> So, I want to ask a question about the convergence thing. So, that’s basically trying to help make sure any new processes are that are generated adhere to the previous adhere to the actual spec. Is that correct? >> Yeah. >> How does that differ from something like

11:45 having prompts or agent instructions in a repo? >> Uh well, the the conversion command is is is a prompt in a in in and of itself, but it’s it it it actually would knows about the Spec It machinery. So, the conversion prompt I’m going to go look at Spec. I’m going to go look at the plan, and I’m going to look at the task and the code base and determine did I overreach? Did I under deliver? Uh did I break something? That kind of thing in that in that particular sense. So, the reason why it’s

12:16 append only to the task is because it needs to convert like, you know, it’s almost like the you know, a loop [snorts] where you go slide make that tighter and tighter and then you’re actually or a funnel or how how yeah, that’s that’s kind of >> [laughter] >> that way, right? So, that’s the >> approach. >> Yes. >> Uh >> Um >> Cool. >> So, now everything that I’ve said obviously is core Spec It. Uh and you all might

12:47 wonder why I’m not calling this here. Well, >> [clears throat] >> one of the things that I looked at Spec It and you know, we’re at GitHub. So, we have access to our co-pilot and co-pilot app, right? Um so, I wanted to think about, “Hey, how can I actually drive Spec It itself in natural language?” So, that’s the first uh thing that I wanted to see because that would actually, “Hey, I don’t have to remember the commands. I can just, you know, either chat or type

13:18 in the in in my in my co-pilot app and then, you know, it will just execute something. So, what we did there is, you know, create a separate sibling repository that is Spec It Co-pilot. The reason being is is that, you know, the core of Co-pilot I’m sorry, the core of Spec It stays agent agnostic, as we promised. If you’re on the Co-pilot side of things, you know, that particular repository will help you actually layer in, you know, specific plugins, specific specific presets that

13:50 will make your experience in Co-pilot better. One of them being indeed the Co-pilot plugin that allows you to just drive Spec It completely in natural language. You know, so that you don’t have to remember what does this command do. Now, you can just say add this particular preset in natural language and underneath the covers the skill will actually call call Spec It CLI and then just execute it for you. That was my first foray in like, hey, how can I make the Co-pilot experience better? But then, you know,

14:20 we started talking and you know, Nicole and I uh uh had a lot of fun talking about stuff and we were thinking about now, well, how can you actually make things visual? And that’s where I will hand it off to Nicole cuz she did some amazing work there. >> Quick question though, when you approach these sort of design sort of discussions, like how how long does that take? Do you sit in a room and just like write a bunch of stuff on a whiteboard? Do you use Spec It to work on Spec It?

14:51 >> [laughter] >> What what does that look like? >> For the core of Spec It, there are some features I’ve done that, but remember it’s an open source project, so I get a lot of PRs in, you know, and obviously I don’t know necessarily when they’re whether or not they used Spec It Spec It to deliver it because the PR doesn’t necessarily contain that. You know, and if it does contain it, I would say please scrub it because I want to make sure that the repository stays, you know, clean in that regard cuz

15:22 I used specs as uh, you know, uh, mostly transient in that sense. I I like what the in the spec it uh, the repository is described as flow forward. So, you’ve delivered something, you create a new spec for the new work if it’s that large. If it’s not that large, then yeah, I mean, you don’t want to create specs if, you know, the change is hey, I need to if you say for instance doing a web app, hey, I need to create a different color for a toggle button. Button, that’s not really a spec, right? So, um,

15:53 but yeah, uh, mostly over Teams cuz we’re both remote and uh, you know, we would just talk things through and, you know, uh, obviously then try things out, you know, uh, it’s a it’s a lot easier now to just say, you know, what? I’m just going to give it an a day and you know, help help throw the LLM at it uh, and steer and guide it and see where we’re at. >> But, uh, I have definitely been throwing like I feel like when I have had a bunch of projects that go on the like your

16:23 to-be-read list, but like to-be to-be-done projects, uh, yeah, having having Copilot or some other sort of LLM to just kind of cook on that in the background while I work on other things is uh, definitely pretty helpful. Uh, cool. So, I’m going to we’re going to swap over to Nicole then cuz I’m really curious how this um, like so, I full disclaimer, I have used spec it uh, for basically since it launched. I remember when it launched, I

16:55 was I was super excited. We were talking about it. I was telling folks that about it at conferences. Um, but I haven’t used it in the last couple months. So, like I’m really excited about these kind of new I mean, I I need to sit down and read through all of the cool new features and capabilities, but especially since I am working more with Copilot, what does it mean for for what does it mean to work with Spec it and Copilot? So, I’m going to we’re

17:25 going to share your screen, Nicole. >> Yeah, sounds good. You can hear me okay? >> [clears throat] >> Yep, we got you. >> Okay, perfect. Okay. Yeah, so I wanted to first just show the the Spec it Copilot repo that Mom for talked about. So, if you want to learn more, I definitely encourage you to come come check this out and look through the read me. The key points that I want to call out here is is this talks about the five plugins that we’ve added for Copilot. The first one is the one that Mom for talked about quite a bit with this

17:57 called Spec it Copilot and so this is what enables you to use natural language to drive Spec it. I’m going to show that right away today. And then the other four are more are more focused on what’s called a canvas app. And so canvas apps are really cool feature in the GitHub Copilot app that allow you to easily create web apps for really any type of UI that you want. And so we’ve used that capability to provide really nice visual guided experiences over Spec it. So, we’re really excited to to share that

18:27 today. So, with that I’m going to go ahead. >> Oh, sorry. Keep going. >> I’m question from chat. Is this only for Copilot then? So, kind of how how can people work with Spec it then? With all the options. >> The core of Spec it is obviously agnostic. The canvases are not. This canvases require the Copilot app. And so does the the Copilot plugin because obviously it says Copilot.

18:59 >> Makes sense. Just wanted to help answer folks’ questions. >> Yeah, no worries. So, yeah, I switched to my machine. Hopefully you can see this okay. This is a GitHub Copilot app. In In you’re not familiar with it, um, was released, I want to say in the June-July timeframe. Really, it’s geared towards agent first development with Copilot, and so, um, the great thing that you see over here on the left-hand side is you can open up multiple different repositories or projects, which allows you to easily switch, you know, in between them and have multiple different uh Copilot

19:30 sessions running, um, so that you can just, you know, move very fluidly between different projects that you’re working on. And so, you can see in my left-hand pane, um, I have a couple of repos open right now. I have the Spec It repo, um, and I I’ll show that in just a moment with the canvas apps, but the spirit behind that, right, is is that you can use Spec It to actually, um, maintain Spec It, and we do that, um, to some degree. So, those examples will kind of help, uh, show that. Um, and then I just have the scratch one that I’m going to start out with and just kind of give you a feel for what the, um, underlying skills are that we talked about that you can use natural language

20:01 with, uh, to drive Spec It. Um, so, yeah, I already have those plugins installed here, so that’s what’s enabling all the all these capabilities today. Um, but I’m going to go ahead and use the forward slash, um, and just take a look at some of the, uh, skills that are being enabled through this plugin that we talked about. Um, so, there’s quite a few different, uh, you know, skills or commands, really, uh, that you have available to you, but I’m going to just start with the Spec It init one. Um, so, this this allows me to initialize Spec It, right? And so, I could use the forward slash command, uh,

20:32 but again, just to show off the natural language, I’m just going to say, you know, initialize my project. And I typing in a demoing is always hard, by the way, but initialize my project with >> You never type and talk at the same time. >> tough. I know. So, I’ll I’ll go ahead and, uh, and enter that. And we’ll just give it a a second here, but we should see it, um, load that skill here behind the scenes. And so, yeah, you can see it’s loading. And so, what this is doing, right, is it’s setting up my project with Spec It,

21:02 and so, there’s certain files that Spec It, uh, scaffolds to basically get that ready. Um, as it starts to make these changes, we’ll go ahead and get some of those files on disk too just so you get a better understanding of really what’s what’s being scaffolded here. This should just take a few seconds to run and you can see changes are starting to happen. So again, the GitHub co-pilot app is really helpful so you can view exactly what’s happening. So I’ll expand this and then we’ll take a look at the the files on disk here. But before we do that, I just want to mention here so the initialization is

21:33 done and it tells us what to do next. And so it basically alludes to the fact that it actually is scaffolding more skills to help me drive co-pilot. And so that’s why it’s saying you need to reload the skills. So I’m going to go ahead and do that. And then the other thing it’s telling me is that I need to set up a constitution which is typically kind of the first step in in your spectrum and development process so that you have your your governing principles in place. So I’m going to go ahead and follow the instructions there. But again, because I just reloaded my skills, I’ll see some

22:04 new skills that are available to me. So I have one specifically for creating a constitution here. Again, in the spirit of just using natural language, I’m going to get rid of the forward slash and let’s just uh I’m going to copy and paste some stuff cuz I know I know my limitations with typing and demoing. So I’m going to copy this out of a notepad I have but this is basically setting up my constitution and I’m just going to give it a really really simple governing principle and say that for every single change, I want docs and tests added. Okay? And you know, in the

22:34 real world you’d have you know, more principles than this but for the demo today, that’s what we’re going to stick with. So I’ll go ahead and press enter here and we’ll just let it go. You can see here again it’s inferring the underlying skill. And while it’s doing that, I want to take a look at the files now that have been created. So I talked about the scaffolding that happens with SpecFlow Kit. And so when it’s doing that scaffolding, it’s actually creating additional skills for you to help drive that default SpecFlow driven development process workflow. And so you can see here, here’s that

23:05 that constitution skill that we just used. And it also includes skills for each of the phases for the spec driven development process starting with the specification going all the way through implement. So these are these are how those skills get created. And then here I also want to point out in the specify folder, this is where all of the artifacts from the different pipeline phases that you were on, this is where they get created. So for example, right now we’ve got our constitution being created and so I’m going to go ahead and take a look at it.

23:36 And I’ll open up this pane so hopefully you can see it. But this is the constitution that it just created and you can see under the core principles that it it added what I I specified in the chat window, right? So that you have to have documentation for every change and likewise you have to have tests. So yeah, so this is this is what the spec to code pilot plugin allows you to do is is to use natural language to to drive this process. And so if you were going to keep going, right? You would then continue on with the specify phase

24:07 which then would allow you to progress through the rest of the pipeline. But yeah, so that’s that’s what I wanted to show with just the the co-pilot plugin. The next thing is the canvas apps. I don’t know if there’s any questions but if not I’ll just kind of switch gears. Okay. I’ll pause. >> [laughter] >> Yes, so we had a great question from Amy here. So what challenges did you run into like when we’re So this is obviously the the co-pilot specific piece but how did you

24:39 go about trying to kind of separate what fit what parts should be agnostic coding agent agnostic and when you’re building skills and kind of these constitutions, how do you like how does that work with with kind of broader AI landscape? >> Um so the core of Spec it always has been agnostic. So what what it what it means is that when you’re running the Spec of IC to actually initialize your project, you you say I want this particular coding

25:10 agent. So if you have Copilot, you specify Copilot. If you have Cloud or Codex, you would specify that. And it actually would generate appropriate skills for that particular agent set. Now, you can actually use multiple agents in the project. But then you just have to specify which ones you use and you have to switch between them. So the key thing really is is that the directory that Nicole showed where that the dot specify is, you’ll notice that the commands, cuz that’s the Spec it

25:40 terminology for a skill, the commands are there. Once you actually say you want to scaffold them, then they get scaffolded into the directory of your coding agent. So that way it stays completely agnostic because when someone delivers an extension, they deliver it agnostically. When it gets installed, it would create a skill file for that coding agent. >> That’s super cool because then if you have folks on your team that are using different coding agents, you still have the same kind of core instructions,

26:13 uh but they kind of translate to each different coding agent, right? >> Right, right. And and and this particular Copilot plugin actually builds on top of that. So it’s actually the it’s a really nice layering that goes, you know, up. So the Copilot plugin is the is the next layer and then obviously the canvases that Nicole now is going to show is the layer on top of that. >> Okay. Cool. That Great question, Amy. Yes, so let’s uh Canvases. Let’s talk about canvases. I

26:44 love canvases. I’ve been making them like I’m pretty sure I make two to two to five canvases a day. I am >> [laughter] >> super So just to explain to folks who maybe haven’t tried out the the desktop app I have shared links for how to get that in in the comments. I will also throw it up again one more time. But basically canvases are a way to um I mean in in some sense they’re an embedded web browser, but they basically a lots of apps now have this concept. I

27:16 know Slack has canvases different kind of communication tools have them, but it’s just like a nice blank canvas as a way to share you know something more than just text from a chat or something like that. So I tend to find myself using them to make dashboards all the time cuz I’m a data science person. So I love a good dashboard, but yeah, I’m especially curious to see what that kind of looks like for Speckle. So >> Yeah, so we we decided to create three

27:48 canvases for predefined workflows. And so again, like I said this is kind of an example of using Speckle to manage Speckle cuz we we do do that. And these workflows are very applicable to managing Speckle. So that’s why I’m going to show these canvases within the context of the Speckle repo. But yeah, so the first one I just want to start with is the spec-driven development canvas. Again, this is for like that default workflow that that Speckle provides. However, as Mumford said you can fully customize Speckle and have the flexibility really create whatever pipeline you want. And then and so

28:19 that’s what the additional canvases will show here in just a second. But for the spec-driven development one, this is what a UI on top of those skills looks like. And so it really provides, you know, that nice guided experience between different phases so that you don’t necessarily have to know you know which skill to to use next. You know grant granted natural language helps, but but this even helps even more, right? If you’re kind of that visual person. So just to walk you through the app really quickly, um I’ve got some badges here at the top that basically show you

28:50 kind of different states of where um different workflow instances I’ve run are at. And so, I only have one workflow, that’s what’s in the middle here, of a feature that I actually started running through. And so, that’s what these counts reflect here is where I’m at. Um at the beginning, too, you can see that you can actually view your constitution. And so, I already created mine. You can come in and view that artifact if you want to. Um and if you needed, you could also create or update it. Um I am going to go ahead and just kick off specify cuz I know it’ll take a little bit and hopefully it’ll it’ll

29:21 actually finish, but um I’m just going to copy and paste a feature ID that actually comes from our community. I just picked a kind of a random issue that I saw today. Um again, using Speckit on Speckit. Uh and I’ll go ahead and just uh hit this run specify uh button here. And you can see in the chat window um that what’s happening here. And uh let me just go ahead. I think I might have run this before, so hopefully it doesn’t cause a problem, but um but what we should see here, right, is is it’s going to actually then

29:51 invoke that specify skill um that exists underneath to then uh basically execute this phase with the the additional prompt that I provided there with the the feature description. And so, then once I do that, um it’ll run the specify phase and then it’ll lead me basically through the next steps of the pipeline as needed. And so, you can see here with this existing one, um I ran specify, um I then went to the planning phase tasks, and I I think I stopped at checklist. But, if we just open one of these just to give you a feel, um I can come in

30:22 here for one of these phases that already ran and I can view the artifact. So, this is my my implementation plan that got created. Just scrolling really quickly. Um and then also uh sorry, going back to the plan. Um you can also come in here and and rerun a phase. So, if there’s additional context that you want to provide, you can certainly do that. Um but yeah, you would just kind of work your way through the phases as you normally would, but um, all using a UI to do that with. Uh, so that that’s what this looks like. Um, the other two canvases that I wanted to provide or an overview of really

30:53 quickly is uh, there’s this idea assess canvas. And so um, Manfredi talked about there’s these two extensions that we just added recently and this is one of those, right? So um, it gives you an entirely different pipeline than the the uh, built-in spec through implementation process, right? This is uh, taking an idea and then researching it, defining, shaping, and then finally coming up with a verdict of a a go or no-go decision. Um, and so in my case I’ve created a couple um, of workflow instances here

31:23 where I’ve started to run through um, those steps. Uh, so for example, this one you can see I’ve actually ran through all those different phases um, and ultimately I I did get a decision. Um, and so if we look at that artifact, um, it actually says that I need uh, clarification. So it didn’t even give me a go or no-go decision yet and that’s because if I scroll further down, you’ll see that it actually needs more context. Um, and so this is a really another cool feature of Co-pilot is is there’s a tool underneath that allows you to ask for additional information that Co-pilot can do that with and so that’s what this is leveraging. Um, so there’s there’s

31:55 questions I need to come in here and answer and then once I submit those answers, I can rerun the stage. Um, So yeah, that’s that’s kind of how the uh, the assessment pipeline works. Um, the third one, oh yep, did you want to say something? >> Yeah, I I want to make sure that people understand that this is the visualization. The actual uh, commands are run by Spec it. So if you do not have the Co-pilot app, you can get the same artifacts generated by using, you know, the

32:25 Spec it, you know, the way you always have. >> Yep, that’s a good point. Um, and then the third canvas app I wanted to show is the bug fix one. Again, uh, this is another one of the extensions that Momford talked about um that basically triages a bug um starting with the assess phase and then um assuming it’s a valid bug cuz that’s um basically kind of the verdict that it arrives at. Um then you can go on to fix and actually test it within this repo. Um so again, I I picked um an actual bug from from our

32:55 repository and started to run through the assessment pipeline. So that’s what we see here in the middle. Uh so if we look at the assess phase again, you can view this artifact and um at the top here, you can see here’s that verdict, right? That it’s it is in fact a valid bug in the repo. Um so we could then go on and and and continue with the fix and the the test phase. Um so yeah, so hopefully this gives you an idea of like how you can actually create a more guided experience um using canvases. Uh there is a fourth canvas

33:26 that I want to show you, but it’s different from the the first three. Um and this is what we call our wizard canvas. Um and the whole spirit behind this is that this is really meant to be more general purpose and for exploration. Um so it allows you to basically peruse all the different presets that we have available across the different catalogs. So we have a built-in catalog, we have a Copilot-specific one, and then we also have community-contributed presets. And so you can come and scroll through all of these and choose to install them. Um

33:57 likewise, you can do the same thing for extensions. And so you’ll see actually I’ve actually installed a few, so um or actually one one extension here, but I installed the idea assessment pipeline, which is the same extension I just showed in the canvas, and also the same extension that Momford talked about. So we’ll we’ll show how this shows up in the the wizard. Uh but the idea here, right, is is that you can come in and install um whatever customizations that you want to tailor the pipeline to your scenario and your process, right?

34:28 Um so with that said, I’m going to jump to this phases tab. And this phases tab is what allows you to really build up your pipeline and then actually execute it and try it out. Um so if you remember like I said I did the the intake assessment extension and so um this this extension has the commands that are shown down here, right? It starts with intake, it then goes to research, uh then define I think it’s shape and then define and then decide. Um and so you can actually modify this this pipeline up here. You

34:59 can remove um different commands. Uh you can come and then add additional ones. And so what I’ve done here is I have added the commands that that I’m interested in specifically for the the intake pipeline. Um and at the end of this I actually then added the specify uh phase because that’s really the next step once you uh reach your verdict for your idea. Um so again this this is just uh meant to be a tool for you to build up your pipeline, visually see it, and then you can come in here and actually execute the pipeline and try out the steps. Um so I’ve I’ve already executed

35:31 it but like if you look at um I wanted to show you specify cuz I did run this. Uh if we view the artifact here, uh this is the actual specs then that were created based on the original idea that I put in at the beginning of the pipeline. Uh the other thing I’ll just mention too is is I I have a fun preset installed. Manfred actually created this preset but it’s called pirate speak. Um and so this really shows how presets can, you know, modify your pipeline to however you want. And so this this particular preset called pirate speak uh basically outputs all of your artifacts as though it’s a pirate,

36:02 right? Uh uh writing it. And so like if you look at some of the the verbiage being used here it’s it’s all in the spirit of a pirate. >> [laughter] >> So again it’s just really showing off, you know, the customizations that you can do um with with Spec Kit. Um so yeah so any questions about that because otherwise I have one more final thing I wanted to show if we have enough time. >> [clears throat] >> Um >> I just want to clarify something for folks as well. When when Nicole talked about the pipeline here,

36:33 you know, you can put things in there and take things out, obviously. But the key thing is is that those presets that were on the previous in the setup side, that is where you can layer in your own functionality because that’s where she actually selected that pirate speak one. So, you know, if you have standardized ways of working, you know, that’s where you would do that and you create your own preset and then you would be able to say, “Hey, I want to now construct my pipeline. Oh, I need a next particular thing that the core of Spec It doesn’t

37:04 do. Okay, I can create an extension and then layer that into your pipeline.” >> Yep. >> [snorts] >> Um I’m going to there is a number of folks asking in chat. Yes, this is So, Spec It is an open source project and a lot of the stuff that I think pretty much everything we’re showing here in the canvases are tied to basically Spec It is underlying running these processes and commands, right? >> It is. >> Yes.

37:34 >> And and and the Co-Pilot specific project is also open source. So, you can see how this was built out. >> Yeah, so even if you don’t use Co-Pilot, that’s totally okay. You can still leverage everything cool about Co-Pilot. Maybe there are cool ways to make canvases or something similar in your favorite AI IDE. Um but that’s what’s awesome about I love good community kind of like supported like templates or or ways that basically folks can share how they’re using a

38:05 tool. Um do you have a community page or like a a place where people can maybe find maybe I can find it uh find all these cool um configurations? >> Um it’s actually on the website of the for the Spec it core you know it’s the documentation website and then there will be a community tab there that folks can click on and they were they’re categorized by area so you would have one for extensions, one for presets, one for workflows and so

38:36 forth and so on. And we also have a friends page. That’s where folks can list their project that uh does use Spec it cuz that’s kind of the requirements but it’s friend to it. So folks that actually created terminal UI for instance around Spec it you know so that’s a friend of ours then. >> Got it. Okay, cool. Yeah, I I’ll share some links in chat for folks and I’ve got the link to the main repo up right now. But yeah, I think that addresses most of the current open comments so.

39:08 >> [laughter] >> Thanks to Cole. Yeah, it was just one last thing then that I wanted to show and this is this is what I would say is probably extra experimental at this point. All the functionality that I’ve showed you so far is actually checked in to the GitHub Copilot repo so you can access that functionality today today. What I’m about to show now is not yet. We’re still working on it and if all goes well you know we’re hoping to eventually you know get it into the the Copilot repo. But for now I just at least wanted to show you kind of where we’re headed with things.

39:38 So again I have the wizard open and it looks pretty much identical to what I showed you to before. However, the key differences is there’s this new button on the phases page called generate canvas. And so our idea is is is that you can build up whatever pipeline you want right according to your process with the presets and extensions that you need installed. And then once you have that pipeline built up you can come in here and click generate canvas and it will actually generate that canvas app for you so that you don’t have to manually create it like we did.

40:10 And so it’s it’s a you know great example of where you can take all this rich metadata that we’re building up in was the wizard and then hand it off to co-pilot to build this app. And so to do this um the generate canvas button um allows you to launch what’s called the canvas designer and I’ll just go ahead and and launch that. Um this will take a couple of seconds um to do because it’s actually going to create a nested session um that we’ll see here in the left-hand pane underneath the wizard. So that takes a little bit of time to do that. Um and then it hands off all that metadata that I talked

40:41 about from the wizard so that it can then generate the canvas um or get it ready for generation. And so in the interest of time I did go ahead and uh basically pre-launch this canvas designer. Um we’ll see a second instance probably spin up here in just a second um as this is working. Yeah, you can see that here there’s the nested one that’s creating. Um but yeah, so this is what the canvas designer would look like and really what it’s meant to do is is that there’s a couple of required fields you do need to set. So like the canvas ID which is the name of the file, right for your canvas app. Uh you also do need

41:13 to provide it a title. Uh but then I just provided a few basic fields um that you can use to kind of customize the generated canvas app just a little bit more. Um so for example, you can provide a description. You can also provide like the header. Um there’s also, you know, we showed badges before. There’s uh different badge rules that you can specify to have those appear. Um and then for the appearance, there’s some um general settings here but one of the more interesting ones, right, is being able to add your own logo for example. And then finally when you’ve just kind of finished setting up those basic

41:44 settings, you can come in here and choose to generate a canvas app. And again, it’s it’s based on that metadata that you’ve already configured in the wizard and passes it to this, applies just these basic customizations to it, and then outputs a canvas app for you. And so to show you an example of one that I actually generated, um again, this is following the bug fix assessment pipeline that we’ve we’ve already been talking about throughout a lot today. Um but this is what the generated app looks like and I I wasn’t really happy with the badges cuz I was testing it out. But uh but you can see these are the

42:14 different badges I added to the designer. You can see my logo. You can see my custom description that I put in the workflow the workflow header name. But beyond that, you know, the behavior is just the same. It’s driving Speckit underneath to call those skills that we talked about to do the bug triage where you go through assess, bug fix, and test. And so anyway, so we thought this was really cool that you could, you know, use Co-pilot to basically generate a canvas app from all that metadata from the wizard. And just one more quick thing is is we

42:46 also decided to make this even more customizable. Like if you really wanted to get advanced and make your canvases even more customized to your needs because, you know, just assume those basic settings I exposed in designer aren’t enough for you. You can actually use Speckit preset presets and extensions to further customize the actual app that gets generated. And so just to show you an example of that is I created a preset that takes this horizontal display of this control and instead displays it vertically. And also

43:17 use autopilot mode. And so that’s what this app looks like. So it’s the exact same metadata. Um Everything’s the same except I applied this this preset that changes the control layout. And so here you can see it’s it’s the same steps but it’s vertical and then this runs in autopilot mode. So you can just put in your you know, your bug and let it run. So anyway, it’s just just some cool things that we’ve been trying that we wanted to share. So >> Oh yeah, no I am super keen on trying

43:47 this out because I’m dashboard very visual person and I think it’s really awesome that we have a way to Don’t get me wrong, the CLI is great. That’s how I was using Speckit before. Um That is totally awesome. And I’m actually kind of curious to check out that community project that gives it You said it’s kind of like a GUI for the CLI. So, I do also like those. But, yeah, no, I think especially with how much you can customize this, like you can really make

44:18 it feel like it is it is your project. These are your workflows. Like, um and so when you have new people come in and join you, like it I don’t know. It feels it feels polished. It’s like when you take and actually customize a website template, and so it looks like your project’s website and not just like every every other, you know, one of three basic templates for a a generator. So, that’s super cool. I’m excited about that. Um I’m going to take a look if you have

44:49 any questions for folks, uh please make sure to put them in chat. Uh I’ve been watching. Um but yeah. Um >> [clears throat] >> Yeah, I think kind of one of the main main threads from chat is folks trying to understand which which pieces here are Copilot specific and which are not. So, uh I think we’ve >> It’s relatively easy to to know if it’s in the in the spec of Copilot repo, it’s Copilot specific. If it’s not in the

45:20 core, cuz the core we we didn’t bleed it over. So, the core stayed the core, and the Copilot side is literally in that repository. Now, I do want to call out for for folks that you know, obviously are interested in any of those new flows. Yes, you can just do the same thing that Nicole was showing in the UI completely on the command line. The only difference is is obviously you don’t get that nice UI around it. Uh you know, and yes, the designer and the wizards are not available to you because that’s

45:52 really specific to Copilot use because it requires the canvas and the ability tend to generate the canvas and then show it, you know, while using Copilot. Um I I do have to say it was really interesting when we had conversations cuz you know, asking about how we designed this, you know, is something that um especially the last part that isn’t checked in yet, that was really the the biggest one is like how do we how how do we think about, you know, you can generate the app, but then it’s static.

46:23 So, how do how how can we layer in the same kind of concepts so that folks can uh generate their app in their in their style, you know, with the with their requirements that need to be there. And you know, I I asked Nicole, is it possible to use presets? Cuz that it really literally was a question from me to her. And and we were like, let’s go find out. >> Yeah. So, yeah, there’s the actual preset that I created, right? Um so that when you launch the designer, you can apply that preset, and then the designer installs that basically when it’s generating the

46:54 canvas. Nice. >> oh okay, actually probably when I explain that, the designer then runs obviously a Spec-it workflow to do that because otherwise >> Yes. >> it wouldn’t work. >> Yep, all this functionality’s packaged to in its own extension actually, and it’s you can see that here. That’s that it has to be installed cuz that’s where all of this is at, and then you can override it, like I said, the behavior with presets and and more extensions if you want. Um so someone has raised an interesting I I

47:24 don’t know if this will render nicely cuz it’s a longer question, but I think the kind of general gist of it is um token usage and working with Spec-it. Are there, you know, has there been kind of any consideration about how to be more efficient about that? Like, you know, I I think something that’s neat about using some of these workflows is >> [clears throat] >> I don’t have to go like I’ve made a lot of canvases.

47:55 It is quite a few tokens for me to go back and forth to change colors, change alignments of things. And being able to have presets that can I think skip a fair number of those tokens. But yeah, kind of how did you go about thinking about that with the design of Spec kit? >> Well, the design of Spec kit obviously started with a as the decor SDD workflow. It is >> What is What is SDD for folks? >> Uh >> Spec driven development.

48:26 >> [laughter] >> Spec driven development. >> Yes. >> Yes. So the core workflow obviously delivers a lot of files because you’re trying to do define API contracts. You’re doing research. Then obviously you create your spec based on all of that and you have to plan it the create implementation plan then you’re creating the task list that actually needs to be done. Now, yes, one of the things once we introduced preset it’s like okay, you know, for a small feature

48:57 that seems rather large. So there is a preset it’s called lean preset that only creates three files. The spec, the plan, and the task you know, yes, that one consumes a whole lot less tokens but also is less specified because of the fact that you don’t have your API contracts. There you don’t see the research, you know, obviously the LM did something but it didn’t actually formalize it in a file. Okay, so it all depends on how much you need and how much your organization requires of you to specify certain

49:29 things out. Now, if you’re talking about a small small thing, a small feature, then yeah, use the lean preset, right? Because it’s it it it is way leaner and it’s quite a quite a bit leaner actually. Now, does it come with some certain things that you need to be aware of? Yes. It doesn’t do certain things that the original core does and that’s by design. But it’s also an illustration of if you want to influence how the process works, you know, you can create your own preset

49:59 and say, “Hey, you know what? I don’t want API contracts, but I do want all the other stuff that happens.” Okay? Override and decor with a preset. And you can do that. Now, um what was the secondary question? Sorry, cuz I kind of like >> Um I I think it was kind of just mostly you know, if if you are token conscious while you’re using SpecKit, what I I think you answered pretty well what the gist of what we’re asking about. Um

50:32 there was an other interesting question uh which was um so, I think the internals of SpecKit are written in Python. I’m a Pythonista. Uh I have no problem with that, but we’re at kind of what was part of the design decision around uh those sorts of just uh architecture choices. >> Um I can really not answer that question cuz I um I left when Dan left our company. So, why he picked Python, probably because he was comfortable [laughter] with it.

51:03 You know, and obviously you’re not going to rip the you know, the the foundation from under people. So, yeah, Python has to stay. >> [snorts] >> I mean, I think uh something I always tell people since I’ve used Python kind of as my primary language for about 15 years, it’s the second best language at everything. Probably for every task, like it’s a Swiss Army knife. So, like you know, probably for any specific task, there might be a better more [clears throat] optimal choice for speed or something like that.

51:36 But whenever I need a library for something, there already is one in Python. >> [laughter] [clears throat] >> From the standpoint of building things, I feel like it’s a and not to brag, but it has a great community, too. So, [laughter] any any project that kind of gets to lean into those communities, I think will benefit a lot. We’re very friendly, so. >> And and the the point out, I mean, obviously, when you run the actual commands, Python is

52:07 not in play. It then it becomes the agent that actually just reasons about the prompt it receives and then, you know, whatever tooling it has, you know, it could be a Python tool that it calls. It could be, you know, a bash script. It could be a PowerShell script it calls. You know, then it depends on what the command actually specifies. So. >> Yeah. I They’re all fully decoupled, basically. You use this to generate your specs for the spec-driven development. And then, you use those, you know. I I

52:38 have definitely prior to knowing about SpecKit sat down and written many specification documents. And it unless you have like company templates or something like that, it can it’s like a lot of full-time work. So, [laughter] it is a great great thing to be able to kind of start with the scaffold and then tweak as you need to. So, that >> [clears throat] >> that was why I was originally so excited about SpecKit cuz I’m like, >> [snorts] >> we can have something to talk about

53:08 that, you know, is already, you know, I can turn on and off features if I don’t need the API contracts, stuff like that. I can just turn those off. Um so, yeah. Definitely highly suggest, especially if you are starting new projects. Um Try out SpecKit. See, you know, kind of see what it helps helps you think through. Like, I think it also kind of can sometimes help prompt folks to things that they >> [clears throat]

53:40 >> It’s morning. >> [laughter] >> Things that folks might not be thinking through Uh kind of completely when they’re designing a project. I know it has definitely helped me think about, oh, there’s actually an interface there and I definitely need to kind of flush out more what what are the rules around this. So. >> Which which actually ties back into that idea assessment extension because, you know, before you implement something, you know, that one actually is also asked the question, is it even something you should be doing? But the nice thing

54:10 >> [laughter] >> the entire process of that one is is that it actually asks you clarifying questions. Now, you can completely ignore those, obviously, but it is interesting to see those clarifying questions because it will tell you like, did you consider this? Did you consider that? And you’re like, oh, the oops, I missed something. And that’s, you know, that’s the thing when we’re talking about an idea, it’s like, did you really think about everything that that is involved in in in what you’re asking, you know, to be changed here? And that’s

54:40 something that, you know, if you’re working on multiple projects, it’s not always an easy thing. And, you know, uh yeah, if you go if you if if you onboard and you just have the ability to say, you know what, this is an an issue that was filed. Let’s go see, is it something that actually fits the code base or is it something that I should push back on? >> Mhm. It’s I feel like, and you can correct me if I’m wrong, it’s like having a pocket PM. >> [laughter] >> It helps you like, I’ve done some PM

55:12 work, that’s not my forte, but I do definitely understand the value of having a great PM when you’re working on a project and they often are the kind of person who asks those questions like, do we really need to be doing this now? Um, you know, have we thought about this? So, yeah, I think it’s a really cool structure to leverage LLMs with to help you work on developing your projects. >> It’s funny that you call Pocket the LLM sorry Pocket PM, my bad.

55:43 >> [laughter] >> Yes, I mean I honestly I would actually if you are a PM use it. You know, you can make it your own just like the the SDD part. You can make it your own with your own presets for that one as well. So, if you feel like okay, this is this process doesn’t entirely adhere to how I would want it done, you can actually you know, tweak it. Now, obviously themselves that are that you know, are currently in the assessment pipeline are you know, there. Do you need to execute

56:15 all of them? No, but that obviously that depends on how you actually would look at your own process. Which is you know, yes, we started as Spec It. You know, it really has now become a process kit. >> [laughter] >> A random I feel like I might have to contribute this as a community thing, but I used to work on medical devices and the software processes and requirement like you there it was like a binder of all the things you had to check that every time somebody made a change to code like all of those rules

56:47 were followed. That feels like something that would be perfect for this. Like I’ll just hand hand Spec It this manual and be like help me make a preset that will when we make changes to the code actually make sure we’re adhering to all of these steps. So, >> It’s funny that you mentioned it. Um way back earlier in my career, I actually did work with a more regulated industry. It happened to be the VA. Um they actually have what they call a NEEM

57:19 model that is for information interchange. Um so yes, I I I I just created the presets for it because I was like, you know what? Let’s see if I can actually do that cuz you know, it’s and then know then actually see what it does. It’s very interesting to see. Now, I’ll try to plug that everyone should use it or anything, but you know, if you are interested in that, you know, go check out my personal repositories cuz that’s the thing that I always want to call out as well. I’m just as a community member as everyone else when it comes to

57:50 extensions and presets. You know, we’re trying things out and obviously you know, because of the community and which we really love, we can see what sticks and what doesn’t stick cuz you know, the more extensions, presets, bundles, workflows, and workflow steps that that that and canvases that come around, you know, the more you know, we can all help each other out. >> Yeah. I that’s one of the my favorite things about the open-source community is like

58:21 yes, we can talk about all the cool things that this does, but you can also make it you can add new things yourself. You can see how they work. You know, it’s not just some sort of magical black box. So, um when something usually in my case I break something, it gives me an opportunity to go before bugging a bunch of people go see um what what did I do wrong? >> [laughter] >> So, yeah, definitely if you are interested in this project at all, please go check out the repo for either

58:53 we’ve got Spec Kit. This is the main repo for the kind of coding agent agnostic pieces. And basically the actual computational back end for the Co-pilot integration for Spec Kit, which basically gives you lots of nice options in the Co-pilot app and other Co-pilot tooling to leverage the agnostic. So, not there’s not really special features or additional things. It’s just ways to make core Spec Kit work well in

59:24 Co-pilot. Hopefully that’s an accurate characterization. Um Yeah, so I will close out um >> [clears throat] >> with a fun question. Um What’s your favorite ice cream flavor? >> Pistachio. >> Pistachio. Okay. All right, that’s a solid choice. >> Mine would be cookie dough. >> Cookie dough. Yeah. >> And yours? >> that’s Mine uh well,

59:58 nominally like a sea salt like or sea salt caramel sort of thing. However, not sponsored plug >> [laughter] >> for Salt & Straw, which is like a a chain in the US. Uh But, they have a Halloween flavor right now called graveyard cookie crumble that is truly the best ice cream flavor I have ever had. I am doing everything I can to not go there literally every day this month cuz it rotates out at the end of the month. >> [laughter] >> I am strongly considering buying so many

1:00:28 pints and stocking my freezer with them. >> [laughter] >> But, it’s like really really dark chocolate with cookies in it cuz I I do love you got to have like the kind of cookie crunch in there. Uh and like dark like bright red white chocolate swirls cuz it’s a a graveyard. Anyway. >> [laughter] >> Uh well, thank you so much for joining me Manfred and Nicole. >> Yeah, thank you. >> You. >> Um and thanks everybody for joining us uh online and uh watching the recording later. You can always catch up on more

1:01:00 open source Fridays. We’re here pretty much every Friday. Um I think just literally opensourcefridays.com. >> [laughter] >> Um and you can check out our whole catalog and I look forward to trying out all kinds of new SpecKit stuff this week and also uh what comes next for the project. And hopefully, some of the folks here who have uh caught up in chat might be contributors to the next next iterations. That would be super cool. Just a thought. >> Mhm. >> [laughter] >> Anyway, have a good one everybody and uh we’ll catch you next time.

1:01:31 >> Thanks. >> Bye.