The SDLC You Build Yourself — Andy Wong, Andrei Bocan & Shane Wolf, Atlassian

AI Engineer · 57 min · 660 sentences · from YouTube's caption track

Each timecode opens YouTube at the start of that sentence. Line anchors (#s42) are the cue ids in the WebVTT, and every line carries its start and end seconds. All transcripts has every talk, and the whole corpus as one file.

00:13Welcome everybody.
00:14Welcome to the AI Engineers World Fair.
00:17Uh thank you everyone for joining for our workshop today.
00:20Quick introduction for myself.
00:21My name is Andy Wong.
00:22I've been working at Alaskan for four years now and I've been working in the AI space for two and a half.
00:27Joining me today are two of my wonderful teammates.
00:29We have Shane Wolf uh manager down in Devi space as well as Andre Bohan down here as well.
00:36So today we're just going to go over a lot of the things how how the agentic systems are kind of changing how we are uh handling the software development life cycle and how orchestration does change everything.
00:59Can you repeat?
01:00Yeah, he's Shane just is gonna walk around give hand out numbers.
01:04Uh it's going to be a little bit weird at the beginning, but once we finish up the uh introduction talk, we'll be it'll make a lot of more make a lot more sense after uh after the kind of like the introduction talk kind of finalizes up.
01:15Um but we just want to quickly get started on the first intro slides.
01:19Um I do want to run through these probably quickly because since it is the AI engineers welfare we want to build.
01:25So I just want to give a quick rundown of everything we've been kind of building in Atlassian for the last I would say one and a half years or so.
01:32Uh just to give you like a highle overview of all the tools we will be offering you within Atlassian itself.
01:38So I think as as you guys might know AI agents have kind of changed the software development life cycle like drastically.
01:45Um, as time kind of goes on, you can kind of see that as we have more AI crewing tools, kind of productivity gains from those tools are kind of tapering
01:53um, as we have more and more tools coming out.
01:56And thus us at at lasting we believe that what our biggest opportunities are tapping on the on the everything left and right of code.
02:03So everything involving planning, orchestration, uh, the code review process as well as the operations part and such.
02:11We've been developing a lot of the things that um we can say like are not as apparent during the software development life cycle.
02:19Uh we have things such as the teamwork graph coming out as well as a bunch of different agents that we've been trying to do left or right of code um such as AI planner, AI autodev,
02:28reviewer, s sur um and and central of all that is something we call the AI context machine.
02:34Just to quickly run through all those tools, what we we what we will predominantly be testing out today is something called the Jura coding agent.
02:42Uh that's our that's the Atlassian coding agent natively uh that has deep organizational knowledge within the Atlastian ecosystem.
02:48So it's good it is a deep enterprise context agent that has uh a deep understanding of everything within the elastian ecosystem whether or not you have juror tickets confluence documents
02:58um it will read through email slack slack uh slack messages as well um to be able to gather all the context it needs to kind of plan out and implement a specific jur ticket that you might have
03:11uh within the jur coding agent itself uh you are able to uh interact with your pull request within the Jura ecosystem itself.
03:20So you everything is everything you'll be able to maintain and edit within the cloud.
03:28Addition to that you'll be a be able to assign your work to specific agents directly.
03:33Uh so if you don't want to use your juring agent that's fine.
03:36What we do provide for you as well is you are able to assign to uh agents such as claude cursor codeex whatever agent you personally prefer.
03:44uh you are able to assign the jur agent jur ticket to um during during that process as well.
03:52And finally we also have SDLC automations as you some of you might know we do have an automation process within Jura itself.
04:00Uh whenever there is a condition met you are able to generate like you are able to generate code uh using the jur coding agent or whatever agent you you personally decide to use.
04:18As we go further right through code, we do have the AI reviewer.
04:21Uh something similar to kind of similar to a lot of tools you've seen in the in in the industry today.
04:26Uh we are trying to implement uh a review of your agent to kind of go over the pull request to catch bugs, enforce standards and verify requirements every time you do make a pull request itself.
04:38So review does go beyond the code.
04:40Again, we are trying to build something that uh retrieves all the context within Jura Confluence, whatever SCM you kind of use as well as Slack to to develop like kind of like a holistic view of what your code PR needs to implement
04:55um and provide a review on that end.
04:59On top of that, we do have centralized me uh manage standards.
05:03Um every team, every organization kind of has their own standards that they want to implement in each uh each repository.
05:10Um it get it gets kind of tedious and what we're trying to do with standards is trying to trying to make it a lot easier for managers, software engineers to implement those standards in all in all the poll requests that come within a specific rep repository.
05:26[snorts]
05:26And as and additionally, this will work on all SCM platforms whether or not it's GitHub, GitLab, Bitbucket Cloud.
05:32um you will be able to apply standards to your your poll requests.
05:39One thing we are trying to do at Alaskan with our AI reviewer is we're trying to reduce the amount of like amount of noise that comes from the code review process as a whole.
05:47[snorts]
05:47Uh what we do find is that some of the code review processes uh do provide do are maybe a little bit overly nitpicky or like you know provide a little bit too many comments that are not of a high value.
05:59So we what we've been doing at Devi is we uh train a few models ourselves filter those and what we have done is use those models to filter out
06:06uh let's say low value comments or like noise that we have deemed uh and try to and filter those like unneeded comments during the poll request comments itself.
06:17Um as a whole we have seen that there has been a higher quality uh comment resolution as you can see the bottom here.
06:23uh using these models we have been able to get around 40% uh human code resolution rate which is rather close to human baseline of 45 and as always we can always measure and you know how see how impactful all these tools are.
06:37Um at lasting we have been building a lot of like dashboards for our various tools to see how many reviews or how many code like uh how many agents you've run
06:46uh to kind of give you an idea of what the value of all of our tools are uh within the devi space.
06:55Going on to a few like the final few few products we've been implementing recently.
06:58Uh something called AI autodeev.
07:01Um it's pretty self-explanatory what it can do.
07:04Uh it goes through your jur jura board uh identifies which uh which which tickets are I would say able to be able to be autonomously ran.
07:15Um applies all the standards initiatives um and then creates a poll request ready for review.
07:21Um, all of that is triggerable within the Jira ecosystem itself.
07:25Um, you can configure all a bunch of different needs such as maximum number of credits you want to use per month.
07:31Uh, maximum number of like tickets you want per per user to run autodev on.
07:37Um, and it's able you can also specify which repos you want to have this running on or if you want to run it on all repos within your your organization itself.
07:45And as always, we have a dashboard to show all the uh all the autodev runs, how it's kind of working, um how how many autodev pull requests are need to review and how many story review requirements.
07:56As always, we're trying to give you like as much uh metadata as possible for the engineers, for the teams to really understand like how how autodev is overall performing for you.
08:06[snorts]
08:08One other thing we do have autodev integrated on is slack.
08:11So once you do have uh autodev integrated within the repository itself um it will start working on a specific work items based on whether or not it's available it's able to use
08:22um or if there's any items that you need to uh any input to finish them more ROI insights.
08:29Um as always we'll we will try to measure as much as possible uh to see how what is the overall ROI you are seeing from autodev as a whole.
08:38And finally AI sorry uh we've all been through incidents we all understand how difficult is to diagnose and kind of go through it especially at like 2 3:00 a.m. at night.
08:48Um what AISR aims to do is try to grab all that context again um within previous incidents jura confluence um and try to root cause why an incident is happening and provide suggestions of how to remedi uh remediate the specific
09:05uh specific incidents once they do come out again it is integrated with Slack uh so you should be able to get a quick summary of what's going on uh understanding
09:16whether or not it's impacting multiple external customers or if it's only impacting a small section of customers and it will provide you uh detailed reason detailed uh detailed steps of how to possibly mitigate the issue that you could you are seeing
09:32as you can see here uh it will provide like like I said we'll provide root cause analysis um it gives you a bunch of different theories of how to possibly remediate the issue as you as you see as you see it coming
09:44I've been saying a lot about context and how the context will be helpful for agent itself.
09:48What we've been trying to build here is the AI context agent.
09:51Um, many of us have been working in large monor repos and all that context is scattered within the whether or not it be different different repositories itself or different pieces of documentation
10:02all throughout whatever uh all whatever workspace you kind of work in.
10:06What the AI context engine is aiming to do is try to grab all of that and create, you know, like a standardized context for you for all of your agents to kind of run in kind of like a more efficient way.
10:18Um, I know I've said context a lot.
10:20That's the biggest thing we're trying to do in the last year.
10:22We're trying to grab all the context as possible and try to create uh a really powerful engine for us to power our agents in the future.
10:29Um so as you can see here AI context engine all of our tools Jira confluence um all of our previous instance everything uh will be will be run using semantic semantic code search and we will be able to provide all that context to whatever agent you kind of use in the future.
10:46What we're trying to do is also make a lot of our tools like uh plug-in accessible.
10:50uh you so hopefully what we can do is uh allow the allow the engineer to simply just add the tool to their browser to their to their IDE or to their CLI
11:00and allow them to use it out of box as simply as possible.
11:05As I said, I wanted to keep this really short because again we're at the AI engineers world's fair.
11:09We want to build.
11:10So what we got to do here, what I do provide for you here is a small workshop for you to play around with Jurro coding agent.
11:15Um, you might see Shane walking around with a specific number uh specific number uh and handing handing that over to you.
11:24Um, that is going to be your account number.
11:27So, let me just quickly go over here.
11:33How this will work is that that number will be tied to your login and the login is the following.
12:01Apologies, some Wi-Fi issues.
12:09So, it's not just me.
12:11All right, got it.
12:21going back to this just so just so everyone does have the QR code available to them.
12:37Yes.
13:00Yeah.
13:01Yeah.
13:01Yeah.
13:01Let's see.
13:04Yep.
13:05Let's see.
13:07I will stop sharing my screen.
13:13Yeah.
13:13Yeah.
13:13Yeah.
13:14I'll do this.
13:21Oh god.
13:24Yes.
13:28Not quite.
13:31Is the internet really slow?
13:33Let's see.
13:36Oh, the Wi-Fi is quite slow.
13:38Did you connect to this?
13:47Oh, there you go.
14:09network changed.
14:15Do you still have cloud fire on?
14:16Try try turning it off.
14:20It's off.
14:22It is off.
14:23Let's see.
14:33How's it going?
14:36Yeah, we're trying to do right now.
14:37Yes.
14:52Yeah.
14:52Yeah, that's the Wi-Fi.
15:14Is that not working?
15:15Yeah, it's not working.
15:21Okay, gotcha.
15:24Yeah, I'm going to try mine.
15:36Got it.
15:38Yeah.
15:39have any questions that we can
15:41[snorts]
15:42talk through while we're figuring out Wi-Fi.
15:49Oh, fair, but also like um
15:54or stand right in front.
15:57Stand right here.
15:58Okay.
15:59So, while we're figuring out Wi-Fi, as I was saying, anybody have any questions that we can talk through in terms of I don't know where we're thinking of going with like the the product line, what stuff is available, what stuff people are keen on, workflows, so on.
16:13Uh, happy to go through any of that.
16:16Sorry, go for it.
16:19Tell us a bit more about the context engine.
16:22Okay, so essentially with a context engine, so um we have a number of I guess graph databases inside of Atlassian that connect to all of the data that you end up plugging into an ATL, uh into your Atlassian products as an Atlassian customer.
16:38So like Confluence, um Jira, Bitbucket, all that stuff is sort of connected together.
16:46We ma in make those connections either through inferences as you use the product or you can like have the connections be explicit.
16:55So like you're working on a Jurro ticket.
16:56The Jurro ticket is based off of like some spec document that is in Confluence.
17:00We also link those to like third party um um what you call a third party product.
17:06So like we will integrate with things like notion.
17:08We'll integrate with things like Slack.
17:10All of that stuff kind of wraps into this whole tapestry of context.
17:14And at the end of the day, the thing that we're uh we're building now is how do I word this?
17:20Uh so it's it's a gentic search on top of all of all of that context.
17:27And as you're this happens a lot like internally in Atlassian like we end up building on one side of the uh on one side of the ecosystem in our platform and that needs to pull in context from either
17:40some other team like some platform documentation.
17:43You have uh some other initiative that's happening on the other side of the organization that somehow happens to impact this bit of what you're building on.
17:51and through the context engine all of that stuff gets pulled into um your current agentic coding session like that is the stuff that Shane who's currently working on Wi-Fi so connectivity
18:05basically is his through line um that that's the stuff that's that he's his team's working on the things that I think we've had a lot of success with um
18:15has been uh so we have a new code search product that we're um we're launching um that's made it so that I think historically we've uh we've seen a lot of teams
18:30try to do like opt for monor repos in order to like have all their their agents like have the context available at all times that has some shortcomings and like means building out like a bunch of tooling
18:43uh stuff gets slow I don't know internally it's been the proc opting for a monor repo ends up meaning that you're opting for the debt of having a monor repo building the tooling around it.
18:54You spend a lot of time building tooling to build the thing.
18:57Um, and I think with a with a code context engine, we've been able to like just have product repos like monor repos because you can pull in that engine at hawk.
19:08It's super fast like you get code level um code level access and that means that any logic that you have that would otherwise be potentially stale, right?
19:18So like a conflence document is just a snapshot of a thing that happens once but like the truth is in the code.
19:23Being able to like access that at high throughput and like having uh good search semantics around it is been really useful.
19:29Sorry that was like very soap boxy.
19:31Does that did that answer your question?
19:33Okay cool.
19:38Anything else?
19:41Please.
19:47For example, if you have a GR
20:09uh sorry, so I'll I'll restate the question.
20:11So the first question was uh if we generate the context on the fly.
20:16Sorry, what was the second one?
20:17The second one is we get conflict information.
20:20Oh, how do we resolve conflicts, right?
20:21Because like Yes.
20:22Okay, that's great.
20:23Uh and I think there's going to be probably a bunch of talks at the conference about about memory.
20:27And I think uh that's kind of one of the the the hot spots of like agentic memory like you have agents that derive knowledge from like a bunch of different sources, but like at the end of the day, you need a human to tell like yes, this is actually true.
20:38this thing is like incidentally true or like it's true for like this specific case.
20:42Okay.
20:42Um in terms of deriving the information um it's uh usually going to be uh you don't snapshot it to some extent because of that like I think the the value of this ends up being that you have the information on the fly.
20:54It's available like it's the latest uh latest and greatest.
20:57um if you have it if you cach it then like it risks being stale and like there's all complexity around it in terms of um what the the way that we we've historically dealt with um
21:12authoritiveness I guess for lack of a better word um has been to provide that as uh as a feature for like the the human in the loop to like be able to say like yes cool this is verified
21:24this has higher authority than this other piece of content um that's harder to do with things like I don't know juror work items.
21:30I think juror work items in general.
21:33You can do it but it's just more manual work and like the risk by automating that stuff is like oh you end up like having things automated automatically end up being authoritative which is weird because you can end up in that in that same situation.
21:50All right.
21:51Are we interneted?
21:52I think so.
21:53I think we should be interneted now.
21:54Yeah.
21:54So, the p I think the page is like hopefully working for everyone now.
21:57Again, I'm going to put the I'm going to put the go link plus the uh QR code back up right now.
22:03Um, but just to clarify really quickly, everyone should have a number.
22:07Um, and that will be your login.
22:09Uh, so we will be using something called like Alaskan University.
22:12So, whatever number you have, it should be student and include and and then the number of whatever number that like Shane provided you earlier.
22:19Um,
22:19[snorts]
22:19that will be your login.
22:20Uh, that will be your login to test the jury coding agent.
22:23um and kind of just play around with it within the juror ecosystem itself.
22:26Uh so please go to uh so from that confluence page you should have the link to our uh to our kind of like test jur workspace um as well and from that you can kind of just like build whatever you kind of like um we do have a
22:41starter repository for you guys um uh it's called like Atlassian swag within bitbucket itself um I can show that in a little bit but please feel free to like once everyone logs in uh start playing around with it and start playing around with the agents directly.
23:01Oh, the the network is still not working.
23:08Oh, you are you on the same network?
23:09I'm on the same network.
23:10Yeah, it should be on the It take a little bit of time.
23:13It take a little bit of time to load.
23:15Yeah, we we Yeah.
23:16Yeah.
23:19Yeah.
23:19Yeah.
23:19Yeah.
23:22I think we've been we we talked with the uh the IT folks here and to see if we can uh see if there is any any resolution on that space.
23:28I do apologize for that.
23:36Yes, I can put up the QR code again.
23:40Yeah, no worries.
23:42Uh numbers, numbers, numbers.
23:43Yeah, sorry.
23:44Let me
23:45Did Shane go away with the numbers?
23:46Pardon?
23:47Uh where's Shane's numbers cup?
23:49Shane's number cup.
23:54Good question.
23:58One second.
24:02Need more numbers for the numbers.
24:10Okay.
24:18The best of numbers.
24:20I think that's a prime.
24:22Maybe
24:28I don't think anybody's able to connect to the Wi-Fi.
24:30Yeah.
24:30Yeah.
24:30Yeah.
24:30I think that was like the the the great unsolved problem of uh 2026.
24:34And I think if if anything's going to save us from uh AGI, it's going to be the Wi-Fi.
24:41Or if they fix it, then we don't need savings, so we're good.
24:47Uh
24:52[laughter]
24:52just you gota you got to tear the traffic, you know, you got to optimize it.
24:55Uh,
24:57okay.
24:57Until we figure out the Wi-Fi.
24:59Any other questions that I Oh, please go for it.
25:08Verification requirements.
25:10Okay.
25:10Sorry.
25:10So, the question was like, can we talk a little bit about the verification of requirements aspect?
25:14Um essentially that that boils down to uh when we have like a specific work item that uh the juro coding agents work agent works on that goes through and
25:25um like the juro coding the we expect the work item to have a checklist of hey these are the these are the acceptance criteria and as the juro coding agent uh generates the code for that particular task it'll go in and like verify
25:41those specific things once it thinks it's um given that it has like a the way that the jury coding agent runs is like it has a full sandbox.
25:49It can install things, it can run things, it can do whatever whatever it is with that specific repository.
25:54So as long as you have the instructions in the repository in terms of oh this is how you run this, it can pop open a browser and then like go and like run that those verification steps.
26:04What is the sandbox?
26:06Uh what is the sandbox?
26:08Where is it?
26:08Is it like hosted?
26:10Yeah.
26:10Yeah.
26:11Yep.
26:12uh
26:13or is it something that you can bring in?
26:16We don't have that ability yet.
26:18So it's like it's it's a it's a proprietary Atlassian sandbox.
26:21Uh it's again it's a container so you can mostly run like whatever it is that you want like it has Docker enabled and like the rest of stuff.
26:26Yeah.
26:27Go for it.
26:28Just related to billing and that type of thing your own key.
26:32It sounds like some if you wanted to all of these agents can use your API keys with with anthropic and open or they currently hosted models.
26:43Uh right now you can use uh so the billing is through Atlassian credits currently Shane am I wrong?
26:52Uses an API key so it uses your own API keys.
26:55So there's
26:56every one of these agents can be pointed officed wherever okay but they don't need to be
27:02yeah the juror coding agent supports like multiple uh I think right now we have uh all the anthropic models all the open AAI models and uh I think that's it.
27:10Do we have Gemini as well?
27:12I don't remember.
27:12I think we can guarantee both the cloud and the open AI models.
27:20Yeah.
27:20Yeah.
27:21Yeah.
27:21It's credits.
27:22It's a slightly different but like Yeah.
27:23Same same thing.
27:24It's like pay as you go.
27:29On top of that,
27:30there's usage.
27:31Yeah.
27:31Yep.
27:32Is this all available now?
27:33Sorry.
27:35Uh jeroding.
27:37Yes, it's out now.
27:37It's been out for for a hot minute.
27:40Yeah, go forward sir.
27:41So mention credit usage credit like for the developer like at the end of the month in terms of from a company perspective like you need to report like what kind of work that you have done or it's just
28:00some
28:04uh the if I recall correctly and I may be wrong I need I might need to follow up.
28:08Sorry, the question was uh how do we do uh credit reporting at the end of the month given that there's like multiple experiences that where people could you be using AI credits for right
28:18um if I recall correctly the way that that gets that gets split split out by product so like specific user has used a JCA with like this specific in this specific experience
28:28sorry JCA being juro coding agent right in this specific experience and like it uh that's kind of the the level of um uh detail you get right now not like a test based.
28:41Yeah.
28:41No, no, not yet.
28:42um that is something that we're working on because at the end of the day I think the thing that like will uh it's kind of becoming apparent is as uh we have like more and more people using
28:52uh uh agentic coding uh you want to make sure that like the ROI is there that like people when somebody is working on something is going to going to spend like I don't know uh 1 billion tokens on a thing then you want to make sure that like that is something that you actually want to have done
29:07as opposed to like somebody's open claw running somewhere off in the uh in the periphery.
29:18Uh, sorry.
29:19So, like what's the latest and greatest on the Wi-Fi?
29:33Uh, any other questions while AGI is coming?
29:37Yeah, sorry.
29:38Go ahead.
29:41Whoa.
29:42Sorry.
29:44Uh, login, Andy.
29:46There we go.
29:46Right.
29:46Right here.
29:47So, there should be a link.
29:49You can just like type this out yourself.
29:50Uh, 1-ish-devo.atlassian.net.
29:56Um, with your with your number, uh, your login will be student your number at university.
30:03And the password is right here, uh, ZXC ZXC on the academy.
30:08We're going to get it wiped at the end of the um at the end of the session.
30:11So like if it's open.
30:12Yes.
30:13Some questions matter.
30:14Doesn't matter.
30:14Okay.
30:15Thank you.
30:15No worries sir.
30:19I got you.
30:25So let me see really quickly.
30:26Let me come down.
30:33Yes.
30:36Oh god.
30:37Can you number you go to made with you?
30:40Made for you.
30:41Sorry.
30:41Made for you.
30:42Here.
30:44Made for you.
30:44Yes.
30:45There.
30:47Oh, there's
30:51this is the closest.
30:52Oh god.
30:54God.
30:55One second.
30:55Please.
30:56What is which?
30:56What is your number?
30:5795.
30:58Give it a second.
31:12Yes.
31:15Yep.
31:26That's a great question actually.
31:28So I think right now um Andre you actually I think you have a little bit more here.
31:32Uh Sorry, the the question was uh whether we do anything smart about model routing.
31:37So right off the bat with uh the geocoding agent currently I think we let you select which model like you want to route to.
31:46Yes.
31:46But I think that's the uh you can also with uh um sub agents and skills you can decide like oh this particular uh skill needs to use like so and so agent.
31:58I don't think we do any sorry so and so model.
32:00Uh we currently don't have anything out uh that would do be doing like doing smart routing.
32:07That is something that we're working on with uh in in autodev to help you essentially figure out like oh this specific task this is a high cool level.
32:14We've seen success with it and like give you the metrics in order to be able to like have that decision and make sure that that's a policy that you want to enforce or not.
32:22Please stacking that in terms of like you've got your agents then you've got your skills.
32:29Do we define the agent or agents preset by you and then we define the skills that they might automot?
32:44Uh the skills are basically up to you like we have like a bunch of bunch of skills but specifically for the juro coding agent there's a um bifurcation to some extent between like
32:56rode dev
32:59uh as as they were called previously um and with the juro coding agent that's essentially we uh give you a harness and there's model selection at the top level of the harness and then with the skills you can figure out like right all right like this is for this specific task, for this specific skill, I want this
33:17model, this sub agent needs to be using blah blah blah.
33:25Um,
33:28yeah.
33:28Yeah.
33:29And that is up to up to the model.
33:32Yeah.
33:43Oh, we're still doing Wi-Fi.
33:45Oh.
33:47Uh, okay.
33:48Well, we're still doing Wi-Fi, please.
33:56Yeah.
33:56Yeah, that's probably what we're going to do.
33:58Uh, sorry.
33:59The question was like, can we keep the demo site open for a hot minute so that like when people actually get to Wi-Fi, they can actually go through it?
34:05Yes, I think that's a that's a reason to ask.
34:07Thank you.
34:10Um,
34:10how long will this be?
34:13We can keep it open for as long as we really want.
34:16I think maybe for maybe if like for the duration of AI AIS fair,
34:20uh, we'll keep the we'll keep the workshop open and so when we do have Wi-Fi, I do apologize for the Wi-Fi issues, you can kind of like play around with for yourself and kind of like play around with the different tools we we will we do have.
34:30Uh, one thing to sort of keep in mind as you're like inputting stuff onto these sites, as you can tell, the login are like very secure.
34:39So, like don't put in anything that you wouldn't be comfortable putting on a GitHub gist somewhere or like on a a bit bucket snippet.
34:47So, yeah.
34:48Uh, keep that in mind as you're kind of adding stuff to and and messing around with all these things.
34:55Um, all right.
34:57Any other questions that we can go into?
35:00Please, sir.
35:05What's the importance of cleaning up stale context?
35:10Interesting.
35:10Can you say more?
35:11Like what do you what do you mean?
35:14Atlas for 15 20 years.
35:17There's tons of product
35:23out of all the things.
35:26It's a good Okay.
35:26Sorry.
35:27So like the good question.
35:28Um the the the question is essentially so like you have a large organization that's been using Atlassian for like 10 15 years.
35:34Um how important is it to like prune some of that stuff so like it doesn't end up showing in search results or like the agent doesn't end up picking it up while it's looking for random stuff to give you an answer to the question.
35:45uh
35:47[sighs]
35:47uh our experience has mostly been okay with this by in in large part because we tend to favor like more recent results.
35:56Uh I think that's where authoritiveness sort of comes in where like if you have something that's been around for 10 years and it's still like the whatever tip of the spear is like that document for 10 years ago.
36:06um those I think we currently shipped a thing in uh well shipped a thing a while back in in um conference that's the verified badge like that's basically the thing that we use internally to say like yes cool this is like still the latest and greatest and that this context
36:19should be favored uh across others Hello.
36:43Okay, awesome.
36:44I think we've had a few questions about standards.
36:47Uh, unfortunately, the standards aren't built into the demo itself.
36:49Uh, but I do want to take a little bit of time just to like show you how standards kind of work within, uh, the Robo Dev product itself.
36:55uh and kind of kind of like how we would be assigning standards overall uh to your specific repositories itself.
37:02Um so just a quick demo here.
37:03Uh so this is the standards page.
37:05Uh we have a few built in already, but what I want to show you right now is like you know how to create your own standards.
37:10Um there's also a bunch of different templates you are able to select.
37:13Uh but what I'm going to do is just create a new one.
37:16Uh new standard.
37:18Uh give a quick description of how what the standards really is.
37:22Uh so so this one would be design token enforcement.
37:26So if we go to standard standard name go to call it design token enforcement uh add the description itself and add the rules.
37:35So after you add the standard
37:39[snorts]
37:39and the rules itself you can specify which repositories you want these standards to be applied to.
37:49uh you can select manually all the repositories or repositories matching a pattern.
37:53Um so as these repositories load you can decide okay let's apply to these following repositories and then after every all that is set up you can just create your standard
38:02um like I said within software we have our own standards um instead of trying to enforce it uh manually trying to like enforce that within the PR process itself what we can do is you can set up the standard set up set up which which repositories you want the standard to uh be enforced on and the robo
38:19dev jur code like the coding agent itself will be running through the pull request process and adding those uh adding comments to see if those p if your pull request isn't adhering to all the standards that
38:30um are kind of the specified within the standard itself.
38:33Um a few other things I did want to demo uh something called AI planner as well as like autodev.
38:41Um we can't give a full-fledged demo but we do have what we do want to iterate that like you know we're trying to go from the left or right of code and I would say like AI planner and autodev are more on the left of code.
38:51So, um I don't know if we have a full demo, but I can kind of talk about it with the uh slideshow itself.
39:08So, we're going to get start uh autodev.
39:10Actually, Andre, uh you've been working on autodev pretty heavily.
39:12Can you kind of like give us a brief uh brief overview of what the main goals you want to accomplish with AutoDev?
39:18Um, and kind of kind of works as he sets up his mic.
39:23Trying to figure out how mics work.
39:26Supposed to touch it.
39:29Not supposed to touch it.
39:30Oh, there we go.
39:30Oh my god.
39:31Thank you.
39:32Um, all right.
39:34Sorry.
39:34Let me I'll just be here then.
39:37Um, okay.
39:38So, for for autodev, the this is essentially our for into like the AISDLC orchestration space.
39:45Um the thought process is to basically get you back some capacity from all the in order to let your engineering team focus on feature work like the uh AI the for autodev we
39:58the initial um the initial release is going to focus on grabbing work items from your backlog and grabbing the ones that we feel confident that uh the agents can handle autonomously
40:11and then just putting up PRs for your team to review in order to make sure that like we're working on the right things that we're abiding by the right standards and then sort of iterating on your harness as well so that we can get code bases agent ready in a fairly light touch way.
40:25Um do you want to oh sorry do you want okay and this this essentially kind of goes where uh uh establishes where where we're trying to go.
40:34So um you we right now again the definition is going to be like we will we will grab your backlog and then we will allow you to also help us
40:45identify which specific things you think auto you would like autodev to to grab.
40:50Um we have some some thoughts around like where we would like to this to go next.
40:55So there's standards enforcement um and I think Andyy's going to talk probably talk about that in the context of the review.
41:02So the way that we think about that is I don't know standards are to some extent pretty similar to linting to some degree but they're a lot more uh
41:12loose uh when you have the uh agents enforce them.
41:16We have one part of the I guess the the the linking workflow in forward-looking things in the reviewer right like you add standards to code reviewer and that takes care of any code that will end up landing in your codebase going forward.
41:31uh we have a workflow for autodev where we would be able to scan existing part like your existing codebase and then figure out how we land uh PRs progressively in order to get all of your existing stuff into compliance.
41:44There's some complexity there in terms of like so nobody's going to approve a PR that's like change half your codebase because now you have a new standards uh or nobody's going to approve anything that like causes like a high amount of churn.
41:55So with autodev we want to be mindful of how much work we're actually hoisting on your team.
41:59uh nobody wants to spend time reviewing uh 50 PRs a day.
42:03So like we have configuration for letting you say cool like my team wants to get assigned I don't know each individual gets assigned three five PRs a week and like those are the things so that makes helps you get like incremental progress and also frees up your team to actually do the feature work that
42:20they uh they actually want to do and like may move forward your u other initiatives.
42:25Um and then yeah that kind of goes into like both initiatives as well.
42:28I think the the space there is I kind of think about it as organizational steering.
42:33So this happens a lot in uh bigger organizations where you have I don't know everybody's trying to move from whatever some version of Rails to a different version of Rails and it's a big thing and we have like 10,000 services and like those are pretty similar
42:46uh and some of the learnings some of the memory some of the workflows that you go through are fairly uh sy.
42:53So you would want something that is concentrated on one team and they're just like spinning up PRs and like the agents essentially do the do the work and then you'd only have a human do the coordination
43:03and then we also tie the loop around that with uh some sort of memory.
43:07Sorry.
43:07So again that's that's sort of the vision that that's where we see this going and that's where like we see this being beneficial to our workflows because large organization
43:14bunch of services that's where we spend a bunch of time that like we would like to not have engineers be involved in that deeply and we see agents as like a really big opportunity to to move that ball forward.
43:25I think you can grab on to okay um this is the other thing uh if uh autodev is doing work uh on your team's behalf autonomously you kind of make want to make sure that like you are aware of like what it's doing and want to be are able to interject
43:41when you want to when it's either picking up the wrong thing or like it's not assigning the wrong the right person and we will learn from that experience and uh the reason why you want that to be in your jura board is because that's the thing that most people stare at like it or not when they're trying
43:55to like figure out what's the next thing to to work on.
43:58Um, and we were talking about this before uh just the um the the work in progress limit is the thing that we were that I was mentioning where you want to make sure that like your team isn't getting inundated with PRs or like other work that might not be necessarily the work that you want to drive forward.
44:14Um the other thing that we we're also thinking through is um limiting cost which I think is very much uh in the site guys now.
44:24Like you don't really want agents running roughshot.
44:26Like if you want to spin up a a stream of work for autodev, you want to be able to set limits for how much we want to spend per ticket, how much we want to spend per week.
44:36All that stuff seems like it's uh just table stakes at this point.
44:41Um yeah this is essent like the whenever we whenever our dev starts working on stuff we will um give you like this this broad reviewer like all right how much how many things needs
44:52uh oh man I I forgot to mention this okay need requirements ah uh for whenever we troll through backlogs there's um a very high likelihood that we have work items that are in your backlog that don't have that are underspecified
45:07like there's a thing that somebody put in like that has just one summary line that says do the thing.
45:12Um those can be valuable but we don't know and we don't have the confidence that we can work on them because we don't know exactly what the goals of them are.
45:21And that's where like we will pull a pull in a human in the loop and say hey we have this work item.
45:26What is what actually is this thing?
45:27Do we want it to work on it?
45:29Then can you help me specify it?
45:30And then once we get to a large good level of specification that's when like we can get to actually working on it.
45:35And that's where again needs review is when we pull in a human to to actually do the review.
45:39Um the other bit that we've also noticed with some of these autonomous workflows is the the the reviewing step is where things go to die.
45:49Um because people again are uh busy with other things.
45:53Um, and that's one of the systems that we're putting together in order to reach out to people at the opportune time in order to pull them in and go like, "Hey, can you like
46:01take an eye to take a look at this real quick and like figure out whether um whether it's good to merge or like it's it needs to go away."
46:08And the [clears throat] the chief vehicle for that is going to be Slack for for the the initial release of AutoDev.
46:15Um, and then again, we were talking about this, the ROI insights, you making sure that like you're getting um, you're not overpaying for the amount of work that we're doing and then it's work that you actually want to get done and being able to like sort of push the limits of like how much stuff you actually want
46:28us to look at.
46:30Awesome.
46:31So, I think the one I think the one part that we want to touch upon again uh, really quickly is this part, the the planning part.
46:39Uh, so we do have one more product out coming out called AI Planner.
46:42Uh I would say this is more geared for I would say like um say uh if you have a spike or project managers or say you're a manager that wants to see if something is feasible for your current repository.
46:54That's where AI planner comes in.
46:55you can kind of provide your idea of what you want to build and what the a planner will kind of do again goes through our context engine kind of figures out okay what is in your repository
47:06what is your plan and is it feasible and if it is feasible it will tell you and the second part is it will kind of give you a breakdown of what you need to implement
47:13um if you do want to build out the spec specific uh specific feature or the specific you know idea that you do kind of have it will give you the plan uh and it kind of give you
47:22a few ways to execute it um talked a lot about the context engine.
47:28Um Shane is actually here.
47:29He's working a lot on code chat.
47:31Um Shane, do you want to talk a little bit more about the code chat capabilities and like all everything within like a context engine itself?
47:39Are there slides?
47:41[snorts]
47:41Are there slides?
47:42Uh there's no slides.
47:43All right.
47:44Um yeah, so he was mentioning uh context engine code chat.
47:47The way we've been thinking about it and at at lastian we have many repositories.
47:53is I think we have 15,000 repositories across the company and asking a simple question like where do I implement this change or where is X Y and Z implemented or what services are using this other service
48:05are really really complicated and so at the core of the context engine it is narrowing down the retrieval and the search space to scope it to like where this answer is most likely
48:17uh lives within the code and then reaching out from there and like digging into the code to answer the questions and then on top of that len and all of the
48:26um information from the teamwork graph.
48:27So documentation historically out ofdate and inaccurate.
48:33If you're searching only across documentation, you're going to get something that was true at some point in time.
48:38Um but it's also a lot of really valuable information.
48:40So this project built this thing at this point in time.
48:45It was true.
48:46It's no longer true.
48:47But the reason why those things are decided at the time are valuable for making changes in the future.
48:54And so we can leverage the code as it exists today, the decisions that went into the code as it existed at some point in the past and use that to inform
49:04um implementation paths or or or planning for you know why we should change code in a certain way.
49:11Does that make sense?
49:12Any questions?
49:13But yeah, like historical information about why decisions were made are something that's often left out of um agent generated code, but it's really important um historical context for like not re remaking mistakes of the past.
49:27And so that's how we think about the context engine.
49:30Awesome.
49:30Thanks.
49:31We have a question over here.
49:39Yeah, it's um a lot of it is used like AI planner is leveraging it for a lot of things that I don't know if either of you can speak more about but a AI planner is using it a lot right so if you have something that you're trying to figure out
49:53um how to build it can look at documents for projects related to it context around it and so it's it's kind of at the core of a lot of these new capabilities we're building out
50:02um autodev uses it um and you could probably speak more to it.
50:06Um AI reviewer doesn't use it a lot today, but we're starting to experiment with how to pull more um into there.
50:12And then the other place it's exposed today is within Robo Chat.
50:16So if you're talking to Robo and um we're in early access with a few customers now.
50:22just announced it a couple months ago, but if you ask it code related questions today, it answers those through um mostly Confluent Search, some other uh connected tooling.
50:33And then if you have GitHub or Bitbucket, it will use their search APIs.
50:38Um but if you are you know have this new capability called code context enabled it will have like a more cohesive understanding of the code grounded in the truth today the historical decisions that went into it and answer those types of code related questions more effectively
50:57sorry that access
51:03yeah so the teamwork graph CLI um if you've seen that that's where we're starting to expose a lot of these capabilities Um the context engine itself is more agentic search is the way to think about it.
51:14We have a lot of the core primitives that we use uh that we'll expose to your agents through the CLI but I think over time you know we might consider moving more of those into there as well.
51:24Yeah, team graph CLI has access to all the um code context skills.
51:31Question in the back.
51:33Yeah, I had two questions.
51:34The first was on standards.
51:36Um it looked like standards were like pros and marked down.
51:39Is that correct?
51:40That's correct.
51:40Yes.
51:41Do you have any way to map those to something deterministic like a CI check?
51:48So kind of like kind of like what we say like a check like a checkbox kind of like uh scenario, right?
51:53Um I think just off the top of my head what we would probably have to do is kind of like implement that within a the markdown itself.
52:01But that's actually really interesting use case where you can just provide it like specific like can like binary yes does this satisfy no does this not satisfy.
52:10Um so I currently I don't think so I don't think we can prov we provides that out of box but I will relay that over to the standards team to see if we can get that in the future.
52:18Yeah, that could be really neat to make it a little bit more.
52:21That would be awesome.
52:23Right.
52:27Epic that you can break it down and decompose it into multiple ages like waves of ages.
52:33Usually there's foundational work that will work down the line.
52:39foundational unblock.
52:49Right.
52:50Just to reiterate uh I think your use case is kind of similar like where you have an epic um and you maybe want that epic to be like broken down into like smaller issues and then within those smaller issues you can have like you know like a wave of agents kind of like execute each smaller piece.
53:03Is that correct?
53:04Would you want like maybe like one step further once you implement like that first wave like is it if like if we could get a second wave going?
53:11Yes.
53:11Okay.
53:12Awesome.
53:12Yeah.
53:12Yeah.
53:13Good great question.
53:14So I actually played around with this yesterday.
53:16Um with JCA what you can do you can take an epic uh you can get JCA can you break this down into like smaller issues for me and then
53:24uh it will be able it it is able to break it into like smaller juror task if you provided a jur epic.
53:29After that what you can what you are able to do um there's a few like automation rules that you can kind of do let's say let's say like uh with an automation let's say like you have the J epic kind of like be broken down once it's broken down you can have automations kind of like re start
53:43a JCA uh instance has start a JCI kind of session and it is able to like autonomously kind of like develop each
53:49and I don't think there's anything that would prevent you from kind of extrapolating that further right like here's all the stories I want break these down into multiple epics right the foundational work the like V 0, V1, V2, and then just have it, you know, further refine the tickets into there.
54:03And so I think nothing would prevent you from doing
54:06Oh, you can keep going.
54:07Yeah, you can do flexible.
54:09Keep going.
54:16Sounds more like a JCA automation kind of like I'm not sure if Robo directly could do it, but I definitely think you can do with automations.
54:23Yeah.
54:23And I think autodev might be able to do some of that.
54:27Auto dev.
54:27Yes.
54:28Okay.
54:28I do that with gas.
54:30[laughter]
54:30You guys are doing great.
54:32Thank you.
54:32Yeah, no worries.
54:34question.
54:36I just want to understand that actually my code is I change the code.
54:58So just to reiterate uh let's say you make a code change within like your IDE.
55:04Oh yeah.
55:05Um so I think you have to think about with JCA that like it it is isolated very much the agent sessions.
55:11So if you say like make a change within the agent session itself um then it will be real time.
55:16Uh if you make it make a change within the VS code instance within the cloud then yes all those changes will happen real time and you will be able to be able to reflect that within the cloud session itself.
55:25Um whatever changes you make on the ID it's a separate kind of like a separate instance right.
55:29Uh so yeah with with agents think about it like all within the cloud itself any change you make in Jira session jur agent session itself within jura then then then all of that will be real time
55:42question over here
55:54you I guess you
56:03so just to iterate like so it sounds like your instance is more like on prem and you want to run a bunch of like the agents like on on prem.
56:11Is that correct?
56:12Not even like I guess like how is the develop development environment like is there like a docker file that we would have to use to
56:30that's
56:32that's docker.
56:32It is docker.
56:33Yeah.
56:33Yeah.
56:34Okay.
56:36Cool.
56:36I think we have time for one more question.
56:44That's a great question.
56:45Uh, we actually just started uh supporting skills I think just not too long ago.
57:02Right.
57:02Yes.
57:03So yeah, I think for just AI reviewer itself, we do support skills like uh like I think less than a month ago, we released leads.
57:10Sounds good.
57:13Cool.
57:14All right, thanks everyone.
57:15We got to go.
57:15We'll be around if there's any other questions.
57:17But
57:18yeah, thanks a lot.
57:18Thanks a lot everyone for joining uh joining our workshop.
57:21Thanks a bunch for the uh the tech staff for getting all of our Wi-Fi issues sorted out.
57:24Big shout out to them.
57:25Um hope you guys have a great rest of your week during the uh during the conference.
57:29Thank you so much, guys.
57:45[music]