Episode 434: Business Central, Noobies, Magic, and AI United

In this episode of Dynamics Corner, Kris and Brad are joined by Microsoft MVP Jeremy Vyska. The trio discusses the intersection of AI and Business Central development. Jeremy shares his journey from being skeptical about AI to embracing AI tools, highlighting his innovative Nubimancy Project. This project provides an AI-based team for developing Business Central extensions. The discussion also addresses the ethical considerations surrounding AI, the creation of fictional business scenarios for educational purposes, and the practical applications of AI in coding and project management. The conversation offers valuable insights into how AI can enhance productivity while upholding ethical standards.

Send us a text

Support the show

#MSDyn365BC #BusinessCentral #BC #DynamicsCorner

Follow Kris and Brad for more content:
https://matalino.io/bio
https://bprendergast.bio.link/

Episode Transcript

SPEAKER_00 (00:05):
Welcome everyone to another episode of Dynamics
Corner.
What does MCP stand for?
And what does it have to do withanything in the fantasy magical
world?
I'm your co-host Chris.

SPEAKER_03 (00:20):
And this is Brad.
This episode is recorded onOctober 9th, 2025.
Chris, Chris, Chris.
What does it have to do with thein the realm of a fantasy world?
I don't know.
But we did have the opportunityto learn about that because I
grew up playing those tabletoprole-playing fantasy games.

(00:40):
And I wish that we still livedin a world where those were much
more popular than they are.
With us today, we had theopportunity to speak with Jeremy
Viscous.

SPEAKER_01 (01:08):
Yes.
Hello, hello.

SPEAKER_03 (01:10):
How are you doing?

SPEAKER_01 (01:13):
Uh busy, as always.

SPEAKER_03 (01:15):
Busy?
Always busy.
I like the new background.
It's uh it's different.
It's different.

SPEAKER_01 (01:21):
Yeah.
Well, I mean, you know, thingshave changed up a little bit.
Uh I'm I'm not with SpareBrained anymore, so I don't have
my own uh little studio space.
Uh so you're getting kind ofwhat you get, which uh I'm
working for a Swedish companynow, and uh we are in a shared
office space, uh, so we havelots of little meeting rooms and

(01:42):
whatnot.

SPEAKER_03 (01:42):
But uh it's it's uh getting back to working in the
office, as they say.

SPEAKER_01 (01:50):
In a fashion, definitely.

SPEAKER_03 (01:52):
Yeah, no, that's good.
That's good.
It's I like being in the officeat sometimes.
I I think being remotesometimes, I think we miss out
on a lot of that interpersonalcommunication type things where
you can yell over, or you know,yell over a cube wall or knock
on someone's door, or you do runsome into someone in passing and
you can ask them questions.
Where I think in the remoteworld it's a little more

(02:13):
challenging because you have tocheck are they online, you know,
are they in a meeting, or youknow, if you have teams or
whatnot.
So I think sometimes people area little more apprehensive to
maybe reach out and talk tosomebody.
So I think uh it's it was alittle more challenging.
So I think going to the going tothe office is it has its
benefits and it also has itsdrawbacks.

SPEAKER_00 (02:33):
You got your water cooler too, like it's nicer, you
know, and just stop over andhave some water cooler talk, you
know, as a group, rather thantrying to get teams going, like,
oh, he's green or red or he'syellow, he's not in yet.
But you know, he's actuallyright there.

SPEAKER_03 (02:48):
Yeah, no, it's it's good.
You get some of that small talk.
Um, so again, welcome back.
But uh so much I want to talk towith you about, and uh I I don't
know, it's probably one of thoseconversations where my mind is
going so fast that I won't beable to uh my mouth will have a
challenge keeping up with it,and I'll try not to go out the
place.
But before we jump into that,would you mind telling us a

(03:09):
little bit about yourself?

SPEAKER_01 (03:11):
Sure, sure.
Uh let's see.
Um I'm Jeremy Visca.
I'm an American who relocated toSweden ten years ago.
I've been working with BC now uhfor coming up on 26 years, uh so
a real long time.
I've been a BC MVP for fiveyears.
Coming up on five years,something like that.

(03:32):
Uh in that neighborhood.
Time's a blur.
But um the uh you know, writtena couple of different books
about BC over the time, uhreleased a few odds and ends,
which we've talked about onprevious episodes, uh, that
people can make use of in thecommunity.
Um and these days, which I thinkis led to the uh us chatting at

(03:54):
this particular moment in time,these days uh people are being
amused to watch how quickly I'mkind of running uh with the AI
ecosystem for business centralworld.
So uh lots going on.

SPEAKER_03 (04:07):
Lots going on, lots going on in uh your your API
book I still have as areference, which is nice.
Uh and I do appreciate all thatyou've done for the community.
And as you had mentioned, Istarted watching you with this.
I I've always followed a lotthat you've done, and you know
we've had lots of conversations,but recently you started a new

(04:28):
project, right?
That it's the world of AI, andyou've been working with BC for
a while, and then one day I wasfollowing uh some of the message
that you had, and you referencedto an additional project that
you had started.
Uh, would you mind telling us alittle bit about that project?

SPEAKER_01 (04:45):
Sure.
Uh I mean let me bring you onthe journey, like well, bring
the audience along uh for thefirst time keeping up.

SPEAKER_03 (04:52):
But the first question I have to get to cut
you off, how do you pronounceit?

SPEAKER_01 (04:57):
The the uh newbiancy project?

SPEAKER_03 (05:01):
Yes, that's it.
Thank you.
Thank you.
I want to go on the journey now,but that was like one of the
things I'm saying.
I'm like, I call it like I wascalling it like nubibacy.
Like I was coming up with allthese phrases, and like that's
the first thing I have to getout of Jeremy is how do you
pronounce it?

SPEAKER_01 (05:17):
I carefully test uh and make sure that everything I
bring on the show is somethingdifficult for you to say.
It's it's mission accomplished,right off the show.
Hey, me too.
Just let's say, Chris, watchout.
I know he's from the northeast.

(05:38):
You missed that.

SPEAKER_03 (05:39):
He he he he relocated to Sweden, but he
wasn't too far from me.

SPEAKER_01 (05:45):
Uh which is an interesting aside to the things
we've talked about before.
Uh, for fans of the show who'vewatched these various episodes,
and in the past we've talkedabout Braider uh as that like
API factory.
When I shut down Spare Brain atthe end of last year, I did that
product didn't die.
That became open source.
Uh and it's on App Source stillnow for free with no licensing.

(06:08):
So any of the folks who havewatched the previous episode
where we're talking about whatyou could do with it, just go
hit install and start using it.
Um you can now climb through allthe source code and get to know
how the heck did I do all thatstuff?
Go look.

SPEAKER_03 (06:22):
That's good, excellent.
So Braid is now open source andstill available in App Source.
Excellent.
Excellent.

SPEAKER_01 (06:28):
So all right.
So the uh the the actual storywe wanted to talk about.
Okay.
So um conference season inEurope happens pretty heavy in
the spring.
We've got a lot of greatconferences from March through
June.
Um, there's a lot of thingsgoing on in that time stretch.
You know, the holidays arefinally over, it's nice enough

(06:49):
to go outside, and and boy, dothings get busy.
So there's all the days ofknowledge and dynamics minds and
BC Tech days and all thesedifferent great events that uh
happen throughout the springseason.
And um I had the pleasure ofattending one of the days of
knowledge uh events in thespring uh and spent a lot of

(07:09):
time with Tina um from uh I wantto say continuum?

SPEAKER_03 (07:14):
Companion, I think.

SPEAKER_01 (07:16):
Companio.
Thank you.
Yeah, that is correct.
Um and uh, you know, he and Ihave always gotten along pretty
well, and he did some uh greattalks about AI is more than just
fancy autocomplete.
Um which up until that point, myGitHub co-pilot experience in
Visual Studio Code was it's justas annoying as when you're

(07:36):
trying to type on your phone andit's suggesting all sorts of
nonsense to you.
It had as much value to me asthat.
Um it didn't really feel like itwas earth-shattering, and yet
we're you're hearing keynoteafter keynote that's talking
about all these things that AIcan do for you.
And I'm like, what's thedisconnect?
How is it possible that 30% ofcode is coming from people

(08:00):
hitting tab?
That doesn't make sense.
Exactly.

SPEAKER_03 (08:04):
And yeah, it's funny that you mentioned that because
with Tina, that same session,and he has to um polish it every
couple weeks, it seems like,because of the way the content
changes.
But at Directions North America,I saw his session on that, and
that also what drove me into it,where I was floored at how you

(08:29):
could use AI with ALdevelopment.
So that video is available.
The BC Tech Days video did comeout uh recently.
So anybody who wants to watchthe session or listen to the
session uh and watch the screen,um, that's also on YouTube,
which it's well worth thelisten.

SPEAKER_01 (08:46):
Yep.
So the that was the seed ofunderstanding for me was get
copilot out of ask mode.
It's not just a conversationalpartner, it's not just an
autocomplete that's waiting foryou to hit tab and you know,
you're just trying to indent aline, the thing's getting in the
way.
It's it's a very differentexperience when you switch over

(09:08):
to that agentic mode.
Um and kind of hearing that, Iwent, I still don't I still
don't feel it.
Like, you know, I I I've beendoing this a long time.
It can't possibly be faster.
Frankly, I'm a very fast coder.
A lot of the times I have to crcross-check what I'm doing with

(09:29):
someone else to go, what is areasonable time estimate,
because I think I can do thisfaster than probably I should
offer to people.
Um and uh so I I was prettysuspicious.
So uh since that was in May, um,and I was speaking at Tech Days
in June, I set myself up withthe challenge, uh, and I didn't

(09:52):
tell anyone about this untilhalfway through the talk.
Uh, in BC Tech Days, um, I waspresenting a whole bunch of
stuff on making job queues runin parallel um and how to make
uh the most of the fact that BCgives us the ability to run
multiple tasks per user, andthey really opened up the job

(10:12):
queue to really handle a lot ofbandwidth.
What can we do to use and abusethat level of capacity?
So um, you know, I did I plan todo this talk on the parallel
development.
Um I've gotten feedback sincethat some people have gotten
operations that were nine-dayoperations down to six-hour
operations.
Wow.
So uh it's making a bigdifference for some folks.

(10:35):
Um but I set myself out thechallenge of if I'm gonna do
this, I want to challenge me,not just my audience.
And I made it that I am notallowed to touch or write the
code.
I have to only writeinstructions for the agent to
understand what I'm trying toaccomplish.

SPEAKER_03 (10:57):
That is an interesting challenge to
yourself that somebody who'sbeen coding for as long as you
have, who has the ability tocode quickly and understands the
language, to now basically gointo instruct mode.

SPEAKER_01 (11:12):
Hands off.

SPEAKER_03 (11:13):
Hands are off, and now you have to write, which
that's a whole other topic iswriting instructions because I
was I've been following thisproject and what you've been
doing.
And I I just I don't even I'mspeechless with this whole
instruct mode and uh and such.
So that's uh that's a challenge.

SPEAKER_01 (11:31):
So um so it was a little bit interesting.
I I've happily at the beginningof that tech day's talk, uh,
where I was gonna show off allthis code and how it works and
everything like that, which youknow, um that demonstration went
really well for you know thehour or so.
Um I I happily uh lifted from AJKaufman uh his opening line that

(11:52):
he likes to give for some of hissessions.
That uh congratulations, you'rein a session that is a
co-pilot-free session.
And just as he's experienced,the crowd went kind of crazy,
like happy because they're sofatigued about hearing about AI,
because they had the sameexperience I did.
It doesn't provide value to me,and yet that's all I can hear

(12:14):
about.
That's terrible.
Yeah, it's really frustrating.
When you're like out of thatlockstep and suddenly all of the
mainstream content means nothingto you, it doesn't feel very
good.
So there's a natural pushbackand reaction to that, and so
people you know applauded withthat.
Um the fun moment for me wasthen an hour, hour or so later,

(12:38):
to be saying, by the way, I liedto you at the very beginning of
this.
Uh, not only is it not acopilot-free session, every
single bit of code we've justlooked at, and the product of
that code was copilot written.
And then we spent some timetalking through how that
actually came about.

SPEAKER_00 (12:59):
Behind the curtain, there's actually a copilot.

SPEAKER_01 (13:04):
I actually did get a couple of negative reviews from
that of I can't believe that youwould do that to us.
It's like it's still real code.

SPEAKER_03 (13:15):
I I think that I think that's a little fun.
So so let's take it through uhthrough your journey.
So Tina sort of sparked, I guessyou you sort of sparked uh your
uh inquisitiveness for AI, uh, Ithink it's it's it's funny.
It's uh I had the similar I hada similar experience, and that's
also what started my journeywith this was seeing that.

(13:37):
Um so you then had a sessionthat you did hands-off.
How did that go?
And is that what also sparkedthe project that you have?
Um because you had someinteresting blogs on there about
your journey, uh blog articles,excuse me, on that project
website about your journey aswell.
Is that together?

SPEAKER_01 (13:55):
Um, that was uh I kind of think of it a little bit
as Tina uh planted the seed andthe challenge to myself, let it
take root.
And um after that experience ofokay, it could do the job.
Like at any point in time duringthat week, if it really couldn't
do the job, I still had enoughtime to write the things myself.

(14:18):
I could go hands back on thewheel and and not worry about
it.
Um but it succeeded.
I got what I needed out of itfor producing decent code.
Um, and I genuinely was well alittle bit surprised by that.
So I started pushing myself umuh a little bit more to explore

(14:38):
it in the day-to-day job.
But you know, as always, whenyou get back from a conference,
you've just got the pile ofthings you need to get done.
Um so I didn't I didn't reallyhave time to pick it back up.
And then um in July, I took offthe first half of July for
vacation.
Uh, but when I came back andeveryone else was still on
vacation for the rest of July,making it a really nice quiet

(15:01):
time, I challenged myself to aJuly of AI.
What stuff am I missing that Icould be doing better?
And that was the beginning pointof where the blog series that
you're referring to kind ofpicked up on the uh
newbimanse.com site.
Um, and that was also the seedof the idea of that project uh

(15:23):
at that moment in time of umthere's always more to be
learning and everything likethat.
Um, you know, there's a lot ofstuff that we're doing in our
organization with DevOps andpipelines.
And, you know, over the summer Ihad been meeting with a bunch of
people talking about all thethings that we're doing in
pipelines and how can I bringthat information to the

(15:44):
community at large?
I can't share our organizationDevOps.
So newbie Mancy as a project andthe AI development in hand in
hand kind of happened at thesame time.
Of what can I do to share allthis stuff that's in my head?
Um, what can I do to bring moreknowledge about all these

(16:06):
different pieces?
Is it unfoldingconstruction-wise as a story and
a narrative?
Uh no, not as not as smoothly asI would like it to.
Um but uh in July I spent a lotof time trying to figure out the
AI did an okay job for me.
It did okay, but I feel like itcould do better.

(16:29):
And it was right around thatsame time uh Dmitry Katzen was
working on the AL guidelinessite.
He was trying to put togethersome instruction set that would
help the AI understand AL bestpractices better.
Um, he was contributing that tothe ALguidelines.dev project
under the vibe coding section,which you can go grab now and

(16:51):
plug into your agents and allthat sort of stuff.
And there was some really gooddiscussion in that thread of
that pull request of what arepeople using for instructions
and what makes goodinstructions.
Um and out of that, I startedrealizing what made good
instructions.
And so by the time my team gotback at the end of July, I had

(17:15):
this massive pile ofinfrastructure and education
that the co-pilot had beenthrough.
And the way that training allthese agentic coding things is
as long as you've built up thatinstruction set, every agent is
kind of a disposable thing.
It picks up where right whereyou've left it every single time

(17:35):
it's starting at the beginning.
So if your instruction set isright, right at the beginning,
that's where it starts everytime.
And my team started trying toget their hands around it, and
it was it was a learning journeyfor them because they're in the
same position you are of this isjust annoying and it doesn't do
anything.

(17:55):
What's the point?

SPEAKER_03 (17:56):
Yeah, so let's take a step back as you're talking
about um instruction set andagents.
Uh so for those that arelistening that may be in the
same position that we were, whatis an agent and what is the
instruction set that you'rereferencing?
And maybe we can, I'm sure we'llget into how do we give that

(18:20):
agent those instructions.
And you're also talking aboutmultiple.
I I see I told you I'm justgonna run with this, but we'll
start with that.
That's when you started withthis infrastructure for your
team.
Sure.

SPEAKER_01 (18:32):
Okay.
The one of the things that Ilearned early on with agentic
coding, it's like you're adevelopment manager who's just
been given a really enthusiasticrookie on a lot of coffee, ready
to go, let's do this.
But they they they've read onlythe textbooks in school about
development practices.
Maybe they've heard of BusinessCentral.

(18:54):
And but they're ready to go.
Um it doesn't always work great.
So um what I kind of think ofinstruction files as, and this
uh is jumping ahead a littlebit, but also MCP servers, it's
like taking that juniordeveloper and saying, Okay, I'm
glad you're here to report forwork at 8 a.m.

(19:15):
all espressoed up and you'reyou're ready to jump in.
Uh first go read thisinstruction manual and then come
back to me.
And now it's time for you topick up your task for the day.
And so when I'm talking aboutinstructions, it's literally a
pile of uh usually files uh thatare basically you are a business

(19:37):
central development helper, andhere are some of the rules that
you need to follow.
Here are some of the things youshould and shouldn't do.
Here's some of the stuff youshould know about, and some of
the stuff you should know not todo.
And that's just a language filebecause these are running on
large language models.
So uh one of the recurring jokesthese days is what pro what

(19:58):
languages do you program in?
English.

SPEAKER_04 (20:01):
Yes.

SPEAKER_01 (20:02):
Um so instruction files and again MCP servers are
basically you've got that happyjunior uh go go read the manual,
go take this class, and when youcome back from those, then we
can start working.

SPEAKER_03 (20:17):
So you you you mentioned the MCP server as
well.
Again, it's it's it's all Ithink we'll come full circle to
others.
So you have instruction sets,which are something that you
give to an agent, which we'llequate to as a person in this
scenario that you're tellingwhat to do.
So then you can have differentinstruction files for different

(20:37):
agents to do differentfunctions.
Um, and in here in the casewe're talking about a business
central developer, but you couldhave someone who is uh someone
who's documenting code, or Imean, how would you break those
agents up?
And then how does it fit with anMCP server?
And you know, I'll ask you the$10,000 question.

(20:59):
What is an MCP server along withall of this?

SPEAKER_01 (21:03):
So yeah.
Um okay, so to take those inturn, yes, uh instruction files
are basically you've got thisLLM, it wants to help you out.
What does it need to know tohelp you out?
And what are the things it canand should do, and what are the
things it can't do?
So, for example, the sales orderagent in Business Central, the

(21:24):
way they've built that up isthrough very careful
articulation of instructions.
It's following instructions justlike our agents are in GitHub uh
in the uh GitHub chat window inyour Visual Studio Code.
Um uh that sometimes is referredto as the prompts, prompt
engineering, and all that sortof thing.
That's the what instructionsdoes the model understand to

(21:48):
accomplish and all that.
Um so effectively, yes, you youby giving different instructions
to the same engine, you give ita different job.
So uh that's why I kind of liketo think of the agent as that
enthusiastic young person readyto show up and do whatever.
What are you doing today?
Well, today you're adocumentation guy.

(22:10):
Today you're a sales orderprocessor, tomorrow you're gonna
be an expense agent.
It it really doesn't matter, uhit's still it's showing up and
what is the uh uniform you'reputting on it on it when it
shows up.

SPEAKER_03 (22:24):
Okay, okay.

SPEAKER_01 (22:25):
Um then to the uh MCP server question.
An MCP server, uh it's annoying,uh it stands for model context
protocol, which means nothing toanyone uh unless you're really
deep in the weeds.
It's it's one of my challengeswith AI at the minute is that
there's such a big gap betweenthe leading edge people and the

(22:46):
rest of us that they'll oftenforget that we'd have no idea
what you mean by rag versusvector.
Like what what what language areyou speaking?
Um the uh the thing about theMCP, it's basically just a
toolbox that can contain theinstruction manuals and also
some tools that the uh guyshowing up that day for work, he

(23:10):
can unpack the toolbox and let'sgo.
That's it.
That's really all it is.
Uh MCP servers allow the agentto understand how to do stuff on
your behalf, whether that comesthrough knowledge or whether
that comes through talking tothird-party systems.
For example, an MCP server I'musing all the time is connecting

(23:32):
to Azure DevOps or connecting toGitHub.
It understands how to log in asme on my behalf.
Jeremy sent me to create thisissue here, have an issue.
Okay.
So the MCP ecosystem is how manydifferent toolboxes can we
invent for these agents so thatway when you show up, uh when

(23:54):
you get there and you'restarting your workday and the
agent shows up and it's ready togo, what toolboxes can you give
them?
And the cool thing about an MCPserver is that these are just
bits of code and bits ofsoftware, so there's an infinite
number of them.
Once someone makes that toolboxavailable, everyone can use it
and everyone can say, here yougo, to your agents.

(24:14):
And that's pretty amazing.

SPEAKER_00 (24:16):
I like that explanation.
It's probably the easiest whereI can now picture what is an MCP
and what it does.
Because that it's it everyonetells me different things.
I'm like, okay, can you justexplain?
That's probably the best oneI've heard so far, Jeremy.
So thank you for explaining thatto me.

SPEAKER_01 (24:33):
The reason I also use that metaphor is um MCP
servers, you get you once youstart getting used to them, you
can go, oh, I should install ahundred of them, and then I've
got all these different tools.
The thing is, is that when youum install these MCP servers for
the agents to work with, um theagent has to decide what's the

(24:54):
right tool for the job.
And that thinking process isburning tokens.
It's it's using up some of itsability to get the job done, has
to stop and think about whattools to use.
And so tools like Visual StudioCode will actually yell at you
if you try to use too many uhMCP tools.
And I like to think of that asthe look, the intern can only

(25:16):
fit so many tools on his beltbefore it's just gonna weigh the
poor intern down.
Don't give it a narrow focus ofwhat tools to think about,
because if it has to try tofigure out between 50 different
toolboxes what should I use todo what you just asked me to do,
it's gonna spin for a long timeand it's gonna be very
inefficient, and inefficientcosts a lot of time and effort.

(25:39):
And part of being ethical aboutAI usage, which is a challenge
uh for energy and waterconsumption, part of being
ethical is also making sure thatif you're using AI, you are
being efficient with it ratherthan having to argue with it for
a couple of hours to get whatyou need.

SPEAKER_03 (25:57):
Yes.
Okay, so just if I can peel itback from your explanation,
which is as Chris had mentioned,was great.
MCP server is basically astandard set of tools that a an
agent, which I'll call it, uhthat eager, that eager uh uh

(26:18):
employee that shows up onemorning to be able to do their
task.
And that tool can be somethinginternal or it can be something
that they can use externally,such as like a telephone to call
somebody to make an appointment.
That telephone's a standard thatanybody can use, and then
someone on the other side willpick up and say hello.
So it's a standard way ofcommunicating, and then you can

(26:40):
have multiple of those running.
So each one of those has a setof tools for the agent, and they
the agent will choose whichtools to do the job.
So if we have an MCP server foruh hardware tools to build a
house, if they're building ahouse, they'll go to that
server.
But if there's another one forreading instructions or manuals,

(27:00):
possibly, you may have a serverfor that.
Another thing that you hadmentioned, which I hear a lot of
with AI, is tokens.
And you mentioned that you canburn up a lot of tokens.
What is a token in the contextof MCP AI instructions?
Is it money?
Effective is it's a gold token.

SPEAKER_01 (27:22):
Right?
I know.
It brings us back to the arcadedays, right?
Burning quarters.

SPEAKER_03 (27:27):
Um the um we put them on the pinball machine when
it's your for your turn.
You have to remember do youremember that stack in the
quarters?

SPEAKER_02 (27:34):
Yeah, I got next.
I got next.

SPEAKER_03 (27:38):
Just I don't want to digress.
I know Jeremy, you want to jumpinto but just think about what
like someone who's youngertoday.
Well, we're talking aboutputting quarters on a pinball
machine or even a regular arcadegame to get next.
And they grew up with an entirearcade in their hand.
It's pretty incredible.

SPEAKER_01 (27:59):
It is incredible.
100%.
So the uh the token thing, justto kind of speak to that, is um
just like with water and power,you kind of like meter your
usage, and like uh for yourhome, uh you buy a washer or a
dishwasher that maybe uses watermore efficiently.
Um it's the same sort of idea.

(28:21):
A good set of instructions ismore compact and it is very much
more precise around what shouldthe agent be processing.
And the amount of informationthat goes from your local
discussion point to these LLMservers, that that's the
underlying tech behind theagents, um the uh the amount of

(28:44):
information that's going, that'suh there's no weight to it,
there's no data volumeprecisely, it's it's measured in
tokens.
So, how heavy is thatinformation stack that you're
sending to the AI platform,GitHub or Co uh Anthropic or all
these other things, how heavy isthat data stack?

(29:04):
And the amount of data that's inthat pile is the amount of
effort that the LLM has to gothrough all of that stuff.
And the measurement that the LLMuses is tokens.
And so that's why people talkabout that.
Um, and so uh oftentimes ifyou're using GitHub uh chat and
doing the copilot things inVisual Studio Code using GitHub,

(29:26):
um you might have premiumrequests which are using tokens
talking to companies that aren'tincluded uh up to a certain
point.
Um and tokens are how they meterthat, the same way you would
meter electricity and water andthose sort of things.
So a unit of measure.

SPEAKER_03 (29:46):
Okay, great.
And then so it is, it's a meterthat goes through it.
Okay.
So now we talked aboutinstruction sets that someone
can write, and I'm sure there'sa format for that.
We have agents, we have an MCPserver, and we have Jeremy's
journey through this as well.

SPEAKER_01 (30:03):
Because uh well, one of my favorite ways of
developing instructions, becauseI'm not a I again, I I've been
doing this less than six months.
Uh one of my favorite ways to douh instruction tuning is to get
the agent to help me write theinstruction for the next guy.
So what I'll do is I'll set uhyou know, oftentimes, especially

(30:23):
in July when I was new at this,I would say, okay, I've got a
task in front of me.
I want the agent, I want you todo this task.
It did not give me the results Iexpected.
So what I will do is I'll makethe results I expected and then
go, explain to me, like youwould explain to the next agent,
what the difference between whatI now have that I've done versus

(30:47):
what you did.
How would I the next agent toknow to do it my way?
So you can ask the agent, how doI improve it for the next guy?

SPEAKER_04 (30:56):
Okay.

SPEAKER_03 (30:57):
Wow.
That see, this is ingenious.
I'm thinking of all the thingsI'll have all these agents do
for me right now.
Like just do a code review.
You know, do a code reviewbefore you set it up to GitHub
Copilot, do a code review, andalso do some other interesting
things.

SPEAKER_01 (31:15):
So code review and documentation are becoming great
opportunities because they'reoften the things that take a
fair amount of time, but wedon't always necessarily get a
lot of time to do them.

SPEAKER_03 (31:28):
Yes.
And depending upon uh again withany code review, I don't care
what anybody says, if you have4,000 files to go through on a
code review, your level ofdetail is going to be much less
than if you have to go throughone file change.
As a human, right?
I think it's it's the level thatyou can the level of effort and
time that you go through ischallenging, the larger the code

(31:51):
review is for the pull requestthat you made we're reviewing.
So we we talked about all theseMCP servers, and I definitely
want to get into the Nubimency.
Nubimency.
Nubanincy.

SPEAKER_01 (32:09):
I'll make it easy for you.
Newbie, as in like new guy.
I'm a newbie.
Okay.
Um, which is one of the Latinpermutations for cloud and mancy
for magic.
So it's cloud magic.

SPEAKER_03 (32:21):
I was going to ask you where it came from as we
talked about the setup of yourproject and what the project is,
but it's newbie, I got that.
That's the old days of noobs andnewbies and all the terms that
we had for somebody that wasnew, and Mancy for magic.
I knew you would have to havesomething like that in there.
Absolutely just because of uhthe interest that you have

(32:42):
outside of AL coding.
Okay, so now we have these MCPservers, we have these agents,
we have these instructions.
How do we get them?
How do we create one?
Right?
So now we're saying we haveinstructions for an agent, we
have this MCP toolbox.
Where do we put it?
How do we create it?
How do we do it?

(33:04):
Right?
I mean, if you think about this,it's how do we put all that
together?

SPEAKER_01 (33:09):
Well, I mean, the short answer to that, my
favorite answer is when youdon't know, ask the agent.
It knows.
So when it when it gives you theinstructions and you go, that's
really cool, but where the heckdo I put this for next time?

SPEAKER_02 (33:22):
It'll tell you.
See, that's beautiful.

SPEAKER_03 (33:25):
That's the whole you know, I I gave someone an answer
the other day, and you just gavethe answer back to me because
somebody asked me a question andI gave them L M G T F Y, right?
Let me Google that for you.
So it's the same thing, butdifferent.
Now we're getting into askingthe agent how to do something.

SPEAKER_01 (33:44):
Well, and one of the things is true, is like a lot of
us in the business central spaceuse VS Code, and if we are using
agentic coding or even AI at allin our development experience,
it's typically GitHub co-pilotchat because it's easy.
It's it's built right into theOS or Visual Studio Code
environment.
We don't think about it.
But the reality is there aremultiple different providers out

(34:07):
there that provide differentcoding tools, um, and some
people encourage people toexplore which one works best for
them.
And each of those tool sets willstore their instructions in
different places.
So it's valuable to say, it'svalid to say, you know,
depending on the tools you'reusing, the defaults may be
somewhere else.
In our organization where we areusing GitHub Copilot, um,

(34:30):
there's a standard name you cangive the file in the standard
folder, and it will know toalways look at that instruction
set.
So like it's dot github slashcopilot dash instructions dot
md.
It's just markdown files, soit's just English.
Um and anything you put inthere, the copilot will
automatically pick up as part ofjust existing.

(34:52):
So it's pretty nice.

SPEAKER_03 (34:55):
I have to pause because now I'm thinking about
all this.
So that anybody that has aproject, they can create a
special folder and you can askthe agent if you don't know what
it is, if Jeremy just said it,and you can put your own
instructions in there, and thenthe GitHub co-pilot chat that's
within VS Code environment, willuse those instructions every

(35:17):
time.
To every time.

SPEAKER_01 (35:20):
So it's like a welcome kit that as soon as that
excited employee shows up, it'slike, ah good, welcome inside.
Here, read this first, then comesee me.

SPEAKER_03 (35:30):
I like that.
And then can you specify whichinstruction set to use so you
can have multiple functions, ordo you have one big function?

SPEAKER_01 (35:41):
What we ended up doing, and what I've referenced
um that I I referenced in thatblog series the after the first
week or so um of using theinstruction set that I baked up,
it was too big.
Um the I built a very nice,lovely book, but the LLM only
looks at so much.

(36:01):
It's not gonna read every singlepage carefully and memorize
everything about a wholedictionary's worth of stuff.
That's too much.
So what it was doing was we werefinding that depending on what
the developer in our team wasdoing at the time, it sometimes
was fantastic or sometimes wouldbe way off course.

(36:22):
And what I ended up realizingthat as developers, we do
context switching.
I might be in spec reading modeand analy uh analyzing mode to
look at what someone's asking meto do, and then I'm trying to
turn that into an architectureand a plan, and then I'm trying
to code it, and then I'm tryingto think about performance, and

(36:45):
then I'm trying to think abouttelemetry, and then I'm trying
to think about testability, andthen I'm thinking about
documentation, and then I'mthinking about extensibility.
All of those are differentcontexts.
We as people slide between themwithout thinking, they're just
so very comfortable to us.
There's no friction of movingbetween those things for most

(37:06):
people.
I won't say everyone, but mostpeople, it's very easy to slide
between those different roles.
Some people are better at thosedifferent parts than others, and
maybe they'll say, you do thearchitecture, I'll write the
code.
It's fine.
What we ended up doing was wegave special instruction sets
that defined each of thosecoding contexts to say, you are,

(37:28):
in this context, only anarchitect.
Don't write me any code, justwrite me a plan of what the
application should be structuredlike and what it should do based
on the specification that we'vegot.
I don't want you to do anycoding work for me.
I just want a plan so that I canvalidate the plan before it goes
into code production.

(37:49):
Um and what we ended up doingwith that, uh, just to make it
very easy to like mentally thinkthrough that code context
switching, uh, it ends upworking really easy to give each
of them names.
So it could be like, I want totalk to Sam about code, I want
to talk to Roger aboutreviewing.

SPEAKER_03 (38:04):
Um, if you don't know about the project, check
out the project and read the thetrail that Jeremy left as he's
gone through this projectbecause that's what it actually
came up to was a bunch of youspoiled it, Chris, a bunch of

(38:27):
names for all of these agents.
It's it's laughable.
It's it's almost laughable.
And what are some of the names?
I do have questions about this,but what are some of the names
that you gave them since Chrisuh mentioned it?

SPEAKER_01 (38:37):
Let's see, Alex Architecture, uh Sam Coder,
Roger Reviewer, Quinn Tester,Jordan Debugger.
Uh we've got a special uh onespecifically to help like people
who are new to BC developmentcalled Maya Mentor.
Um and she, for example, we'vegot new people who joined us
since the summer, and it'sphenomenal because our juniors

(39:00):
have been able to just pop openprojects and then ask this Maya
role, can you explain thisproject to me?
What sh where should I start?
And it understands your role isjust to explain to this new guy
who's never seen AL code before,where do I even start to read
it?
So um we found that it wasreally easy for people to
remember like the names muchmore than I need to talk to the

(39:22):
review guy.

SPEAKER_00 (39:24):
Yeah, you it's just funny because I I did look at
your website and I was like, arethese real people?
Like the project leaders,Mistress, Scopewell, uh, supply
chain, functional consult, pickforward, been right.
I'm like, are these real people?
And I'm like, I had to pause fora minute, and I'm like, okay,
they're not real people.
This is fascinating.
But you give a name, right?
Like it's like you fall in lovewith them.

SPEAKER_03 (39:46):
But it's almost like you build your own team.
And that's the key with this, isI'm following what everybody is
doing with this, uh, within thecommunity and outside the
community, because you mentionedREG and all those other things.
I find myself in the in the whenI'm driving somewhere, I listen
to YouTube now, trying to get upwith all these new terms, and I
always end up forgetting most ofthem because they always have
like a new one, it seems.

(40:06):
So it's like building your ownteam.
So you can have multipleinstruction files in your repo
in this in the context of whatwe're talking about, and those
different instruction files havesomething in that say what they
do.
So when you're talking with youragent, you say you are uh Maya
mentor.

(40:28):
Here's my question, basically,or you are you know Brad, the
best coder, write this functionprocedure for me, right?
And then it will do that foryou.

SPEAKER_00 (40:40):
I I love the reference to the like you said,
you know.

SPEAKER_03 (40:42):
I have to just stop this conversation right now
because I'm gonna go set all ofthis up.

SPEAKER_00 (40:46):
I love the magical references though, like the
fantasy world that you've built.
Yeah.
And as a uh when I was growingup as a gamer, I I love the
references to that because it'slike, oh yeah, I can it's easy
for me to remember thosecharacters, right?
When you're building this thing.
This is fascinating.

SPEAKER_01 (41:05):
So let's wrap the MCP topic up with one last thing
that is on this like journey andeverything like that, and then
I'll talk to the the newbiemancy project and what the heck
is that.
And uh because they're notthey're they're related, but
they're not record dependent oneach other.
Um the uh all of the stuff thatI've learned with like these

(41:26):
instructions and everything.
I also learned how to interpretlike the project AL
guidelines.dev.
All of that wisdom is broughtinto the MCP server that I
released called BC CodeIntelligence.
And all of the AL guidelines arebrought in.
So that MCP, like I mentioned,the toolbox, that includes in it

(41:46):
all of the topics that are on ALguidelines.
And so what that MCP for BC codeintelligence is for, again, open
source, anyone can install it.
If you go to the GitHub projectfor it, I'm sure there'll be
lots of lovely links in thecomments because you guys do a
great job of linking to things.
Um, the GitHub project for theBC code intelligence, there's

(42:08):
literally a blue button thatsays install this into my Visual
Studio Code.
Um click, that's it.

SPEAKER_03 (42:16):
No, I laughed because when I saw the project,
I saw the project, I was talkingwith Jeremy because I'm trying
to figure out, okay, with thisMCP server, where does it run?
What do I do and stuff?
I was talking to him, and hejust wrote back to me, just hit
the blue button.
That's what it's like the likeit goes back to what he was
talking about.
Like the agent would do afollow, and I'm like, okay, what
do I do?
How do I do it?
He goes, here's the link, justclick the blue button, and it

(42:37):
will do it.

SPEAKER_01 (42:39):
Yeah.
So uh for the technicallyinclined, what that does is in
Visual Studio Code, there's anMCP.json for your Visual Studio
Code environment, and that listsall of the MCP servers and how
to install those.
And it does require a little bitof uh prerequisites.
So for example, uh many MCPservers are actually JavaScript

(43:00):
or TypeScript type uh littleecosystems.
Um, so it does typically requirethat you have the node.js uh
libraries installed on yourmachine.
Uh those are usually included inthe README's with people.
So you if you don't have it forsome reason, you can get it up
and going.
So the uh BC Code IntelligenceMCP contains all of those

(43:20):
experts that we were justtalking about.
It contains all the ALguidelines.
Um, and some of the roles uhhave expanded a little bit over
the past couple months.
Like I've got one that'sspecifically for uh code
archaeology, because oftentimeswe take over projects that we
don't know anything about, andit's like, well, I need a role
to help me understand this app.

(43:41):
What's the goal of this app?
Um, and so what you can just dois when you hit that install
button and then you got itonline, you go, Who's on the BC
specialist team?
And you know, it'll give you thelist of all the names, and
you'll be like, Great, let's uh,you know, let's have Logan, the
archaeology guy, jump in herewith me and help me understand
this app, or have Maya help meexplain why is runtime.8 uh 8.0,

(44:06):
why is that set here?
Can you explain what the rightversion of that would be?
So um, and there's a bunch ofdifferent prompts that are built
into that, which are pre-builtinstructions.
So when you've got that MCPinstalled in your Visual Studio
Code environment, you canactually just hit slash in your
GitHub chat, as long as you'rein agent mode.
And there'll be this list of MCPBC code Intel prompts that will

(44:30):
pre-fill in a whole bunch ofstuff that you can just hit
send.
So, like for example, I want youto help me understand this app.
I want you to help me upgradethis app to version 27 and fix
all those number seriesdependencies that just started
breaking.
You know, all of these sort ofthings that are just pre-built
as prompts that you can just hitslash, pick one, and just hit

(44:52):
go.

SPEAKER_03 (44:53):
I'm just not saying anything because I had to fix a
number series yesterday.
And I had some help.
Don't tell anybody.
Uh so you're mentioning theseMCP servers.
So you hit the blue button andinstall.
In the con in this scenario,where does that run?
Because I know there's MCPservers for Learn that you can

(45:13):
connect to.
There's MCP servers for uh someother MCP servers that others
have created.
We're talking about theseservers, MCP server.
We know what it is.
Where does it run when you clickinstall?

SPEAKER_01 (45:26):
Just like the wonderful world of USB-C is
absolutely universal foreverything, right?
Uh MCPs are universal foreverything, right?
Um, there's two main types ofMCP servers.
Uh, ones that are networked andyou can use them in larger
projects and on ecosystems,multiple people can call it.

(45:46):
Um, those are typically thethings like the GitHub MCP or
the Microsoft Docs MCP.
There's actually a serverrunning on GitHub or Microsoft
site that you're callingdirectly, and that server
ecosystem is out there.
Many of the ones that you caninstall and run for yourself on
your development environment areactually another type where

(46:06):
they're a little micro serverthat is actually running on your
machine when you start that MCPuh up by having the chat talk to
it.
Um and that's what the case witha BC code intelligence.
It like many other uh MCPs, itactually starts a little node
process on your machine to runthe JavaScript server.

(46:28):
So while that's up and running,there's actually a little tiny
server running on your machinethat is listening for those
calls from the uh copilot agentuh that will call into that
local server.
So it's on your local machine atthe time.

SPEAKER_03 (46:43):
Okay.
And then is it only running whenyou have Visual Studio Code open
or is it running all the time?
Okay.
So I just wanted to getclarification for those.
I know that many have questionson that on when they're uh
running and such.
So thank you.

SPEAKER_01 (46:59):
And typically a lot of times the MCP servers will
only start once they're called.
Um and uh in Visual Studio Code,you can hit F1 to open the
command palette and say uh MCP,and you'll get a bunch of MCP
options.
And one of the options is listservers.
That'll show you all the serversyou have installed, and in that
little window, it'll show youwhich ones are running and which

(47:22):
ones are stopped.
And that way you can also selecteach one and say, I want to
start or stop this one.
So you can be in very uh carefulcontrol of those as you want to.

SPEAKER_03 (47:34):
Oh, good.
So you can keep are theyresource intensive?
I know we talked about tokens,data, and load.
Are these MCB servers?
Typically not.
Okay.

SPEAKER_01 (47:43):
Typically, they're often just uh many of them are
just reflectors.
Uh for example, talking to AzureDevOps.
It's uh a wrapper that's callingthe uh APIs at Azure DevOps.
It just is a nice way of going.
I'll take what the LLM wastelling me in natural English
and I'll turn that into an HTTPcall to Azure DevOps, and when

(48:07):
it responds, I'll hand thatresponse back to the uh agent,
and the agent knows what to dowith it from there.
So many of them are verylightweight.

SPEAKER_03 (48:15):
Excellent.
Excellent.
Now I know everyone has a betterunderstanding of this, and
everyone's going to go click theblue button and install it
because that's what I wanted todo when I first started reading
this.
So now let's jump into thenewbimancy project.
Right?
It's you had me you you hadmentioned it's it's similar, but

(48:35):
related.

SPEAKER_01 (48:36):
Um so uh oftentimes I uh I do a lot of presentation
and training and writing aroundBusiness Central.
And one of the challenges uhwith educating on BC, uh DevOps
and all these relatedtechnologies, you can't show
your work because you're doingclient work, you're doing

(48:57):
internal IP work, you're doingthings that you cannot share.
And so oftentimes when you'regoing to teach, uh you're
wanting to share differentthings, you have to duplicate
all that effort in ashare-friendly way because you
can't leak IP or GDPR data andall that sort of thing.
Um I want to create more books.

(49:18):
I would like to show off morethings about uh how DevOps and
GitHub and all these differentpieces can work, but I can't do
that with my company's data.
I can't do that with my clients'data, I can't do that with our
products.
So I went, I need to be able todo this more regularly.
I I speak at conferences everyyear.
Uh every year I'm having tostart over and coming up with

(49:41):
like a new like fictionalbusiness and a fictional story
behind this.
And I'm kind of a nutter.
I like to cross-pollinate fromsome of the different things
that I enjoy in life.
Um, and one of the things Ienjoy a very great deal is
tabletop gaming, um,particularly role-playing games,
DD, and that sort of thing.

(50:01):
Uh, I've DM'd for a long time.
And so I I enjoyed the idea ofwhat if I took both of those,
uh, this thing that I love andthis thing that I keep
struggling to fulfill.
What if I crossbreed those?
What if I do something sillywith those things together?

(50:21):
And so Newbie Mancy is a projectwhere I'm building a fictional
business central partner that isfictionally helping a group of
retired heroes.
These retired heroes help savethe world, and now they've
ridden off into the sunset, buthappily ever after means what?

(50:44):
So each of these five retiredheroes has their project that
they've always wanted to do, andnow that they've helped save the
world and they've found theirfortune and all that sort of
thing, what do you do with yourlife now?
And so they're trying to runtheir respective businesses, but
you know, running a businessstill takes a lot of work.

(51:08):
So the idea behind uh the newbiemancy project is I'm going to
fictionally treat those five uhchallenging customers as a
family of companies that we'reproducing um some app source
apps for, some uh PTEs for uhdemo data, testability, all of

(51:30):
that stuff.
And because this is all 100%fictional, I can do this in uh
an open way.
So, for example, I'm buildingDevOps up with only public
projects for this area.
So that way, if you're trying tofigure out, okay, how is this
guy doing pipelines in AzureDevOps?

(51:51):
Well, all of the source filesare there.
You can just go in and look atthem, you can see all of the
flow and everything.
So it it gives me a platform bywhich I can share more about how
to do all these differentthings.

SPEAKER_03 (52:08):
I think it's amazing.
And it's funny because when Ifirst looked at the website when
you first started promoting it,I start with I saw a bunch of
people on there.
And I'm like, oh, is Jeremydoing this or are all these
other people doing it?

SPEAKER_01 (52:19):
So it was uh You know, I'd have to look.
I it's it's something in theneighborhood of like 30 or 40
fictional business centralpractice people who are gonna be
implementing all this.
And as I'm fleshing out thestory this fall for the heroes
and their businesses, no man isuh an island.
So each of those uh businessesis gonna have their own little

(52:42):
cast of characters for each ofthose businesses as well,
because why not?

SPEAKER_03 (52:47):
It's it's a fun, it's a fun project to follow,
it's a fun site to follow.
And I also enjoyed the blogarticles following the you know,
from AI Skeptic to uh you know,in 90 days was your you know, I
think your title for that, oryou had a few of them on there.
I can't really call them all, Iapologize, but I remember the
concepts.
Uh definitely a great follow.

(53:09):
It's definitely entertaining.
And as Chris had mentionedbefore, some of the names are
pretty unique for even on thewebsite now as looking uh as you
expand it, because uh you knowyou have been updating and added
content with it.

SPEAKER_01 (53:21):
So it's um yeah, now it's not a small journey on that
I'm taking uh with some of thestuff that I'm doing.
I've mapped out uh this pastweekend, I was working on it a
little bit.
Uh 38 total apps are gonna be inthe constellation of all the
different pieces to likedemonstrate all of the different

(53:43):
parts.
So, like, you know, the appsource app equivalent for this
uh family of companies, uh a keyfoundation layer, and three apps
that are on top of it torepresent good, you know,
multi-app infrastructure, andthen individual PTEs that sit on
top of each of those partsdepending on the business type.
Um, so I think I've got all 38apps all now scaffolded with all

(54:07):
the basics, and I'm gonna beworking on uh pipelines and um
the next step up for the biglift on this is I'm gonna be
starting to produce all the demodata that will go with this.
Just like uh Kentoso has, youknow, the ability to install the
different modules.
Uh this solution is gonna bewhich of the heroes do you want

(54:27):
to install into a testenvironment to be able to use
their data as test data?
Go for it.

SPEAKER_03 (54:36):
That's impressive.
And I I sit back and I listen toall this, and I'm like, how do
you have the time to do all ofthis?
To to work, to speak, to tolearn, and then also to put
together a basically afictional, I'm calling it a
fictional universe.
I go back because I also grew upplaying tabletop games, you
know, Dungeons and Dragons, uh,magic, you know, the card game,

(54:58):
you know, all of these.
I wish life was still like thatback then, to be honest with
you, because that was so muchfun doing all those.
But it's almost like you'rebuilding a whole world or a
whole universe.
Absolutely.
Yep, how you're doing this?

SPEAKER_01 (55:12):
There's you know, there is a fictional world,
there's fictional countries,there's a whole bunch of stuff.
I've got a whole knowledge basethat for my agents who are
helping me build some of thesethings up, because yes, I'm
using agents to help me buildall these things.
I've got a knowledge repo aboutall of the different countries
and like what are the cities ineach of them.

(55:33):
So that way when we'regenerating the demo data,
there's there's a basis to workfrom.

SPEAKER_03 (55:40):
You know what I think would be interesting from
this?
After you build this out and youhave all these instructions for
the countries and the companiesand the people, have it write a
book.
Have it write a book about thecountries and the people and
their lies, just to see whatlike have an agent write the
book and say, here's a bookthat's completely AI generated

(56:04):
based upon this newbie mancyproject.
I think that would be aninteresting result if you're
building it up this much.

SPEAKER_01 (56:11):
We we strike there into an interesting heart of an
ethical conundrum.
We talk about this a lot in myfamily because my wife is from
an art background, and thechallenge of bringing AI into
creative spaces is that there isa reasonable and angry sort of
pushback from people that ifyou're using AI tools in such a

(56:35):
way to do creative works that itwould be better to utilize a
creative person for, um, thatthat's not uh that that doesn't
feel right.
It gets a lot of pushback.
So for example, I don't use AIgenerative tools for images.
I hire graphic artists.
I don't uh typically hire AIeditors because I actually

(56:59):
really rely on um some of thedifferent people that I work
with to help do a lot of theediting work.
So it's an interesting balancefor me of I don't want to use
I'm okay with using AI to makethe things I would do happen
faster, but if I'm using AI toreplace a creative person, and

(57:19):
that includes, you know, myvoice of storytelling, I
actually would find that thatwould be really challenging for
me to accept.
And there are uh some fairlyfamous incidents uh at
conferences of gamingconventions and things like that
where there was a discovery thatpeople were using AI tools to
make the art that they wereselling at the convention, and

(57:40):
they were thrown out.
Yes.
So there's some very bigpushback.

SPEAKER_03 (57:46):
Yes, I understand.
No, I understand the pushback,and I understand that whole
dilemma as well.
Yeah.
Um because it's uh it's it'sit's a fine line.

SPEAKER_01 (57:58):
Um it is the it's challenging to be pro-AI.
I see the value, I experiencethe benefit.
Uh there's a lot of uh theanswer to your question of how
do you have time for so many ofthese things is because of AI.
There's a bunch of stuff thatI'm multitasking between, and
these tools are helping me dofaster.

(58:18):
Um I genuinely am getting muchmore done than I've ever gotten
done before in the past fewmonths.
And I described it to someoneas, you know how autistic folks
talk about, you know, the aroundtheir friends they can unmask.
They can uh they can be theirtrue selves, they don't have to
like be an interface to the restof the world.

(58:39):
Uh I describe my working withagentic development as one of
the first times in my life whereI'm able to produce code or
content that is code adjacentthat it's as fast as I can come
up with the ideas.
It feels like for the first timeI'm going the speed that I can

(59:02):
be at in my head.
And that's a big deal.

SPEAKER_03 (59:06):
No, that's I I can see it's it's it goes back to
what you would just sort of talkabout.
There's a fine line between thepractical use of AI and how far
you take the use of AI basedupon uh what you're doing with
it.

SPEAKER_01 (59:22):
But that and you'll see like traces of that in some
of the things that I do.
Like I'm still, as more oftenthan not on social media, you'll
see me post like hand-drawingthings because there's an autent
authenticity to that.
That no, that's physicallysomething I just made.
And um for the folks who are AIskeptics or against AI, there's

(59:48):
a kind of a respect andrecognition of like you're you
are not replacing a human withthis work, and you're trying to
be careful.
So, uh, for example, there's agreat business insider about
like the uh article about thedangerous challenge of how much
AI is impacting climate change,which is something I care a lot
about.

(01:00:08):
So for me, building BC codeintelligence isn't to try to get
to people to get people to useAI lots and lots more, but I
know that that's going to happenanyways.
So if people are using it in away to get better results more
correctly, there's less wastefuluse of AI.

(01:00:29):
So for me, by building thesetools, it's uh it's about making
it uh mm as ethical as I canmake it if we're gonna do it
anyways.

SPEAKER_03 (01:00:40):
Yes.
I understand I mean I'm saying Iunderstand and I see and again
you hit it for anybody that maythink or be a skeptic or may
think that what we're discussingabout ethics and AI contradict
each other, that is a great wayto put it that it's going to
happen anyway, so why not do itin an efficient and ethical way

(01:01:03):
to minimize the impact as muchas possible.
Right?
Or the negative impact or theside effects.
You're talking about climate andthis this side effects or impact
of AI across the board.
So um I think uh we cover thatthat it makes you think you know
we we we we we talk about allthe great things of AI but it

(01:01:24):
also does make you think aboutthe AI's impact on uh the the
world not only the physicalworld but also the people in the
world and even um mentally withindividuals now because I've uh
had lots of conversations withindividuals that struggle with
AI and just what it's doing tothem mentally if you take a look
at it you know it's doing a lotof great things but it's also

(01:01:47):
creating a lot of anxiety andangst in many individuals as
well too well I'm a dad and youknow I've got a kiddo in high
school and you know one of hisperspectives as a child growing
up in the AI era is he's takenthe ethical stance of I refuse
to use AI while I'm in school atall so that I'm never in a

(01:02:08):
discussion with anyone about isthis work genuinely yours?

SPEAKER_01 (01:02:13):
Because he wants to have that purity of I can always
say this is something I'm notdoing at all.
That I'm I'm just against usingit.
And I agree wholeheartedly withhis decision to do that.
I mean I would support him if hechose otherwise but he wants to
in these formative years makesure that he's developing those

(01:02:34):
skills himself and that's fair.

SPEAKER_00 (01:02:37):
Yeah that that is that is a good point because I
have two high school kids rightnow that um you know they don't
use like same similar to youJeremy they refuse to use AI.
They want to know it's like uhit's like the analogy of like
you know you can fly a plane andautopilot and you'll be fine but
if it doesn't work you need tobe able to learn how to do it

(01:02:59):
manually.
So you know they they don't useit.
Um the only time they've everused it was they need to
validate something and ratherthan Google they just had to
copilot or chat GPT and that wasit.
That's the only time they haveto use it.

SPEAKER_01 (01:03:13):
One of my favorite uses of it that is phenomenal uh
that is also you know in thatspace of it's about making the
person better um use AI andcopilot tools to challenge what
you're doing or to clarify.
So for example if I'm doingpresentations I will for example
maybe run it through a copilotto say you're a uh consultant in

(01:03:37):
the BC industry I'm talkingabout this topic this is the
context here's the PowerPointdeck what questions would you
have that I have failed toanswer this is my premise have I
really kind of hit the points ofwhat I'm explaining so I do
often use validation toolsaround Copilot to also help me

(01:04:00):
make me a better communicatorand make me think of the gaps
that I'm missing.
And BC code intelligence as anMCP is that same sort of idea
that it's it's about trying toclose those gaps.

SPEAKER_03 (01:04:12):
What are the things I forget what are the things I
don't think of can we closethose up yeah great great great
uses for those it is it is andthis has been a great discussion
uh uh newbie mancy I learned howto say properly I think everyone
should follow that project thatthe team and I think also um the

(01:04:35):
uh code uh mcp that you havewith all those agents is is also
something that everyone shoulduse and also help contribute to
uh you had mentioned it's opensource so if you have uh
contributions uh we can make ita little more efficient for
everybody and I I have to createa Brad one i i think I'll have
to come up with some sort ofagent that's a Brad one a Brad

(01:04:57):
the I have to follow the namingthe BC code intelligence MCP
also supports um layering tech.

SPEAKER_01 (01:05:04):
I've talked to a couple of early adopters if you
want to add stuff that'sspecific to your practice or
even specific to your projectthere's a guide in the wiki for
that project that says here'show I add stuff specifically for
my organization and here's how Iadd stuff specific to my project
because there might be goodgeneral guidelines that you do

(01:05:25):
something slightly different andthen you do something even
slightly more different in theproject level.
And that includes adding morespecialists.
So if you want to experimentwith it and add a new specialist
type called Brad the superheroyou absolutely can do that and
the MCP server is built tosupport loading that as an extra
layer much in the same way thatBC does extensions, the MCP

(01:05:48):
server does layers so that youcan just add stuff in.

SPEAKER_03 (01:05:52):
See you thought of everything when you put that
together the layering is greatbecause like you had mentioned
if somebody has any someindividual specific rules that
are unique to them that they maynot want to share they still can
use the project.
Oh that's the wicked on therabbit hole of object naming and
object numbering and affixes andnamespace.

(01:06:15):
We could spend all day talkingabout those.

SPEAKER_01 (01:06:17):
Well Jeremy thank you again for taking the time to
speak with us this has beenextremely informative and I
appreciate your time and thankyou for sharing with us all that
you have done and all thatyou're doing uh in this episode
but as well as what you're doingfor the with the community as
far as your speakingengagements, your book writing
and now uh offering theseprojects uh to individuals

(01:06:40):
because I know I learned a lotuh following yeah what you had
put together as well too Ireally would have to get uh on
those characters do you have areyou gonna put images of them
like uh like an employee uhcharacter is a phenomenal artist
I know in town who loves to doRPG character art so I'm I'm

(01:07:00):
hoping that as a uh Christmasgift to myself I'm actually
gonna buy a portfolio from himof I would love to have
character art for each of theseuh wonderful little characters
and we'll flesh out theirbackstories over the upcoming
years.
Yeah that'd be awesome.

SPEAKER_03 (01:07:16):
Then you can make a game a tabletop game with these
characters.
See that's what this I wantsomething like this.
I'm seeing this whole coolmythical world because I like
those types of worlds too rightit's I can just see this whole
mythical world coming togetherfrom this project.
I'll tease for the next time I'mjoining you guys on the show in

(01:07:36):
the uh probably after the newyear of uh I I've got some l
groundwork laid for uh BCcharacter sheets we'll have to
we'll we'll schedule that upsoon to get you back on and also
see you know I want to followthis project and see where it it
grow it goes and and how itgrows um it's um I I think with

(01:08:00):
technology as a change isadvanced I think you know give
it tomorrow and there'll be newtechnology that we can utilize
with this.
But if anyone would like to getin contact with you to learn
more about Nuba Mancy, learnmore about your uh MCP server
knowledge experience uh what'sthe best way to get in contact
with you?

SPEAKER_01 (01:08:19):
Uh these days the most often place I'm I'm
chatting with people tends to beLinkedIn.
I am on Blue Sky so you can findme there as well um but LinkedIn
has been the most active for melately.
That's where the Microsoftpeople hang out so I'll join
them there.

SPEAKER_03 (01:08:36):
Great great thank you we'll have links in the show
notes and then also we do have aguest profile so your guest
profile will be attached to thisas well with some of the
additional information forothers to get in contact with
you.
Thank you very much I'm going togo sit down and start thinking
of Brad the superhero and I'mgoing to work on layering it to

(01:08:56):
see what Brad can do for mycoding projects.
Sounds good thank you talk toyou soon take care thank you
Chris for your time for anotherepisode of in the dynamics
corner chair and thank you toour guests for participating
thank you Brad for your time itis a wonderful episode of

(01:09:17):
Dynamics Corner Chair.

SPEAKER_00 (01:09:19):
I would also like to thank our guests for joining us
thank you for all of ourlisteners tuning in as well you
can find Brad at developerlifecom that is D V L P R L I F E
dot com and you can interactwith them via Twitter D V L P R
L I F E you can also find me atmattalino and my Twitter handle

(01:09:51):
is Mattalino16 and see you cansee those links down below in
the show notes.
Again thank you everyone thankyou and take care of