Read the transcript: 200 turns, 6,196 words
00:06Abrar Mahmood Hello and welcome to the AgentSee podcast. I am Abrar and we have.
00:11James Hartt James, hi.
00:12Abrar Mahmood Hello. And today we're going to be talking about the questions before you build. I'll go over to you, James, now.
00:20James Hartt Yeah, so this kind of, comes out of, basically the millennial toy of the playground of the sort of early 2000s or maybe even Late 90s of the little sort of circular ball with the else marquee LCD screen that asked you Questions to see if it could guess what your what you were thinking about it could have been like an object or a or I don't think it Was always an object I don't know
00:45Abrar Mahmood Yeah.
00:45James Hartt Anyway software you could say at the end of the the day becomes an object that you can somewhat hold so that's basically What we've tried to do is come up with you won't necessarily have to answer every questions but if you did Then it would have a pretty good idea certainly a frontier agent frontier LLM front would figure out Exactly what kind of software project product you would want So yeah that's kind of where where it came from and the questions are built on both of Mine and Abrar's experience. We've got a number of sections with four questions in each of them. And we're gonna go through all of them and figure out why, yeah, Why we think if you do answer all of these questions, you're setting up the initial conditions for a good successful software product. And you can do that. Yeah it's something that I think maybe some people that are not technical might miss And then if you do end up making the wrong decision then it can really come back to bite you later on down the line So yeah let's carry on
02:02Abrar Mahmood So perhaps we can go into Where it came from so I think in terms of my ideas I've been working with businesses That have been vibing their own workflows quite a bit and of course as we know we're now in a place where engineering and the Actual grunt work of it isn't necessarily the bottleneck like that's almost it's almost like for me I see it as the code being a bit Almost like a byproduct
02:31James Hartt Yeah.
02:31Abrar Mahmood So I think that I've spoken about this previously but like we've had like the area the error of like compiled code which We Kind of seeing like code as it's written by agents at the moment as almost like compiled code but a bit more Readable and like
02:50James Hartt Yeah.
02:50Abrar Mahmood We don't always like read it line for line and write it line for line but we do look at it when we see relevant so going back a step I Think it's now more important than ever to ask the correct questions to start with and I think that's kind of where our experience sort comes into this Going into the questions so that's kind of like where I find it fits in really well and I know there's other tools out There and that do a similar sort of thing that I've personally worked with a few of them but I think the ones that we have are Very applicable to where we are taking AgentSee and so I think yeah it will be really good to start going into The different sort of areas that we ask our questions on So I think we'll start with the idea
03:42James Hartt Mm-hmm.
03:42Abrar Mahmood And so do you want to go into that a little bit
03:45James Hartt Yeah so I think that this is the question that really can only be answered by a client it's something That yeah what should it do like it's kind of rough it will immediately touch the potential industry Like or sector or whatever you want to like yeah there's lots of whys and you might put in an example or Something it's that there'll be the most personal questions and there'll be the ones that have probably Occupy your mind the most as someone either that's non-technical or even is technical you're thinking about yeah the problems Of your given sector so yeah four separate questions that kind of does it like live alongside an Existing workflow that's very important to know like is it brand new like obviously there's a term called legacy like will there Be sort of legacy code or like APIs that needs to work with also think again things that you might want to consider of Like has it been built before and was it a success or did it peter out or did it just spectacularly fail those are sort of Things that you do need to contend with and they're actually the things that certainly research agents are absolutely brilliant at Figuring out what has already been made So yeah it really just comes down to the idea whether it will like it's this whole tool is meant to it's not meant to Get in the way it's meant to sort of support and encourage the idea rather than sort of say no because everything's possible in Software and everything is becoming more possible with sort of agent workflows so yeah the idea is this obviously it's the first thing that you need to start Talking about
05:33Abrar Mahmood I think I think the other thing that's really important is we're starting with the idea and I think that's a really good place to Start because now that it's more easier than ever to build software with agents like the temptation is often to just Go in and start knocking something but I think like starting with the idea is a really good way because it kind of it kind of gives you a time To reflect on what the problem is that you're actually trying to solve so like coming back to like the implement well the coding Not being the bottleneck it kind of like reduces the barrier to entry to build something
06:10James Hartt Absolutely.
06:11Abrar Mahmood But actually what you're building like is it solving the problem that you are actually intending to solve and thinking About it from this point to start with is definitely a really good place to start because yeah like just don't want to go in and start building something and realize Actually this doesn't really solve the problem
06:29James Hartt No.
06:30Abrar Mahmood That we intended to solve so we've just gone in and we've created a database and we've just like written Some APIs but does it actually solve the problem I think that's I think that's something that I've personally experienced in the Past so I personally also start with the idea first as well and I think it's definitely the way go so that's Yeah that's really important to make sure you've got nailed down before you start going going forward So the next one is that is sort of a care needed around the domain and the idea so perhaps we can go into that A little bit
07:10James Hartt Yeah absolutely I mean hopefully the idea has narrowed a lot of that down and in fact an agent if you said oh I want to make A heavily regulated health tech idea it and I want to launch in the UK like you're going to be you already know some of the Names like the national clinical excellence or whatever lots of the different regulators that might exist and procure NHS procurement and stuff like that you're gonna have to possibly contend with and that obviously comes with the highest form Of care that's required and every single feature will have to be reviewed to make sure that patient data is not not being Leaked or and loggings like redacting things correctly and yeah a lot a lot more care and attention needs to be spent whereas if you're Not working in that or tangentially I don't know something to do with fitness like fitness would be a lot less like it's important it's health Data you don't want it to be leaked but it's not someone it's not it's not going to ruin someone's life if it gets if it gets leaked
08:13Abrar Mahmood Yeah. I think it's also worth mentioning that for particular domains, for example, in health or even in finance or In any of the other domains, there may be particular regulations around the data which aren't necessarily making the data more Secure from a technical perspective, but it might make it there might be particular regulations in order to comply with How data is like data retention policies So like perhaps for example if the data needs to be anonymized within like days of it being recorded for that particular domain Like these are the kind of things that we need to think about to start with to ensure that we comply with this going So that's a really good reason To go into this as well.
09:00James Hartt You've got ISO standards for this kind of stuff and different like compliance things that there's lots of SaaS products for that we could Use them you could again I feel like quite a lot of it could be handled by agents like on our side it's that it's that Kind of diligentness that you could really just chuck a load of tokens at to kind of just kind of make sure that Someone's always looking at this like Because in the old world, things got missed because
09:28Abrar Mahmood Yes. Yes.
09:30James Hartt Of human inattentiveness rather than Something that can just be Pointed out. And that's their sole goal to look at every single diff that comes through and say, oh, is there something that Needs to change based on a given standard that it can always load in?
09:50Abrar Mahmood Yep.
09:50James Hartt So so yeah and that it kind of like that's the rest of the section like the care needed it's it follows on from the Idea really nicely yeah again personal it's really just around personal information potentially and money and yeah if Anything's involving children like that's obviously a an extra like level of sort of safeguarding stuff that you need to consider Lots of other data stuff and it is really the most it's the stuff that requires the most diligent attention
10:22Abrar Mahmood I think
10:23James Hartt So
10:23Abrar Mahmood The other thing is that it's one thing complying with it but like these standards and regulations change
10:30James Hartt Yeah
10:31Abrar Mahmood So keeping up to speed with it is an overhead in itself as well so like if we like if we need to Be considering regulations like keeping up to speed with it and making sure we are adapting that going forwards and if anything That doesn't comply with new regulations we need to ensure that we've got our existing Implementation brought up to that so that's another thing that of course that we need to be taking into Consideration quite seriously
11:01James Hartt Yeah and it's something that only this these kind of models can really like offer in a in a like as These kind of models of developing software can offer it's just a yeah a huge step up in potential Quality if you're constantly asking the question on every single change something that you just wouldn't be able to do anyway yeah so I think I Think we've covered that roughly like
11:28Abrar Mahmood Yeah.
11:28James Hartt Yeah it's something that would be client by by client always Idea by idea
11:38Abrar Mahmood So the next section is the running conditions.
11:41James Hartt Yeah.
11:42Abrar Mahmood So what that looks like and the sort of questions that we might want to ask in terms of when it's running, what
11:49James Hartt Yes
11:50Abrar Mahmood Sort of constraints do we need to work Okay.
11:52James Hartt So this again it's kind of informs us for infrastructure decisions but also architectural decisions Like in our like I've in my career I've seen examples where scale was built for but it never arrived and Then that made I guess they need to somewhat needlessly made it hard to iterate whereas if you Don't build for the scale and the scale arrives then you often have A lot firefighting and distractions that result in again you sort of moving at a snail's pace potentially on Other priorities when they come at the same time so it's yeah it's sort of asking what does your busiest day look like and for some clients it Could be this is an internal tool we'll never see more than a handful of people using it at the same time to social media virality Which obviously requires you to think like every single feature needs to be thought of could this hit scale Do we have load testing for it like can we simulate the load because sometimes you can play is even an absurd to simulate the Load or it's not a real test So yeah it's kind of like running conditions are something that if you get wrong at the outset you can have some major Headaches and I'd and you have to make some sort of big potential changes and that would again sort of slow you down so Yeah and also uptime as well so the scale and uptime like if you want the highest scale and the highest uptime you're talking about sort of sort Of insane levels of architecture you really do need to test whether you need that and make sure that you do because again it will require Much more much more care and attention than their sort of simple app would and then yeah the platforms as well that we run on you can have Any platform like from a sort of I I certainly with these tools I feel so confident in building front end for any platform like it could even I would go far as Saying like I'd be very comfortable making say mods for a video game which was probably way outside of my comfort zone like just Because the agents they join the gaps like you know exactly what you want but you can join the gaps but having said that I think there are some Interesting decisions that you certainly have to make like say if you're doing a mobile app your most people want them on all the Popular platforms but
14:21Abrar Mahmood Yeah,
14:22James Hartt I think
14:22Abrar Mahmood Why wouldn't you, right?
14:23James Hartt Yeah exactly but you would you can certainly make some more interesting and subtle decisions today where you could Mainly develop or say develop for an iOS app and then every single diff that happens you send off an agent it's like Repeat this work but for Android and that would be enough of a prompt to then have two native apps that means you could get all the benefit of native Apps with basically next to none of the overhead like you still have the sort of some sort of testing overhead but like These apps app platforms are not so different that you couldn't work in A really nice workflow like that could get you all the benefits for none of the real drawbacks that
15:05Abrar Mahmood Yeah.
15:05James Hartt Used to exist and yeah things like do you need to support accessibility do you need language support do you need a full Offline mode like that's something that you really need to decide up front because if you're trying to hammer in later it gets quite it can get quite Messy
15:22Abrar Mahmood Yeah.
15:24James Hartt And then yeah sort of integrations like what do you see it integrating with existing software like Sas products like do you expect to put inference out there as in like LLM inference out there do you expect People to have the ability to call expensive APIs potentially expensive APIs and how do you protect yourself so Yeah like think that. Roughly anything that I think I've missed there Abrar
15:55Abrar Mahmood I'll tell you you've covered it but I do have like some personal experiences which I think are definitely worth pointing out A big one is of course offline functionality so
16:07James Hartt Yeah.
16:07Abrar Mahmood I've had I've had a number of projects where I've of course when we were working at the agency previously as well as Some more recent personal experience where offline has been a requirement that would come in At the eleventh hour
16:25James Hartt Yeah.
16:25Abrar Mahmood And of course Especially when working on mobile apps you kind of need to architect the entire thing to work offline
16:34James Hartt Yeah
16:34Abrar Mahmood So I think like with our sort of setup of having our questions I think like offline functionality is definitely a real big One that is going to completely shape the way the application or
16:46James Hartt Because
16:46Abrar Mahmood The
16:46James Hartt Yeah
16:46Abrar Mahmood PWA
16:46James Hartt You
16:46Abrar Mahmood Or
16:46James Hartt Sync
16:46Abrar Mahmood Whatever is
16:47James Hartt You've got
16:47Abrar Mahmood Built
16:48James Hartt To sync stuff conflicts like yes it's a
16:52Abrar Mahmood Yeah
16:52James Hartt You've got to decide like every app might be different in how they decide how their offline and functionality works
16:58Abrar Mahmood Yeah
16:59James Hartt So and that's often not a technical decision it's a product and yeah client decision
17:05Abrar Mahmood Yeah,
17:05James Hartt So yeah
17:07Abrar Mahmood I think also from a testing perspective, I think offline functionality is something where it can look like It's working, but then because you've got so many edge cases, you've really got to test for every edge case to ensure it's truly working.
17:22James Hartt Yeah And
17:24Abrar Mahmood And,
17:24James Hartt Yeah
17:24Abrar Mahmood Like...
17:24James Hartt Automating automated testing offline mode is always a spoof right like because
17:30Abrar Mahmood Yeah.
17:31James Hartt You've got to be connected to To a device somehow or like
17:36Abrar Mahmood Yes.
17:36James Hartt Or yeah it's gonna it's gonna you're gonna have some level of like not a true test and then yeah
17:41Abrar Mahmood And
17:41James Hartt Offline
17:41Abrar Mahmood Even
17:42James Hartt Mode essentially always has to be truly tested offline But yes
17:48Abrar Mahmood Just put testing for that. Bad network connectivity if it's
17:51James Hartt Yeah
17:51Abrar Mahmood Not truly offline that's another thing which
17:53James Hartt And
17:53Abrar Mahmood Can be a little bit tricky as well
17:55James Hartt Even yes even simulating bad network conditions
17:58Abrar Mahmood Yeah
17:58James Hartt Is tough it never really behaves like a shoddy of 4G connection
18:05Abrar Mahmood No
18:06James Hartt So
18:06Abrar Mahmood Yeah
18:07James Hartt Sorry.
18:07Abrar Mahmood You often end up with just slow network requests which isn't really the same thing is it Yeah no I think that's yeah I think that's good I think that's definitely what we want to be covering in terms of the running conditions And like I guess like someone who's got some technical experience with the cloud like in terms of scale like why not just put it On like serverless compute that automatically scales like there's reasons to not do that as well so
18:34James Hartt Well yeah I've seen horizontal scaling go horribly wrong and it's not yeah if you've not got your checks and Balances then
18:43Abrar Mahmood Yeah
18:44James Hartt You end up with a absolutely ginormous cloud bill
18:47Abrar Mahmood Yeah
18:47James Hartt So
18:48Abrar Mahmood Yep
18:49James Hartt You don't want those happening especially to yeah you don't you just don't want those things happening so you need to nail these things down and then If you are you can you can answer these in any sort of way it's like we could see viral of social media load but we'd prefer it to sit Everyone behind a queue or
19:07Abrar Mahmood Yeah.
19:07James Hartt We'd prefer it to or all of these paths that will get hit by behind the queue and then you can have like pretty modest infrastructure To serve all those people but then again there are some use cases that That you still need a huge amount of vertical and horizontal like capacity like it's just yeah unavoidable so yeah It really comes down to your idea and where you want to take it
19:32Abrar Mahmood I think I think also in terms of like some of your suggestions there so like one of the things that you mentioned was we do want to support Viral scale but have everyone sit behind the pew like I think some of our customers might not necessarily know that's an option so I think one of the really good Things about our questions is that they are non-technical so if like for us like that's kind of like an Example of where our expertise can come in to offer these suggestions of how we can solve these problems so like we don't need everyone To know like that is an option so yeah I think that's a good thing to point out
20:10James Hartt I think yeah when you can sort of like we're going to update these questions as people use it and as we Get some feedback and as we use it ourselves like future episode will will be us using these questions to see if it results in a In a better output for a an actual idea and yeah the all these the shapes of each of these sections really sort of Really sort of dissect an app and what it should do
20:40Abrar Mahmood So the next area is essentially the shape
20:44James Hartt Yeah
20:46Abrar Mahmood Of the Thing we are looking to build. So, do you want to go into that, so that in terms of what it talks to?
20:53James Hartt So obviously we're at the entire point of AgentSee is that the agents like competitively note I very much doubt Anyone in the next few years will not be writing most of their code with agents but having said that if there are Bits of the app that like someone really needs to know intimately then you need to like that needs to be Outlined and then again that sort of harks back to like potentially regulated industries and certain things like that but like we Would obviously like you're working with us because we expect AI to respond to your Your wants and needs on the product so Yes we would be heavily saying yes it should be drafting and generating a bunch of the code for example all of the code But could it like be involved like will LLMs be involved in the product itself is that a good idea are you Using the right tool for the right problem like yes ChatGPT and Claude and someone can do all of these amazing tasks but they Sure as hell can like spend your wallet so we need to sort of check the shape of your like of the AI usage like other things Of like how often do you want updates to the app like will it change a lot after you launch it like some people will Know that this is just an internal tool and it just needs to do this and it needs to stay secure and get updated very rarely Or when a CVE comes in a sort of critical do you know what's it what's the CVE stand for it's critical something Like
22:34Abrar Mahmood Critical,
22:36James Hartt Vulnerability something
22:37Abrar Mahmood Yeah, that was
22:38James Hartt Yeah
22:38Abrar Mahmood It, yeah.
22:39James Hartt And yeah so things like that again these are brilliant for agents to monitor so that no one's really putting in that effort they just look at the Dependency list see what's going on and then and then change it and patch it so that those are the sort of things that we will that will be offered Without even thought but like do you feel like you will have to make constant feature updates to satisfy A market like that's the sort of question that we're like wondering like and that's down to the idea that's down to the Industry that's yeah we just it's something that will sort of shape the app and then yeah things about and some of these questions once we're Getting to them they could be answered quite easily by a previous question so if it sounds like we're repeating yourself it's because possibly are but like what's The absolute worst thing that could happen with your app and for every app that would be a different thing it could be that it goes down when People are in the office it could be that personal data is leaked you could select them all and say I absolutely do not want any of these things To happen At all ever and yeah we can we can make some fairly strong assurances in on the outset that won't Happen that you'll have sort of protections around those things happening and then finally yeah sort of infrastructure costs like some people will have a An operational budget an opex of a given amount and they can't stray from that and yeah if you've got a really simple internal tool you can Be talking about pennies a month and you could be going up to hundreds tens of thousands if you're using some expensive APIs out there That still need to be part of the product like I can think of sort of anything to do with location is very expensive for example like if You're wanting to use sort of Google APIs for they're always very expensive because they've invested the time in all The those silly cars driving around and photographing everything and listening to everyone's Wi-Fi so Like
24:40Abrar Mahmood Yeah.
24:41James Hartt You yeah those are the sort of things that you can capture reasonably easily up front so that there aren't any shocks and it's just Useful to know so yeah I think that's kind of like roughly like these are really getting into the weeds like with this tool may Not even answer this like ask you this question because you might have might have figured out it
25:01Abrar Mahmood Yeah
25:02James Hartt Might have figured out already what the answer three is.
25:06Abrar Mahmood I think these are good prompts so when working with AgentSee it gives us the information we need to help you Build a product that will actually work sustainably so an example of that is like if we're Building a product and it's of course an online product What does a offline flow actually look like and that's something we can help our customers with so like god forbid the App is offline for a period of time and it is an absolutely critical workflow but it could even be as simple as Where our customers where our customers intend usage of the application the Wi-Fi drops out or there's a power cut now what does an offline Flow for whatever it is you're doing look like so I think like that's something that I've actually been quite inspired by at Amazon So Amazon. Are quite big on making sure their online tools have an offline equivalent so if their systems go offline for Whatever reason like what does it look like to do this on pen and paper so I think that's like a really big thing to ensure that you can continue with Trading and business or whatever it is that you are doing and I think that these sorts of questions can help us figure out what that Is and I think that having these ideas again is kind of where our expertise come in rather than just going off to an agent and knocking Something out
26:33James Hartt Yes,
26:33Abrar Mahmood So I think that's really important as well So the next section is the reality so should we talk a little bit about that
26:43James Hartt Yeah I think I think it's this is sort of a this is really if it this tool hasn't really nailed down absolutely everything but maybe there's a couple In there it's really like Like it's kind of How do you measure success of something? Like it could be, yeah, like how you know That the effort was worth it For this app. And again, for an internal tool, that could be very easy to quantify. But for a sort consumer Product, it could be a bit difficult to know in advance what success looks like. Like again, spend As well. So, like what's your, like, if you do know your budget, like, you will sort of welcome to tell us, or you can, you can Sort of, give us a range of, like of what you feel comfortable to talk about. And yeah, like, sort yeah, for the Minimum viable product and then the very finished like not a lot more to do I guess it would sort of cover the like Yeah what's that like and then also sign off as well like do you have a team that needs to confer To figure out oh is this what we want or is it a single person or is it no one like it like we're Getting into that software realm where if you are if you like reasonably have described something you can Be pretty sure it's going to make what you want or at least a very good approximation that needs tweaks rather than Full rewrites so it's I guess it's these are some fairly vague questions at the end of it and then Also yeah ongoing maintenance
28:29Abrar Mahmood Yeah.
28:29James Hartt So if you have a semi -technical person or someone that their role could be like restoring from a backup Or running a script or again getting an agent to run some of the assets that have already been created for the app and then Trying that first and then if it if it all goes to pop then you do come to us to figure it out like is there someone that can be the Sort of first port of call for like someone and that may be a no and it may then fall to us to figure that out for you and again Agent like there are a lot of agent workflows that can be execute now to Self-heal a system when certain things happen well so again how happy you are to rely Upon that depends on your idea depends on your industry depends on all of the previous questions and like yeah that it's It sort of rounds off the questions and figures out like your appetite for owning the app when it's completed
29:32Abrar Mahmood Yeah I suppose the other thing is because agents as you said are quite good at self-healing maybe it's even We work with someone who's completely non-technical and part of what we're doing needs us to help whoever We're working with in order to deal with these situations like I know of course like it's quite easy to go to Go to your agent and say oh it's broken fix
29:59James Hartt Yeah,
29:59Abrar Mahmood It
30:00James Hartt Yeah.
30:01Abrar Mahmood And more often I guess like as the agents get better it's going to get better at fixing it which is which is really interesting as well but I think Like perhaps it's a little bit more more than that we need to do to work with our customers to say okay so We've got a problem with the data what does it look like to actually restore this and I think like part of what I've done personally is try To get to a point where the documentation we have in our code base is very good for agents to read So when you do go to your original and say there's a problem I need fix it knows where everything is and it's laid out in a sensible way so It can go and figure out how do a database restore and promote restored database for example so that's yeah that's a really big thing as Well that I think I've done quite a bit of past which has which is just getting better and better as the as the agents improve really
31:00James Hartt Once you're done with this like this tool if you've used one of the frontier models anything I mean this should behave just fine on Opus Or Fable or like all of the yeah ChatGPT products it should it should do it reasonably well you Should come out with a brief and the brief would will obviously encourage you to send it to us like via email or you can reach out on the socials And like basically have a chat about your idea and see if it's right for us right for you and yeah there's Certain things that we would always do Like for every single product like yeah things that keep us sane as developers you would always be able to walk away With your code there would be a bunch of there would always be automated tests of some Variety automated review Like there would be a tracker that either you use that we integrate with or we use and you'll have full access to and yeah we try and Keep everything in plain words and come to an agreement and retainers and things that you need to really support a full software Product and everything hopefully like yeah above board and you're happy with and know exactly how things would proceed
32:18Abrar Mahmood I think the other thing to just highlight it's now becoming more important than ever to be on top of security
32:25James Hartt Seed
32:25Abrar Mahmood Like
32:25James Hartt Yeah
32:26Abrar Mahmood Now I think these days like literally every other day you see something on X or whatever
32:32James Hartt Mean we've had
32:32Abrar Mahmood Or whatever's being hacked
32:33James Hartt I think we've
32:34Abrar Mahmood And
32:34James Hartt Had what is it, 50% of all CVEs ever created in the last three months
32:40Abrar Mahmood Yeah
32:40James Hartt Is because
32:41Abrar Mahmood Yeah
32:41James Hartt Of just the sheer like it's a flood and this is going to be a theme of the next two years I would say it's a Y2K level event Where like every single software product that is out there is scrambling to plug up all the holes that can
32:55Abrar Mahmood Yeah.
32:55James Hartt Now
32:56Abrar Mahmood Yeah,
32:56James Hartt Be revealed by agents so
32:59Abrar Mahmood And
32:59James Hartt You
32:59Abrar Mahmood I
32:59James Hartt Best
32:59Abrar Mahmood Think
32:59James Hartt Believe
32:59Abrar Mahmood There's also,
32:59James Hartt We'll be
33:00Abrar Mahmood Sorry.
33:00James Hartt Making sure that yeah if a CVE rolls in then it's patched very quickly and automated like and it all Goes out very quickly so
33:10Abrar Mahmood I think there's also given the frequency of the attacks there's also a lot more Targets because there's so much that's vibe coded out there
33:21James Hartt Yeah
33:22Abrar Mahmood Without the technical experience backing it to ensure best security practices are put in place
33:29James Hartt Yeah.
33:29Abrar Mahmood And I think of course like we don't really like I think like at an agency we don't really talk about that too much typically Traditionally but that is something that you get when you have experienced software engineers working with agents To build like there is a level of security in place that ensures that we have we have our best chances of Not falling into these situations.
33:54James Hartt And These are all things that we're setting up on our own open source stuff so
34:02Abrar Mahmood Yeah.
34:02James Hartt There's there's evidence that we're actually doing it to our own products so yeah it's all like that's the thing that I Enjoy the most about all of these new workflows is that a small team can do things that an Absolutely giant team was capable of like that sorry in the old world only a giant team was capable of that Obviously moved at a glacial pace so it's yeah it's the beauty of these tools yes they've taken as they've taken An amount from the sort of enjoyment of software engineer but they've also given a bunch back because you can as a small nimble team you Can do you can achieve some frankly insane stuff like
34:47Abrar Mahmood Yeah
34:48James Hartt Then insane levels of quality and diligence that everyone's going to need in the years coming So
34:56Abrar Mahmood Yeah I think yeah I think just to that point as well like Working with a massive with a massive company also has its costs as well whereas working with a smaller team like us you Get you get the experience of software best practices and product best practices without having the Overhead of a massive engineering team so you can scale you can scale that up with your agents quite well but of course in Terms of costs like this
35:26James Hartt All
35:27Abrar Mahmood We're not paying for a massive for a massive skyscraper
35:31James Hartt Yeah absolutely and yeah we're constantly going to be working on these workflows and figuring out the best way to Interact with clients and if a client can make changes with agents like they're going to be near free like And we're going to find out ways to work in that workflow so that we can yeah best support best support clients
35:57Abrar Mahmood Anything else or
35:58James Hartt No I think
35:59Abrar Mahmood Should
35:59James Hartt I think
35:59Abrar Mahmood We?
35:59James Hartt That's pretty good and yeah I've enjoyed chatting about our little questions it's a nice whimsical way of defining Software
36:08Abrar Mahmood Yeah.
36:08James Hartt And who doesn't need a bit of whimsy in these trying times
36:11Abrar Mahmood Yeah.
36:11James Hartt So
36:12Abrar Mahmood Yeah.
36:13James Hartt Yeah
36:14Abrar Mahmood Cool. Well, thank you very much, everyone, for listening in today. In the next few episodes we will be going in A bit more detail and perhaps even trying out an example of the questions so we're looking forward to seeing you there thank you very Much
36:30James Hartt Absolutely
36:30Abrar Mahmood And see you
36:31James Hartt And
36:31Abrar Mahmood Next time
36:31James Hartt And yeah we're gonna also be inviting on guests starting from sort of when this episode airs so or hope to hope to get on some people Either potential clients or people that run agencies themselves and yeah really looking forward to just having a chat with people and yeah Should be fun
36:48Abrar Mahmood Thank you very much, bye
36:50James Hartt All
36:50Abrar Mahmood Bye!
36:50James Hartt Right cheers bye