전사 🔗
WEBVTT
00:00:00.473 --> 00:00:03.533
What you think when I say best practices?
00:00:03.693 --> 00:00:34.823
How dare you it means somebody somewhere did thinking for you[UM] No thinking required. that's that's entire job that phrase does. So it's equivalent of secondhand thinking Oh boy. Welcome back arguing Agilent. I'm host product manager Brian Orlando. this my cohost, enterprise leader and the boss best practice, Mr. Om Patel I heard I get promoted every week, here! That's right. That's right.
00:00:34.817 --> 00:00:35.279
I like think myself keep going.
00:00:35.878 --> 00:01:02.595
the pragmatist of practices. But hey, the boss of nobody the boss of bad habits. That was guilty charged probably. So today we're be arguing about best practices. Are they really the best or just secondhand thinking Before we start, you like our channel, please don't forget like subscribe and [UH] put in your comments in the comments section let's discuss best practices in three parts.
00:01:03.015 --> 00:01:42.519
First, what phrase actually does when it's used like meeting and why shows up mostly for things that it explicitly can't handle second, why best practices never work. we'll use example that if you're watching [UH] our podcast, you definitely have story about it. And third, we'll discuss the memo that your CEO probably already sent you [UH] work for, know, some big company maybe rhymes with Shopify. that's not rhyming. That's just saying name the company, Brian. do you anyway. Because, because the modern equivalent this is[UH] ask the first.
00:01:43.019 --> 00:01:44.918
Yeah. Yeah. It's same play we think.
00:01:45.218 --> 00:01:47.618
But we'll see. We'll see what you think the end it.
00:01:48.180 --> 00:02:00.513
So Brian, are you claiming the phrase best practice is receipt for someone else's thinking Yes. And yes. And it's really the podcast We've got that in there[UH] welcome Julie.
00:02:01.054 --> 00:02:22.774
Oh, claiming that there may been things cut out, but can tell That's not only I you the term best practice does not get uttered when the team is designing something new or something novel it gets [UH] gets uttered when decision is defending and works because nobody room has reconstruct the reasoning behind best practice.
00:02:24.033 --> 00:02:33.700
the issue phrase that shows up on big irreversible context dependent decisions you know, the exact ones that shouldn't be used for.
00:02:34.375 --> 00:02:38.281
know we're talk about this a later section so I'm not to deep dove here.
00:02:38.480 --> 00:03:05.134
But if you're working on something new, you know, something nouveau[UH] novel, whatever your objective oftentimes teams use best practices as a way to moderate the risk that they're taking on because they're saying this isn't groundbreaking. This is something that's been done. We're going adopt these practices that worked over there somewhere [UH] over yonder beyond horizon And know, so that's think true to certain extent that people do use that.
00:03:05.194 --> 00:03:17.115
I've seen that specifically in the pharmaceutical industry where people are using models that have worked before saying well it's worked before so this isn't new again. Risk reduction strategy.
00:03:17.661 --> 00:03:26.181
So for this section I have a citation two and I thought would fun to share this one as a Lisann Lisann Bainbridge Ironies automation.
00:03:26.381 --> 00:03:34.813
found this Lisann Bainbridge ironies of automation from 1983 maybe 1982, definitely 1983 I'm read section.
00:03:35.514 --> 00:03:36.675
because it's a boy.
00:03:36.835 --> 00:04:06.900
It's a zinger. OK, here we go. The important ironies the classic approach to automation. It's also that I'm pretty sure this is a pretty sure someone UK wrote this paper anyway. It strikes me as very UK the important ironies of the classic approach to automation lie the expectations of the system designers and in nature tasks left for human operators to carry out. The designer's view of the human operator may be that the operator is unreliable and inefficient. So should be eliminated from the system.
00:04:07.379 --> 00:05:11.646
There are two ironies in this attitude. One is that the designer errors can be major source of operating problems [UH] that's such good one. The second irony is that designer who tries to eliminate the operator still leaves the operator to do tasks which the designer cannot think how to automate the reason found that paper is because it's referenced by this Lee et all Microsoft Cambridge paper from where 2025 they surveyed 319 people which was a self-report survey. So don't like I'm not going dig into too deeply. I have paper here so can look at it So in the paper the author Lee et all here, the paper, it says, as Bainbridge noted, A key irony of automation is that by mechanizing routine tasks and leaving exception handling to the human user, you [UH] know why a user has an h to the human user. You deprive the user of the routine opportunities to practice their judgment and strengthen their cognitive musculature, leaving them atrophied and unprepared when the exceptions do arise.
00:05:12.307 --> 00:05:21.225
And just case you thought was miss chance to top rope with Deming on this one [UH] I'm not because again anything you only handle exceptions.
00:05:21.846 --> 00:05:25.105
Anything inside the system should be considered stable and not problem.
00:05:25.485 --> 00:05:27.466
Exactly. It sort of weaves in with this.
00:05:27.846 --> 00:05:30.565
know, mean Deming wasn't saying automate away all work with A .I.
00:05:30.966 --> 00:05:33.600
Exactly. And have steel men for here.
00:05:33.781 --> 00:05:35.800
Right. So let's let's go those.
00:05:35.940 --> 00:05:36.940
Let's go with the first one.
00:05:37.201 --> 00:05:41.367
Oh boy, we should start first one because first one is this first. No, no.
00:05:42.028 --> 00:05:50.194
I was say that the first one is like it's, it's, it's the softball of it is of
Load full transcript
pushback on this one says, look, everyone else is doing it.
00:05:50.992 --> 00:05:56.526
there's low risk in adopting standard, you know, it's there's low risk adopting best practices.
00:05:57.245 --> 00:05:58.805
This point was making earlier.
00:05:58.946 --> 00:06:05.286
Probably, know, somewhat prematurely. But yeah, lower risk is risk mitigation strategy. It's best practice. So we're to do this. Right.
00:06:05.706 --> 00:06:12.060
And then everybody who is kind of, know, antagonized by this no, it's best practice. Makes sense. Let's this.
00:06:12.980 --> 00:06:35.100
Except there's lot more to it than that which you're find out by keep listening because you're find out more about why this even though the scenic makes a valid point as in it's pervasive, it's not necessarily the best point I cut straight to the practitioner partially because don't really understand the academic but [UH] also mainly because it segues the cynic very well. Yeah.
00:06:35.500 --> 00:07:00.966
With practitioner is saying hey man like I'm overloaded like don't think about more stuff, which translates to actually good, a good pushback here, which is take this cognitive load off my plate. Like I got to work on more important stuff or concentrate on only the most important stuff them. Only most critical things or the bottleneck whatever exceptions or whatever. And that's a good thing And also like just framework tool something like that.
00:07:01.146 --> 00:07:24.454
can use like basically don't make me think. Don't add thinking to my plate. Oh, that's the right. Don't overload me Look, just real quick for those you that are actually watching this. The academic is simply saying if you have a framework that you're, know, that you're considering, you really need some, know, some practical steps. So the heuristic shortcuts are what has worked before a heuristic.
00:07:24.613 --> 00:07:31.894
And then you know that we can practice. that's all that saying. But it's not really that much of value adding that. So we're going to skip forward.
00:07:32.120 --> 00:07:35.925
yeah, mean that one can't be understated is I'm overloaded.
00:07:36.125 --> 00:07:44.584
Like take some stuff off my plate. if it's wrong stuff, then least took some stuff off. That's great. mean, listen, go forward. Go forward, young man.
00:07:45.504 --> 00:07:47.805
Go forth and conquer that kid's going] places.
00:07:48.045 --> 00:08:10.932
Yeah. Yeah. It also gives them a little bit a escape patch well. Right. So that's because can just say look it's best practices worked over there and there over there and it didn't hear because maybe didn't apply right. So again it's for me or us or our teams worked on my box. Do we got better points here's some here's some points. know they're better points. Let's talk them.
00:08:10.992 --> 00:08:14.952
Heuristics might fit routine for reversible calls.
00:08:15.634 --> 00:08:20.194
Yeah. also you can't just lift shift from industry to industry. Something like this. Right?
00:08:20.274 --> 00:08:44.888
Heuristics may work in one industry whereas in another industry [UH] and again all things will equal. We're just saying compare like for like if you're thinking about a regulatory environment, can think about that. Sorry, was stuck on. You can't lift the shift metrics from everyone else's metrics Well, you can. be what is your app [UH] My app is[UH] Private prisons A maximum number of users. I just lifted that dropped in.
00:08:45.227 --> 00:08:46.607
Boom. It's great number.
00:08:47.067 --> 00:08:48.388
Yeah, I use that time.
00:08:52.268 --> 00:08:54.847
Yeah, mean, listen, good luck you.
00:08:54.927 --> 00:08:58.988
You're going make that work you. Let us know how you did that in the comments below kids go in place.
00:08:59.028 --> 00:09:09.095
Oh my goodness. one better points though is next one. think best practice getting invoked when context matters most almost read that wrong [UH] That's not drug. mean bug.
00:09:09.888 --> 00:09:23.769
Listen, mean could be, it could be, could be Context is king. Right. To kind of paraphrase a number people that have quoted this before. So yeah, they get invoked in context matters most without context.
00:09:24.109 --> 00:09:25.208
What you really doing?
00:09:25.967 --> 00:09:33.475
Yeah, don't like that point. Number two is kind weak my opinion. think better one is the third one which is[UH] they're saying first have problem with the third one.
00:09:33.556 --> 00:09:48.768
third one says industry standard quote. Industry standard is polite version of quote. I didn't look didn't think whatever. I would push back on that worrying because it's, we're not talking about standard on this episode. We're talking about best practice [UH] already when started, knew should said the opening.
00:09:49.048 --> 00:09:57.634
through quick. But knew I should said the, original promise. We're not talking about standards. Standards is like plumbing code. Like a body people got together.
00:09:57.695 --> 00:10:22.854
Authority has deemed these be standards, which is not the case with you know these things we're talking about. Not not, I'm not talking about standards. You know, I'm not even talking about like general design standards, which I thought about as well when were playing this podcast of you should use first first out queues for messaging. If it's important for you that that things are done certain a like a design pattern, know, that design[UH] and principle.
00:10:22.894 --> 00:10:24.073
People might call that standard.
00:10:24.432 --> 00:10:28.092
I call that standard though. might, might be like, well, that's best practice when dealing with messaging.
00:10:29.232 --> 00:10:35.192
Yeah, And I'm like, because nobody, nobody, no authority said this is how should do things.
00:10:35.714 --> 00:10:43.634
Well, they would say then it's best practice. and then would say, well, now we're back in square one because like there's time to use FIFO queues. Yeah.
00:10:43.714
--> 00:10:57.620
there's time to use standard normal queues that not FIFO. Sure. So weighted queues based on some weight. So your situation, we've think about which design patterns. I would argue that that is not best practice. It's just a design pattern.
00:10:57.640 --> 00:11:45.934
You're applying it for whatever. know, yeah going back to the Lenny and Friends podcast did where Karri Saarinen was we were listening him and was saying when you don't have to think about something, then you do the work but you skip the learnings part That is this last bullet point. Secondhand thinking skips the steps where you would learn And then took away the practice work that goes along that because you automated it or give air whatever. So now that skill is atrophying which the Microsoft survey which is [UH] was self -reported survey was only people. 300 So it's not it's not groundbreaking. But the Microsoft survey seems think that there's something there with skills of that kind atrophying when just hand off the A .I. Of course, were using copilot.
00:11:46.075 --> 00:11:49.075
mean, were here like sticks rocks trying like do you know.
00:11:49.495 --> 00:12:07.735
Yeah, it looks [UH] in the A [UM] .I you know, deference that that is definitely valid. But think it's even without that. I think it's equally valid when you talk about offshoring some that work or offloading, let's sasay near shore offshore whatever. You don't have the learning opportunity there to bring that knowledge back. That's true.
00:12:07.835 --> 00:12:12.654
Somebody somewhere figures it out and then keeps stumbling over the same stepping stones over and over again.
00:12:12.855 --> 00:12:23.428
That's true. There's whole Segway right now could take talk about offshoring or going cloud or [UH] that's a different podcast. bunch other stuff that people just did because that's what other people doing.
00:12:24.081 --> 00:12:24.361
Right.
00:12:24.442 --> 00:12:27.261
Everyone's got cloud. That might the best practice.
00:12:27.621 --> 00:12:46.836
it was the best practice that cloud stuff best practice the offshore stuff was best practice because could get labor at third cost or whatever, fifth the cost or whatever was back day. It's not same anymore And the cloud stuff was because it solved bunch your, know, five nines. And back back day with Sarbanes Oxley and whatever, know, those problems.
00:12:46.895 --> 00:12:53.341
Right. By doing nothing. mean nothing except spending three acts on compute now is lot more.
00:12:53.836 --> 00:12:57.756
The fallacy was well you know the DB skills expensive.
00:12:57.917 --> 00:12:59.236
DBs are expensive.
00:12:59.277 --> 00:13:00.116
just go to cloud.
00:13:00.177 --> 00:13:01.317
don't need all these people.
00:13:01.397 --> 00:13:07.956
I'm not disputing any that but expensive versus my cloud, you know, cloud bill is three X what pay for.
00:13:08.096 --> 00:13:08.496
Exactly.
00:13:08.596 --> 00:13:10.618
So the receipts weren't there.
00:13:11.057 --> 00:13:23.769
Yeah, that's all I'm saying. OK That's, that's whole different discussion is a takeaway here and there is a takeaway, would say stop accepting best practice as an answer [UH] like full. That's it.
00:13:23.270 --> 00:13:40.856
think. Sure if there thanks for coming to the podcast [UH] and then start asking what problem the best practice is being proposed. What problem it solved and maybe for who it solved it for. how know that kind question? know, oh so there's questions screen. Where did come from? Who was built for? Did work?
00:13:40.976 --> 00:13:51.903
Did not work? What happened is decision that we're looking at applying this best practice reversible or irreversible. This is the two way door decision for you Jeffrey Bezos fans.
00:13:52.783 --> 00:13:55.323
And what specific problem this solving for us?
00:13:55.884 --> 00:13:57.884
Right. Right. Yeah.
00:13:57.904 --> 00:14:09.125
And fundamentally, you know what makes best practice when say best. How you, how you qualify that? You know, just because worked for a company or 10 companies doesn't mean it's work you.
00:14:09.345 --> 00:14:10.445
Yeah. Yeah. Yeah.
00:14:10.791 --> 00:14:33.320
So when was last time you heard best practice used instead of designing something with thinking give best one in the comments. That's what I'm asking for adopting best practices, Here's why what we're copying never ends up looking like what's printed on the picture. You know, the restaurant go have nice pictures of stuff and you get the real thing and it's like all wilted.
00:14:33.620 --> 00:14:34.120
Yeah, right.
00:14:34.419 --> 00:14:36.940
Exactly. I call that false advertising.
00:14:37.179 --> 00:14:42.460
So [UH] mean listen, there's And when we are lot other countries call it false advertising. But you[UH] know.
00:14:42.379 --> 00:14:42.659
Right.
00:14:42.820 --> 00:14:45.559
Ha! LAWS? What are those?
00:14:46.952 --> 00:14:53.312
we're adopting best forward. OK, so practices. Do we have podcast on the best practice that got adopted?
00:14:53.552 --> 00:15:13.758
Oh my goodness, think this is is good time that's[UH] right. It's what happens entire industry adopts a theoretical conclusion. Are able to 46 the Spotify model, viral or viral org design that never existed from January 28th, 2026. In that podcast, we found out don't copy the Spotify model.
00:15:13.817 --> 00:15:24.105
That's we, that's we found out You don't copy the Spyfi model because they did because their voice Spotify. All do is you adapt the Spotify vocabulary.
00:15:24.842 --> 00:15:27.524
You're installing whole set of new jargon. the Yeah.
00:15:27.482 --> 00:15:28.543
vernacular changes.
00:15:28.722 --> 00:15:31.414
But underneath that, what you really doing?
00:15:31.434 --> 00:15:32.095
That's right. see.
00:15:32.154 --> 00:15:35.355
Let's talk more about what [UH] secondhand thinking looks like at scale.
00:15:36.274 --> 00:15:37.434
Oh, secondhand thinking.
00:15:37.716 --> 00:15:40.229
with with pushback because people love the Spotify model.
00:15:41.288 --> 00:15:49.548
They'll say, well, Spotify is successful Om! and if be even more cynical, I'll say borrowing from winners rational under uncertainty.
00:15:49.889 --> 00:15:53.754
Really under any case actually they'll probably stop there say borrowing from winners is rational.
00:15:54.143 --> 00:15:58.708
They may say that I'd say it's irrational The winners are winners for reason.
00:15:58.908 --> 00:16:07.869
Right [UH] Or many reasons. But you don't really have those reasons. So also, also want to ask cynic, like sit down, like tell me hurt you.
00:16:08.134 --> 00:16:11.294
who, who's Spotify.
00:16:11.475 --> 00:16:18.687
If you go watch the spot, like if you're like Ed and Ed will[UH] message he'll I watch podcast. We like mentioned something offhand.
00:16:19.087 --> 00:16:29.327
pause podcast and go and do something else reason read whole thing and come back watch your podcast. If you're like that, you can, can pause this now and go listen Spotify podcast.
00:16:29.788 --> 00:16:30.467
think was that long.
00:16:30.567 --> 00:16:35.261
think was lot more hour. But they got like a billion dollars in funding.
00:16:35.322 --> 00:16:36.461
Didn't got like lot money?
00:16:36.522 --> 00:16:39.941
rounds funding. Oh no, no. They got a lot rounds funding.
00:16:39.980 --> 00:16:42.352
But think it totaled maybe lot of money.
00:16:42.393 --> 00:16:44.653
Over billion dollars. don't know was. It's lot money.
00:16:44.913 --> 00:17:23.881
was definitely least several hundred million dollars in funding. So [UH] it's not like there is some scrappy little whatever that know bootstrap their way that. That's, that's not, I mean, maybe that's Spotify story via marketing whatnot. But when you say Spotify successful, they only crossed the profitable line like fairly late. Yeah, couple years ago, think. don't even know they're still there, but a couple years ago so don't, don't tell me that the Spotify model is reason their success. And then again, like said that episode every company that I know that has borrowed all their terminology was a train wreck.
00:17:24.287 --> 00:17:36.920
That's reason because not every company that borrowed their terminology could raise those rounds funding You'd have have the same luck as they did, which is very, very, very difficult. I think say good luck.
00:17:37.380 --> 00:17:46.121
Well, yeah, you look it go, what did they really do with this model in their own organization? come back and go, they never actually used it.
00:17:46.661 --> 00:17:49.661
Listen so [UM] if you're adopt the Spotify model, don't use it.
00:17:49.840 --> 00:17:51.540
That's the corollary.
00:17:51.874 --> 00:18:12.587
you're skipping over the Senate going straight practitioner who says listen squad some cool and Whatever can use to let my management get, know, let me have you know, product teams that stay together and stuff like that rather than using the, know, project model where you run out you get team, them together they do stuff. Then you disband team over over over again. But and also Kniberg said don't copy it.
00:18:12.847 --> 00:18:13.548
said don't copy it.
00:18:13.708 --> 00:18:15.387
So know, people should wise man.
00:18:15.587 --> 00:18:32.133
Listen him. [UH] also like water is wet. Like mean Squads, Guilds, and and tribes. mean, if that's what rocks your boat, go for it, but don't copy the model So the academic here, just to not skip that, don't get the academic here. are they going say certain uncertainty?
00:18:32.712 --> 00:18:54.614
looks like rational[UH] Oh, so, so like we don't know do. So we're just to, this blindly copy it and hope that we'll get same results. Like there's no other environmental factors you're considering OK, I know. I listen the academic. Sorry, I'm not putting any anything on that one because if you're doing that, you're really rolling dice.
00:18:55.314 --> 00:19:07.840
was say [UH] When that way I understand best because you really are completely lost, I start trying to copy. I mean, that's all you're doing is grabbing straws. Right. And hoping for good result.
00:19:07.921 --> 00:19:11.861
mean, look look all, look product people that are now the people. Look that.
00:19:11.921 --> 00:19:12.260
Look them.
00:19:12.560 --> 00:19:16.422
There's all Agiles that were project managers. Not say that trying.
00:19:16.442 --> 00:19:19.040
This is continuation of what we already know.
00:19:19.421 --> 00:19:21.040
shots fired. Oh my goodness.
00:19:21.260 --> 00:19:23.082
Shots already been fired there.
00:19:23.182 --> 00:19:28.788
Put them back [UH] Oh no, you didn't. All right. Here go. So see. Are there are there better points here?
00:19:29.347 --> 00:19:34.968
so under better points I listed arguably Spotify's results came from right. [UH] let's trust and autonomy.
00:19:35.467 --> 00:19:46.079
I don't believe that at all because again go watch the Spotify episode. yeah, the idea was a decentralized movement an org design was not the normal hierarchical.
00:19:46.480 --> 00:19:51.711
know, mean, the, the, the reason like that bullet point right there grinds my gears.
00:19:52.053 --> 00:20:12.885
Daniel Ek was rolling back all that and trying become more hierarchical, [UH] under banner of oh we're just adjusting. But like just like time somebody rolls back any kind of decentralization in order have quote more control I immediately start questioning like, oh, that what you're doing? he really, know, really is about efficiency? that what it's about?
00:20:12.925 --> 00:20:14.185
It's not really.
00:20:14.506 --> 00:20:17.212
It's really about control it. Yeah.
00:20:17.893 --> 00:20:42.714
Yeah. Another good point there. The second one, reasoning learning that produced the model never transfers this [UH] one is a goal. I think, think this the only point on this whole like there's bunch garbage here. think the only one you [UH] just can't copy paste stuff right from
another company's success into your company because of all the things that you already know that different between two works. your cultures different.
00:20:42.815 --> 00:21:09.925
structures are different. Your people, politics processes, you name it. They're all different. So how you expect this to work? It just befollows me that people think this will work. And think it's again going under, you know, under guise of it worked there. It's successful. Let's do this. Right I know somebody who followed the Atkins model. I'm to do that. But hey, listen, I don't ever exercise. It's work for me though. Yeah.
00:21:10.086 --> 00:21:10.786
Good luck with that.
00:21:10.925 --> 00:21:15.107
Yeah, that's old credential parade of[UH] Exactly. They did it. So know.
00:21:15.528 --> 00:21:16.720
Yeah. Yeah. Yeah.
00:21:16.859 --> 00:21:20.680
Renaming teams squads was finally replaced. That's whatever man Yeah.
00:21:20.799 --> 00:21:46.759
know, only, other thing I would have up here is you [UH] know, a savvy viewer might say, well, know, social proof is good for, know, getting these guys changed. I'm oh no, like social proof is good for like restaurants picking out what restaurant you're going eat your friends whatever. Like I don't, know. to base my org design on social proof. know, plus most, most words or design is just authoritarian like yeah, they're impervious list they're just authoritarian authoritarian pyramid most orgs.
00:21:47.019 --> 00:21:47.473
Exactly.
00:21:47.753 --> 00:22:00.499
Hierarchical. Yeah, exactly. Or a matrix of matrices. I've seen that large or it's not just matrix. Or if you zoom out the matrix, still becomes like the it's definitely hierarchical.
00:22:00.618 --> 00:22:03.486
There's doubt about that anyway. Like that.
00:22:03.546 --> 00:22:05.226
Sorry, got, got lost what trying say.
00:22:05.306 --> 00:22:13.452
We're on the take. Yeah, take away. Yeah, I got, got lost from in what were trying to say. So I'll just move on to away, which is stop copying the practice.
00:22:14.038 --> 00:22:25.652
[UH] start by making sure you have the same problem that is solved. So stop copying that practice. But then you look the problem underneath that practice that they were trying solve first place and see that yours same.
00:22:26.231 --> 00:22:31.531
Because even then, if you're copy, least got a shot that you know you're copying something solve same problem.
00:22:31.832 --> 00:22:34.531
But but going back back saying Spotify.
00:22:35.172 --> 00:22:38.852
But you assume that you had the same problem under all the same conditions.
00:22:39.211 --> 00:22:58.978
Or at least the main conditions and there's just so many conditions like how would you even start I think point number one here, which is the same title [UH] you know, stop copying practice unless gives you chance. If you know it's same problem you're trying solve that was already solved elsewhere. It gives you chance to take the practice and maybe mold it your own purpose.
00:22:59.092 --> 00:23:03.711
Right? Right. if if it's not same problem, you're already off. Yeah.
00:23:03.791 --> 00:23:09.352
Right. You've already put your chips or whatever you get double zero on roulette table.
00:23:09.892 --> 00:23:10.731
good luck that too.
00:23:11.192 --> 00:23:17.750
You might, you might spin double zero, but also you might not I only spend double zeros.
00:23:17.971 --> 00:23:21.230
I never do. Even when I bet on that All right.
00:23:21.325 --> 00:23:32.306
right. So next was right. The differences between what they were framing their practices onto their context and yours and see how many common points have.
00:23:32.746 --> 00:23:39.605
Right. If you have more than not. OK, maybe think about it little more if you have less than that though.
00:23:40.465 --> 00:24:20.152
You know, the interesting thing about this we might talk about category three is it requires that you back up directly from the problem and then like wheel back to ask why or to talk to customers or to, something that if you're certain shop or a feature factory type of type of place, you're not asked up there once you try to get up that level. Like it's like you're you're looked at like you're challenging system or whatever. You are. You're not team player. That's right. It's, this another I watched great video today I try send you on the channel or on on slack between because we should totally copy this guy's series videos But this guy talks more about just like working corporate marketing camera.
00:24:20.311 --> 00:24:20.852
He's older guy.
00:24:21.011 --> 00:24:59.746
Can't remember his name but was talking about this basically you're like boxed in by your boss because your boss doesn't know how to define success upfront [UH] they only know success when they see it. the success is defined after the project is finished. Yeah. And then under different definition because now we know what success looks like working manager that like, yeah, don't[UH] like you're along with or know, he was given tips to, to help you have better working relationship them but also not get trapped under manager who's a **** communicator and a person like this who doesn't know to, know, mean they never learned how to define success.
00:25:00.086 --> 00:25:09.398
Right because he's like they're always be they're always going be undermining you basically. You know, they're not not intentionally. They just sure didn't.
00:25:09.479 --> 00:25:27.257
They don't know do do with success up front. So he's like, well, know, when you start stuff, you've got get like written confirmation. And then like even you send something and they're like, that's not right. Then least they're responding with the written confirmation. So they're giving you, know, mean, yeah, not, you're not really tricking them, but you're kind like you're managing them. Let's say that.
00:25:27.337 --> 00:25:33.190
Yeah, right, right, right. I mean, modern phrase coaching them, But yeah, coaching up yeah.
Right.
00:25:33.250 --> 00:25:55.438
Managing up. All right. the point this slide, think to think. I think that's the point this slide. You know, retain the thinking. All Now take everything we just said and swap Spotify for A.I. because that's memo that your CEO may have already sent you work at particular company with somewhere around seven thousand employees to name no names.
00:25:55.637 --> 00:26:00.778
course, we don't do that on this podcast. That's right. Very earlier the pack. mean, except for earlier.
00:26:03.045 --> 00:26:09.505
So ask the first has become the best practice late [UH] but now with humans completely out loop.
00:26:09.565 --> 00:26:11.125
So Shopify is CEO.
00:26:11.444 --> 00:26:23.309
Actually, last year he put out this memo that that oh, have a, know, yeah Toby Luki Luke Luki Luki he put out Shopify memo last year. Actually, was little more over[UH] last year.
00:26:23.549 --> 00:26:24.390
What? 18 months ago.
00:26:24.849 --> 00:26:25.950
Yeah, long time ago.
00:26:26.049 --> 00:26:45.151
Yeah I have tweet pulled up. can read the whole thing but this is core what said. Reflexive, reflexive usage is now baseline expectation at Shopify. and then thing said is before asking for more headcount and resources, teams must demonstrate why they cannot get what they want done using AI.
00:26:45.769 --> 00:27:00.088
so there is April 2025. 7th, It's still up, Tobi Lutki! I just like saying yeah, there's whole thing. It's a tweet. don't understand. This[UH] is yeah, graphic novel over here, man. Like it's not a tweet. understand how there's tweet.
00:27:00.336 --> 00:27:19.641
he wouldn't done that Twitter all day's 140 characters or less[UH] Then maybe there's just like old man yell at clouds here because I guess Twitter has some sort like article writing blog function now. Yeah, I was on Twitter when 140 it characters was I'm good. I'm good mean, don't ever need 140.
00:27:19.442 --> 00:27:20.042
more Good.
00:27:20.490 --> 00:27:31.309
Short and to the point back then. But yeah, now I was say, but then they made big they made lot money, but they never made any profit [UH] That's a different together. Yeah.
00:27:31.710 --> 00:27:49.490
yeah. So the, we read whole memo, the whole, the point me showing it briefly screen if it made it through the editing was that we read whole thing. And this is pretty like I'll talk about citation in the against points because I'm kind actually on side of what's the citation.
00:27:50.084 --> 00:28:09.317
And then the, the Lee et all paper from Microsoft CMU that we talked about earlier where 319 the workers were polled. That's little bit here to higher confidence in Jenai higher confidence in Jenai is associated with less critical thinking. So more use .I. and more inconfident that it'll it out, the less yeah, less critical thinking.
00:28:09.596 --> 00:28:12.237
mean, the tendency is to accept what it gives you. Right.
00:28:12.944 --> 00:28:16.365
I always find cynic helpful in situations like this what would they say?
00:28:16.605 --> 00:28:17.585
say if the A .I.
00:28:17.684 --> 00:28:21.412
said it, I don't have defend it. True I wouldn't say that.
00:28:21.491 --> 00:28:22.511
That's not cynical.
00:28:23.132 --> 00:28:39.750
I mean, that's a, that's silly thing say. Not cynical thing say it is silly. But surprisingly [UM] I'm seeing quite few people fall for this. know, I asked AI and AI told me this. how know it's great? You know, all these smart people fed AI. Yeah, must be correct.
00:28:39.911 --> 00:28:40.111
Yeah.
00:28:41.050 --> 00:28:41.310
Really?
00:28:42.183 --> 00:28:48.088
know. No, that's not real. That's not real argument. real argument is going be. Yeah, but Om the tools are only getting better.
00:28:48.355 --> 00:28:54.217
Yeah. Yeah. Well, that's their saving thing when you have an argument against what they say. well, they're only getting better and getting better.
00:28:54.517 --> 00:29:07.249
Yes, I know [UH] I think when look it that way, it's if you're going continually trust what AI, you know, suppose that the tools may be getting better, no doubt but the humans are getting dumber.
00:29:07.595 --> 00:29:13.035
What told you you're holding it wrong? Oh yeah, right. mean, well, that's this paper is hinting at.
00:29:13.655 --> 00:29:18.469
I'm sure that we could find a better resource. I mean, this paper I think from April. that?
00:29:18.788 --> 00:29:19.868
Yeah, April last year.
00:29:19.909 --> 00:29:20.048
Yeah.
00:29:20.375 --> 00:29:21.536
You'd have open separately.
00:29:21.694 --> 00:29:23.407
There is a published Yeah.
00:29:23.808 --> 00:29:26.989
That's was published. Twenty five. But don't know when was written.
00:29:27.249 --> 00:29:29.449
Usually it's a few months before it.
00:29:29.788 --> 00:29:41.307
Yeah, yeah, But anyway, probably a few No, it's standing I don't know this, the cynics take your it's, it's, it's little sillier than I think a real cynic would be. mean, I think practitioner here is real cynic.
00:29:41.407 --> 00:30:00.275
I think they've put on funny disguises like Benny Hill and that's a runner on the runner screen smacking each other right now. Like the practitioner says AI, as a thinking partner raises floor. think that's a real cynic here meaning, meaning[UM] meaning, like the thing explain this.
00:30:00.315 --> 00:30:14.535
The lowest level knowledge between us now is much higher for the average person than was before because everyone has same tool. So everyone should give at least the same like floor level answer.
00:30:15.008 --> 00:30:37.087
Yeah, but it's assuming that people know how to use the tool equally. I mean there people that simply just put a very, very weak prompt in and take the first response and run with it. And there are people that will critically judge the responses and tweak their prompts over time and you know iteratively arrive at much more solid sort of conclusion.
00:30:37.567 --> 00:30:43.236
So I know. don't think it's black white
It does raise the floor overall.
00:30:43.715 --> 00:30:50.474
no doubt. And then last part is true AI- native. That's what everybody's looking for. So just, you know, use that your resume.
00:30:50.734 --> 00:30:53.134
It's sort saying it's the new thing.
00:30:53.555 --> 00:31:00.815
It's like the Internet was and whatever else. The practitioner is a blend between somewhat cynical but still largely practical.
00:31:01.567 --> 00:31:10.394
this is an interesting one for the comments. Does I raise the floor of knowledge. Does it raise floor knowledge and why?
00:31:10.555 --> 00:31:17.800
Like because it's it's faster search engine like like taking taking software engineering out equation. Does it raise the floor?
00:31:18.248 --> 00:31:20.968
for knowledge, not for productivity or whatever. Yeah.
00:31:21.147 --> 00:31:30.114
Yeah. So one thing does do is it bridges the gap between you know, the question the answer. Right. It's much faster. But Google does that.
00:31:30.253 --> 00:31:31.733
know, Google did that.
00:31:31.933 --> 00:31:40.433
Yeah, that's Yeah AI applies some some sort reasoning to it. Right It also kind personifies the responses to you.
00:31:40.554 --> 00:31:59.246
Right. So we know that but that's, that's not, that's not really main point. main point is it is looking at various things based on a number different perspectives other than just a dimensional or even a unique dimensional Google search. So from that perspective it gives you options how you interpret them.
00:31:59.638 --> 00:32:01.692
That's up you. Some people don't write.
00:32:01.732 --> 00:32:21.125
They just take verbatim what they get. Some people do interpret them and then they ask more meaningful questions. So I don't think it's black white [UH] think that's raising floor. think this ends up being a completely spun off separate podcast just under the question does I raise the floor knowledge like I guess we'd have bound it like [UH] at a company. Yeah.
00:32:21.445 --> 00:32:29.579
You know mean, kind like looking it from the perspective of just of general person street, because talking about raising bar knowledge. don't know. don't know.
00:32:29.819 --> 00:32:51.845
got, I'm trying come around I'm just it's hard. Let's, let's, let's defer that to that podcast. But end the day, it comes down this. What is [UH] the A .I. is being fed by humans. So if what feed it is, you know, somewhat dubious to begin with, how expect anything better out of it? That's beyond me, but will delve deeper. in that [UH] speaking better, are there better points?
00:32:52.184 --> 00:33:00.483
verification in stewardship assumes you still spot a wrong answer hey, hey, stewardship.
00:33:00.564 --> 00:33:12.584
Those vibe coding kids going be real mad when when they, when they hear this one verification meaning I test the function of what the code puts out. And stewardship means know, I'm responsible that area.
00:33:12.844 --> 00:33:14.743
ownership it as well. Yeah.
00:33:15.144
--> 00:33:32.044
but can't spot wrong answer No. If you've got the blinkers on, you've got blinkers on. mean, no amount verification stewardship is going to necessarily highlight the wrong answer. what will do that perhaps is having a cynical point view and say let's verify from multiple sources.
00:33:32.344 --> 00:34:02.384
I mean, the point, first point here is saying you need have verification and stewardship. And you're saying AI is gonna help you, that the assumption is that you can spot wrong answers because I'm say it's raise floor of human knowledge whatever but it's to give you wrong answers if he's raise floor knowledge like that you need raise floor of what people understand right wrong. First before you know you start firing your slop cannons off your ships or whatever.
00:34:02.564 --> 00:34:22.891
We never had a good[UH] edited that whole episode we never had good, we never landed on good analogy They all terrible the humans can't discern good from bad. Then AI isn't really going change that for you. So I mean again, we're kind of scratching away at that other podcast that are referring to. But we are whoever aspect.
00:34:22.931 --> 00:34:33.851
Whoever you trust most or guess whatever trust most is what stop checking. So apply that to this I know tons developers that don't even look at the code anymore.
00:34:34.751 --> 00:34:49.726
know, I mean the AI guys over here. the first ones saying that. And then everyone thought were crazy. then everyone's kind of come around on the the development sphere on Internet about, I even get code. And I don't look at code reviews. don't do code reviews anymore.
00:34:50.313 --> 00:35:11.014
Yeah, mean traditionally we were used looking at, know, answers that stack overflow and things like that now is doing just that. It's just doing it faster. But if was skipping that and just giving you its own opinions, I would be bit more skeptical and say, wait, mean it does do that sometimes. Of course. Yeah, course it's going do that.
00:35:11.313 --> 00:35:43.206
But you know, the speed is much faster these days than it was. But the interpretation still up to human the project shops that out there versus the product model, the project model shops, they already outsourced what build to somebody else. they separated what build from their development teams, told them what build, they decided what was going get built. And then this like outsourcing the learning and [UH] Outsourcing the ability go find new knowledge outsources. Why?
00:35:43.773 --> 00:35:46.432
saying AI great for project model shops.
00:35:46.893 --> 00:35:47.833
It's great for them.
00:35:48.159 --> 00:35:49.059
They can go so fast.
00:35:49.713 --> 00:35:51.733
need to think at all. can just keep going fast.
00:35:52.146 --> 00:35:56.505
is like dream the the project model shop. Any request that comes in through their pipeline.
00:35:56.771 --> 00:36:00.711
Anything comes their pipeline like something broke whatever, like the[UH] just picks up and auto fixes it.
00:36:00.931 --> 00:36:13.599
And That's, that's dream. These folks have of software factories or factories whatever call them. That's their dream. That's new phrase [UH] So they don't have close tickets manually like they always do. don't do anything. Yeah. They just become a passthrough.
00:36:13.898 --> 00:36:18.378
Yeah. For users to complain then the complaints just get handled by machine.
00:36:19.119 --> 00:36:22.278
That's, that's the dream Boy, is that dream?
00:36:22.778 --> 00:36:31.079
It is in the project shops. That's this last one clunkily says. That's what his last bullet point clunky says is we don't need anybody that answers why anymore.
00:36:31.480 --> 00:36:34.840
Right? Just keep building because buildings quick and fast and cheap.
00:36:34.907 --> 00:36:36.726
Yeah, I mean you enough money.
00:36:36.786 --> 00:36:36.947
Yeah.
00:36:39.067 --> 00:36:46.648
Oh boy. boy [UH] I know. I know start with that one Keep building away because it is cheap fast and so on.
00:36:46.728 --> 00:36:49.193
Right. But is it right thing build?
00:36:49.393 --> 00:36:57.773
So we've just basically shifted the bottleneck from building to deciding. Right what we building?
00:36:57.914 --> 00:37:15.047
You just, you're making up all kinds new podcasts right now because deciding, deciding always was bottleneck. And always these kids never figured it out project model. So right. Right. Let's hit take away us this category because Stop asking whether not to use AI because you're going do it anyway.
00:37:15.126 --> 00:37:20.980
OK and then start deciding which decisions it never gets to make. Or would say what order makes them to.
00:37:21.000 --> 00:37:31.039
That's why. So because what happens is wrong are catch it? even have anybody staff more can catch those issues? and then are these things you're handing over reversible?
00:37:31.413 --> 00:37:32.574
These are great takeaways.
00:37:33.173 --> 00:37:40.153
So which decision at your company now goes to AI before goes human. And then who decided that?
00:37:40.494 --> 00:37:41.235
Get them my office.
00:37:41.255 --> 00:37:41.835
That's I'm saying.
00:37:42.054 --> 00:37:42.875
know and us know.
00:37:43.474 --> 00:37:47.235
let me know after you get them the office. That way can get coffee first.
00:37:47.255 --> 00:37:49.295
know, I'm just step over here get coffee. It's cool.
00:37:50.014 --> 00:37:54.909
And that brings us back to the first thing that we said is that the .I.
00:37:55.028 --> 00:37:55.789
adopted this stuff.
00:37:55.929 --> 00:38:03.489
didn't invent this stuff, know, that we're seeing. So start with claim that best practice means somebody else did the thinking.
00:38:04.309 --> 00:38:17.268
And I know we did we end up with it. Did end in one direction ended like 20 directions at least So 10 ask yourself question what makes best practice best who decreed that. Right.
00:38:17.427 --> 00:38:21.288
Or it just a one of many good practices perhaps.
00:38:21.893 --> 00:38:25.599
And while you're asking, I put screen for you watching.
00:38:25.960 --> 00:38:31.000
And then for people listening, I'll read it here's some, some points from this podcast. Yes, there were points.
00:38:31.286 --> 00:38:38.032
So I'll read now so you can [UH] write it down you're driving. don't do that. Let's go back and listen later.
00:38:38.271 --> 00:39:14.958
Twice, three times maybe first ask what problem the quote best practice solve and for whom number to copy the problem, not practice should be copied. analysis anyways don't copy their problems. And then three, decide which decisions that the A .I. doesn't get to make or least doesn't get make first And [UH] that's it. [UH] you got these. That's your favorite. Yeah, I think top one really, know, ask what problem best practice solved and for whom. And also reflect back and say, did it really solve the problem and to what degree[UM] I like that one best. think that's go on the side my monitor.
00:39:16.139 --> 00:39:17.219
we're done. That's it. We're done.
00:39:17.358 --> 00:39:24.778
All right. that's, All right. Well, thank all for staying with us If you're driving, keep driving If you're watching, please like and subscribe.
00:39:25.018 --> 00:39:29.119
You're driving. that if you're driving. Pull over then like Exactly.
00:39:29.179 --> 00:39:33.987
can that too. And yeah, know other topics you'd like us dove into [UH] in future