Soft Skills Engineering - Episode 478: Can you coach self-awareness and my boss is an llm

Episode Date: September 15, 2025

In this episode, Dave and Jamison answer these questions: Can you coach self-awareness? I manage someone who seems to believe their skill set is on par with their teammates, regardless of their constant PR feedback regarding the same issues over and over, the extra attention they are regularly given to help them overcome coding challenges, and the PIP they are currently on to address these issues (and others). What are some approaches I could take to help steer them to better understand their areas for growth when explicit measures don’t seem to get through? I work at a small 10-person startup. The company has absolutely nothing to do with AI, but one of the founders has gone full evangelist. He genuinely believes AGI is arriving this year and that there isn’t a single job, task, or process where an LLM isn’t the obvious tool. Day in, day out, he’s posting links to random AI products with captions like “looks interesting 👀”. It’s like Clippy got a16z funding, moved to Shoreditch, and now spends his days flogging us apps we didn’t ask for. He also insists we “use AI more in development,” despite not understanding development in the slightest. The routine is always the same: He asks the engineering team how to achieve some goal (always involving an LLM). We give a sensible answer, weighing complexity, cost, feasibility. He comes back with a massive pasted transcript: “here’s what ChatGPT thinks.” We pick out what’s actually useful, quietly bin the nonsense. He takes our response, shoves it straight back into ChatGPT, and returns with another transcript: “here’s what ChatGPT thinks.” This has been going on for months. At this point, he’s basically a human middleware layer for ChatGPT — no analysis, no original thought, just endless copy-paste recursion. I’m genuinely worried he’s outsourcing his entire thinking process to a chatbot and slowly losing the ability to engage with ideas on his own. How do I tell him — politely but firmly — that this is both rude and a bit tragic? And, half-serious: is there a prompt injection I can use to jailbreak my founder back into being an actual founder rather than a ChatGPT relay bot?

it takes more than using technical credit default swaps when trying to explain technical debt to business people to be a great software engineer this is episode 478 of the soft skills engineering podcast where i'm your host jameson dance i'm your host dave smith soft skills engineering is a weekly advice show about all the non-technical things that go into the technical field of software development like analogies to things i don't understand something something financial crisis yeah hey you know what happened to the financial to to mortgage what did it what would it even be you know what happened to like the economy i guess in 2008 that's going to happen to our business because of debt it's just it's the same i swear i i recently heard people starting

to use an acronym for the 2008 financial crisis thing i can't remember what it was if you do remember that have you heard no i haven't heard this it's like the gfc or something oh yeah gfc global financial crisis oh the global okay i'm like oh we're calling that the gfc now like okay maybe i'm just behind the times but i don't remember it being called that until now this is how we age dave where this is gonna be in history books now oh i'm telling you i'm sure it is but i've already got a whole list of things that i lived through and people i talked to every day did not my daughter was making fun of me because i drove her to school today in a car that had a cd player in it oh brutal yep such as life i know comes for us all unless you're one of

these patrons dave do you want to thank them oh i'm pretty sure this exempts you from aging oh yes i i i want to thank them so much but i mostly want to thank you for that sweet sweet segue all right you're welcome here goes the list of these are actually the names on patreon of our supporters who contribute at the level where they get a shout out every single week they are let's normalizing calling sorry i already mess up and i'm only on the first name let's normalize calling curly braces squiggly braces even though they're pretty curly and not so squiggly neverendingstorypointer.com.com has too many words i see all right woke up this morning got myself a gabagool the two things security engineers hate are change and the way things are

My actual name on LinkedIn is Yami Debugging in the Dark Candy The Mobile Development Ordinaire First of his name Quote or one escape One drop table myomics Jacob Channeling Bjorn

There is a very particular smell coming from your desk Please create a spike to investigate The Missing Semicolon Christy The World's Okayest Programmer Noah Labhart What's up Soft Skillet Nation? It's your boys J and D here to drop some dope wisdom

I want to change this but it's so funny making Oh, we'll never know what the rest of that was going to be We've got a single open parenthesis, though, in part of that. Oh, no, there's going to be more, too. I've already looked ahead. All right. Daniel Remsburg, Nick Molyneux. If the service org uses a subservice org and the controls of the subservice org are necessary, pretend it's not.

Javier Gonzalez, Chewy, Ted Timbrel, Princess Simon, Candy Cane, Lollipop, Chicken, Bok Bok, Pop Pop, and Dog. Blueberry Pie, Applesauce, Snakes, Input Length Validator. A single open parenthesis. Dan from Drone Deploy. Never is not just a greater on Mars flamingo emoji. I like chicken. I like liver. Meow mix, meow mix, please deliver,

which I just recently realized is an Austin Powers quote. So from Dr. Evil. What? Yeah, he sings it. I think it was also a commercial, but. Isn't it a slogan though? This is what I'm wondering. Is it actually a slogan or did Austin Powers popularize that song?

Or was it a slogan first? I don't know. Okay. We'll never be able to know. We have no way of knowing. trash panda swiss python conference sorry swiss python python summit every october gotta keep the patreon name latency resilient nice job kyle boss can't see dodds that guy over there jenny kim the stochastic parrot

quinton we have not heard from you in a month please go see susan in hr asap jonathan kings and i beautiful functional user documentation jameson yes dave did you know that this episode is sponsored by hormozy.pro all right it's not we might have to put some boundaries on this you've uncovered the secret way to pay much less for sponsorship yep we've never edited one of these but if claiming you're sponsored i mean i don't know it's it's a stretch gray area oh well there you go hormozy.pro check it out i guess never is not just a creator on mars teddy bear emoji oh whoa someone tell will angel how to get the https redirect working for www.vibechart.net soft wars episode one the phantom manager

brayden canes john grant britney ellick benadryl cabbage patch wimbledon tennis match oh man so good so now we have two open parens yeah and if the show ever ends it'll be because of a stack overflow i guess yeah someone's got to correct that next week yeah and those will accumulate yeah then sony needs to write an interpreter for the language formed by the patreon names yeah it'll do something we've got some listener feedback do you want me to read this yeah sure go for it all right this is from anonymous which i hope they were also anonymous when we answered their question or you could figure it out hi dave and jameson i wanted to follow up on my question from episode 426 a year later i ended up getting the promotion i was hoping for your advice that

getting promoted three times in three years is pretty rare really helped me put things in perspective it made me focus on just doing good work instead of wallowing in my own sadness about the timing of the promotion just wanted to say thank you and thanks for all the wonderful advice you provide here so it was three times in four years then i guess yeah still still still fast so good for you pretty awesome reflects well on you i would have given you the promotion after year three though just so you know you're an easy grader yeah super easy all right should i read our first question that's why i invited you here today okay i will fulfill my role this is from an anonymous listener who says can you coach self-awareness i manage someone who seems to

believe their skill set is on par with their teammates regardless of the constant pr feedback regarding the same issue over and over the extra attention they are regularly given to help them overcome coding challenges and the pip they are currently on to address these issues and others what are some approaches i could take to help steer them to better understand their areas for growth when explicit measures can't seem to get through oh man i'm on a pip i get called out of my prs for the same stuff over and over and i'm doing great man to live in that mind would be so wonderful it's got to feel good right oh yeah imagine if something actually great happens it would like blow that mind up this feels more like a psychology question than a work question

because i think there's stuff going on in this person's brain and as a space psychologist i'm not qualified to discuss what it is on earth i may only speculate you're qualified you're just not allowed by the psychological yeah it might be distorted by the atmosphere exactly that surrounds our planet i assume that part of them knows they have issues and this is maybe kind of a defense mechanism where if if they're putting on a stiff upper lip right keeping up a good front yeah maybe maybe pushing back even harder against the feedback is a way to uh bury their head in the sand and and just and protect themselves from something's happening yeah but if if they're on a pip i mean they're gonna get fired and i don't know like this also kind of feels like that that

cosmic justice scale that we all have in our mind where we're like that guy better get comeuppance for what he did to me you know yeah not not not in this that's not the exact situation here but you're like you're asking boy i wish this person would just acknowledge that they're in trouble or that they're not as good as they say they are and i think the problem here is actually in your mind the one who insists that this person acts like he you expect him to act yeah okay i i think i get your analogy to the cosmic justice it's like there is this gap between their skill set and the real world and they like for for the universe to be fair they need to understand that yeah and acknowledge it like say it out loud yeah like i'm not a good developer and i'm on a pip

and i'm i'm just bad it's like why why do you need that person to say this who cares if they say it i just want them to get better at their job yeah i mean that's probably why is because they want them to get better yeah maybe they're just nervous that if they can't acknowledge it that they won't get better but i i think yeah i mean there is a world where some people could just be like unwilling to say out loud how much they're struggling but still be willing to work on the struggle yeah although this person does not sound like that person yeah i mean this specific person sounds like they're going to get fired and that might be some coaching about self-awareness a very blunt um that might be painful piece of coaching that might be the only coaching that

works yeah or maybe they'll maybe they'll who knows the same mind that is able to pretend like they're doing great might be able to say oh man it's really too bad that my company had to lay me off for purely financial reasons yeah yeah i guess this process is not known for its transparency so there could be a way that you that they lose their job and still think it was out of my hands my my boss hated me or something so maybe a wild idea but what if you just tell them you think they need to improve their self-awareness you could just talk about it directly it's kind of blunt feedback and and hard to give but they're at the blunt feedback stage if they're on a pip so this might be better than i don't know it might work i i have low confidence that anything

will work honestly because i know self-reflection is not a strong attribute of people who lack self-awareness but whatever it it saves you from dealing with the object level problems which is the specific instances of bad code like you don't want to just go play whack-a-mole and squash all those one by one because then you turn into a human linter for this person and you're supposed to be a human linter for llms i think so yeah this is taking up all your llm linting time yeah your token you only have so many tokens yeah yeah so you could try it you could say hey there's a pattern i've noticed that i i think you haven't noticed which is also part of the pattern i think we have these same issues over and over again and i don't see you reflecting on them and growing

from them it feels like every piece of feedback is is disconnected for you when really i think you need to examine why do i get all this feedback and how do i change to make it so that people have to watch my work less closely how do how do how do i make it so people trust my work more yeah and how do i make it so i'm not on a pip yeah which again i think this problem will go away because this person will get fired well in a sense your immediate exposure to the problem will go away but if you really care about this person that problem will not go away probably i mean it might yeah the firing might be enough to snap them in out of it yeah but it might not man it does suck a lot it might just snap them into something else some other yeah defense

mechanism i've heard people ask an interesting question when someone exhibits behavior that is harmful to themselves not usually in this context but you could argue i think that this behavior is harmful to this person because it's preventing them from acknowledging the reality of their situation and then improving that reality and so the person like the question i've heard asked is what are you getting out of not acknowledging these problems or what are you getting out of fabricating a fake narrative about your situation and it's not meant to be a rhetorical question it's literally meant like you need to identify why you are benefiting in some way or why your subconscious is benefiting in some way from doing these these behaviors and find try to find another

way to meet that need because you clearly have some need oh crap i'm this is space psychology but it's coming through in a suborbital plane but but clearly this person has some reason some need and it might be a need that they themselves haven't really grappled with with their prefrontal cortex. You know, it might be a need that is kind of tickling their, their prefrontal cortex from below and making them do this. And so this is again, a case where you might, it might be a permissions based approach where you sit down with someone and say, Hey, I really enjoy working with you. I care about you. Can I, can I share some feedback with you that might be kind of hard to hear? Or can I say something kind of direct, you know, and, and when you read their response,

both their what they say in their body language and it seems positive then go with it and and maybe ask them like you said jameson why why do you see i'm confused why do you seem to believe that your performance is really great when there's all these signals suggesting that you've got some big room for improvement that might turn into you losing your job somehow i missed this or forgot about it but this is the manager asking the question so i assume that they have some influence over the pip yeah they're gonna be the one that has to fire this person this is really important for you to like basically you're i okay this this totally changes the dynamic for me that it's the manager it's like yeah you need to get through to this person you

owe it to them as their manager to help them see this reality yeah one outcome you might get is that they have this crippling self-doubt that they're covering up for yeah does that how does that get better right i mean assuming assume that their performance change is is the same immediately because i don't think the self-doubt makes them write code you don't like you know yeah say say they are now aware how is this better if they're now aware of their bad performance yeah and and they have this crippling self-doubt how is it better yeah i mean they've been covering it up with bravado and bluster and now now that it's revealed you're right i i don't know what i'm doing or i'm struggling or whatever isn't that the first step on a journey of improvement

is acknowledging a reality that's not good yeah it's great i mean now you can work with the person right like that's that's like a achievement unlocked yeah right now you're just trying to get to the point where they'll even open the door to that so it's like to me that's the that's the next rung on this ladder i think that would be a huge win so if if that goes well then i think you would expect to see sustained improvement or changes to their the quality of their output based on feedback that they've been given yeah and or even self-reflection on on it might not even be feedback of here's exactly what you need to do to improve but just now they're they're looking critically at the results of of their work maybe that's a good first step and at this point

as a manager i'd be looking for process changes in the way that they do their work things like hey i'm gonna go have a chat with another developer about my intended approach on this code before i go and send the pr out to the team you know just little things like that so you can get like more real-time high bandwidth feedback on things um i'd be looking for that well have we answered the question well i do have i do have one more thought and it's inspired by someone i used to work with who just was so good at having these awkward conversations and getting people to open up and say what's really going on he had a way about him that he was jovial but also really smart and also really caring and also a little playful and silly and this combination allowed him to sit

down with people and just really look them in the eyes over zoom in this case and say dave why does everyone on the team think your work is so bad you know and i'm like wow like how did you say that how did you actually and then he just sits in silence and lets the other person respond it's like he's genuinely is curious like he's perplexed by this question and he really wants your help answering it and what i found is that it put people into a curious problem solving mode instead of a defensive mode and to me like that's defensive mode is what you got to break through with this person so i don't know try try achieving 20 year mastery level management hard conversations in the next week yeah because that's what this guy did well i mean we just answered their question

So I assume that that will confer the confidence and gravitas necessary to pull that off. Maybe. All right. That's all I got. All right. Question answered. Congrats. Congrats to us and to you.

Congrats to us. Yep. Okay. Jameson, do you want to read our next question? Yes. This is from an anonymous listener who says, I work at a small 10-person startup. the company has absolutely nothing to do with ai but one of the founders has gone full evangelist

he genuinely genuinely believes agi artificial general intelligence is arriving this year and there isn't a single job task or process where an llm isn't the obvious tool day in day out he's posting links to random ai products with captions like looks interesting eyeball emoji it's like clippy got a16z funding moved to shortage now spends his days flogging us apps we didn't ask for he also insists we quote use ai more in development despite not understanding development in the slightest the routine is always the same step one he asks the engineering team how to achieve some goal always involving an llm step two we give a sensible answer weighing complexity feasibility and cost step three he comes back with a massive pasted transcript here's what chat gpt

thinks step four we pick out what's actually useful and quietly bend the nonsense stack five he takes our responses shoves it straight back into chat gpt and returns with another transcript this is what chat gpt thinks oh my goodness this has been going on for months at this point he's basically a human middleware layer for chat gpt no analysis no original thought just endless copy paste recursion i'm genuinely worried he's outsourcing his entire thinking process to a chat bot and slowly losing the ability to engage with ideas on his own how do i tell him politely but firmly that this is both rude and a bit tragic and half serious is there a prompt injection i can use to jailbreak my founder back into being an actual founder rather than a chat gpt relay bot

oh wow this is the very expensive tube metaphor playing out yeah this tube has 45 of your company i will note that this question does have several m dashes in it which is frowned upon as a flag, as a shibboleth for being AI-ridden. I don't think it's AI-ridden, though, because, frankly, there are more human touches of mistakes than I would expect from an AI.

Okay, okay. I think this is just a person who likes em dashes. So it's a really good AI, you're saying, that has thrown you off the scent. Yeah, or that. Yeah, or it's been fuzzed enough. Oh, my goodness. Ugh.

Ugh! You're outsourcing your critical thinking skills to a model that was trained on gigantic corpus of English text to predict the next word in a series. What could possibly go wrong? You know, every time I tweet or write a blog post, I think I'm just helping out the future. You're feeding the machine. I would like to think that my output is better than the median text on the internet. Above average. You are above average.

Yeah. So I'm making AI better, just out of the goodwill of my heart. And thereby, making this question asker's life better. Because imagine if the chat GDT output wasn't as good. Yeah. Imagine if it didn't train on the transcripts of our show. We'd give way worse advice. Oh my goodness.

How to achieve... Okay. So this is the founder of the company, right? Yeah. Oh, that's brutal. One of the founders. So I assume there's maybe another, at least one more founder.

And this seems like the non-technical founder because they don't understand development. I got it, got it. Oh, man. You know, there are founders who find their way into obsolescence through many different means as they try to run a business

that they've lost control They've lost the ability to control and understand. And this is just one track that I think helps them do that. The AI track, you mean? Yep. The copy and paste everything you hear into ChatJBT and then send the output to the other person. It's just a complete abdication in this case.

You know? Yeah. I think I've done this at least once. We were thinking about some product direction and and i mean i think i did a little bit more than this but there's some ideas tossed around i sort of tried to write up my understanding of them and then threw it over to deep research that chat gpt tab i don't know they keep moving it around or hiding it but it takes a long time so i assume the output is better it'll i think it goes and does a bunch of web searches and tries

to source things more accurately instead of just generate the content purely but definitely use more electricity yes yeah more fish are thirsty now because of it and then i read it and then just like copy pasted it and said i don't know here's here's deep research's take on on what we were just talking about and to their credit nobody said hey you moron this didn't do anything they were polite but nobody really engaged with it they were sort of just like okay they're like Yeah, I could have pasted that in too, Jameson. Yeah, I don't think it had any insight. It had a little bit of information, but there wasn't a, it was sort of like, so what?

What do we think or do differently because of it? And because I didn't produce it, I did not say, here's what I think or do differently. I was also like, I don't know, man, this is just, here it is. Was this in an area or a topic that you have less expertise in than others? yeah it's sort of like market research on a product domain that i don't i don't use this kind of stuff and yeah yeah so i i also didn't have the ability to easily tell this is true or false see that's that's what gets me with this whole situation because you remember the jailman amnesia effect that we've probably talked about here but have we ever talked about that on the

show i forgot this is the one where you have amnesia yeah okay i'll just explain it so this is what i think is happening with ai nowadays but this is the gel man amnesia effect which i've heard it pronounced gel man it's it's g-e-l-l i think hyphen man m-a-n-n the jello man amnesia effect i believe that's the one anyway the effect goes like this you are an expert in some area and you read some newspaper article or you listen to a journalist talk about a subject and you realize they have no idea what they're talking about and they've definitely said several false things because it's your area of expertise it's about your yeah yeah yeah then you go and you listen to that same news source say something about an area that you're not an expert in and you believe

it right so that's that's the gel man amnesia effect and i find i find myself struggling with this with ai because i regularly regularly see it produce answers in the domain that i know well software engineering that are just flat out wrong and sometimes very subtly wrong where it's like oh boy that's a that's a gnarly dead end that you just produced there that we wouldn't have figured out for a long time or just straight up blatantly wrong like that method does not exist in the standard library like you cannot call that it does not exist and the reason it does not exist is because it cannot exist. SSL doesn't work that way.

You know, this kind of thing? Yeah. I think I saw some LLM-generated Python that used was as a keyword. I saw that. Or it's trying to iterate backwards or something. I saw, I'm like, wait, does that, I'm like, this has to be made up, right?

No, that's some Ruby nonsense. DHH is so mad at you right now. Okay, so anyway. So I think we have the same problem with AI where in our area of expertise, it produces stuff that's wrong because it you know that's how ai works and then we go over and we do like something slightly outside of our area of expertise like a market research and it's like i just want to trust it so much being ridiculously unqualified to speculate on this i i i encounter

that same phenomenon of something i know well and seeing it be riddled with errors but some of those errors are not material to the point it's it's just a a sign that they don't know the domain super well but like the thing actually happened or i don't know and some of them are more if they use it to base some kind of reasoning or conclusion off of and the the the errors lead them in the wrong direction then it makes a bigger difference so i wonder how much of that is just like oh you use the wrong name for that bolt on the part of the engine that this is what i would do if i were writing an article about an engine i would say they're made out of plato i think yeah it would be riddled with errors that would indicate i don't know anything about engines but if i was

just saying like and this car crashed into the other car with its play-doh engine yeah with play-doh engine somehow somebody died despite everything being made of play-doh then maybe that's okay turns out it's a non-newtonian fluid it gets hard when you touch it fast yeah yeah i wonder if the i wonder how much of that like how much do the errors matter and it's hard to know but well i mean it can really matter i mean like market research it's like oh yeah they have 10 of the market it's like really like how how do you know that that's right you know yeah i guess we're not even i'm i've diverted far afield of answering this question this started off as a story about how i have done this but i i would like to think i don't

regularly do this i do regularly talk with ai to refine my ideas yeah but i am pretty infrequently just saying here's what happened what do you think and then telling someone else here's what it said when i asked here's what happened what do you think yeah i very rarely copy and paste my ai output and just give it to someone yeah there's there's a bunch of dystopian y combinator funded startups to like put smart contacts in and then hook up a chip to your throat that'll speak for you i don't know i'm sure there's some imagine you didn't have to know or think or do anything and just the right answer came out all the time yeah your mouth moves in ways you couldn't predict but it's always beautiful there's a rick and morty episode about that where he has this gem that lets

him see how he will die and so he starts taking actions that lead him to this death that he thinks looks good looks like the way he wants to die so he'll like slowly start to mumble letters until he finds the right letter of the right word to say it's just like stumbling through life towards the good ending that's funny this is both rude and a bit tragic yeah the fact that it's the co-founder makes this tricky yeah it is tricky surely everyone knows though right like if he's doing this for everything i'm sure he's not doing it just for the engineering team i mean this goes back to that cosmic sense of justice problem that we had a minute ago where it's like yeah do you really want to solve this problem or do you want to make this fool acknowledge their foolishness

you know like is that what you actually want because there is a very simple solution to this problem just ignore it because you you know it's ignorable because the dude himself hasn't even read it right he's just copying and pasting between terminals like between text boxes like All I do all day is I take text out of Slack, I put it into ChatGPT, then I take text out of ChatGPT and put it into text, and I make $300,000 a year. It's really great. But the real problem for you doing your job can be easily solved by just ignoring this. Now, do you want to tackle the cosmic justice scales that are out of balance? If you do, then I don't know.

I'm not sure it's really worth it. I'm going to disagree with you, Dave. I don't think ignoring it is the solution because I think the co-founder is meddling. They're not just saying, here's some stuff. You have to read it. You have to respond to it. You have to parse it. It takes time away from other stuff.

It's that whole, it takes a lot longer to refute BS than to produce it problem. And there's another problem, which I have encountered a lot when I was bad at school. It's easy to feel like you are learning stuff, but not actually learn things. yeah you get to kind of pretend develop skills and knowledge and so i suspect this co-founder also thinks like you know i'm really learning a lot about the domain of software engineering from copy pasting things back and forth between chat gbt and they're going to start pushing for you to take real world real life actions beyond just taking up your time with it that i think are probably going to be bad or at least lacking in context and nuance so i i don't know i hope

there's a technical co-founder that's probably what i would do yeah i would try to talk to the technical oh interesting hey we are we are being i edited this out but they have three engineers at this tiny company so hey you're losing i don't know 25 of your engineering capacity because you're responding to this bs pasting yeah like please help and hopefully the co-founders have a relationship where one of them can say hey you're being really annoying to the other one that's true the small the small shop or the small uh headcount does change the dynamic a lot both like you said both in terms of the negative effect on this and also the resolution because yeah you probably could just say hey other co-founder help us out here i wonder why like

i mean to go back to your point in the other question what what need is he meeting by doing this is it that maybe he feels like i don't know what the eng team is doing i don't because i don't understand it so this is trying to help or i just i feel like things can be faster i think it's really easy to look around at some other company or team and say look how fast they go we should be going faster and have a very incomplete picture yeah so like yeah is it just he's really pumped about ai and thinks it's the answer for everything okay also i mean there's there's a there's room to be really pumped about ai and be aware of its weaknesses and i think if you're really pumped about it you kind of have to be aware of its weaknesses otherwise you're just

gonna you're gonna fail pretty hard yeah so there's there there's a lot of content out there about how to use ai effectively that acknowledges it is imperfect it has impacts on how your brain works it gets stuff wrong it can it can lead you down novel failure paths and it just seems like he's at step zero of this which is like oh cool i can i think the first thing i ever did with chat GPT was ask it to write a poem about git and it was like sensible and yeah i was very impressed cool you can you can talk to the computer that's the stage that he's at yeah you could talk to the computer awesome you can talk to the computer and go back and talk to your other founder friends and tell them how cool it is that i am running my business with heavily leveraging ai yeah soon

ai is going to replace all these engineers that keep telling me actually this is harder than you think yeah i wonder if that's it too i wonder if this engineering team has a reputation for pushing back too much and this co-founder is like i'm sick of being told no or i'm sick of being told later yeah i want i want someone to tell me yes be sycophantic i want someone to tell me you're absolutely right yeah what an excellent point i think another way you could approach this is to push on push on him for more original thought or analysis to say hey i understand what chat gbt thinks but i want to know what you think but you really don't right i guess if it's like how should we test our code yeah you probably don't

but i i think there is if he's a co-founder of the company like that's that's part of what you need the co-founder for is to have this vision of where you should be going yeah it is pretty worrying if the vision of where you should be going is generated by an llm yeah stochastic token prediction and then some temperature thrown on top of that yep both rude and tragic yeah i wouldn't say it's tragic i would focus on how it is harming your output how it's how it's worse for the business and what you need would be better for the business maybe you say hey this is rude i kind of agree i feel like there's probably a what is that an etiquette around ai use that is still developing i've seen some people when they write blog posts they'll put a rating for

how much of it was written with an llm from none of it to it was completely generated to somewhere in the middle a two from a one to five is it maybe maybe suggested a topic sentence or something like that and i feel like there's this this seems to fall under a general developing etiquette of it's just annoying as a human to be told here's what the ai said yeah it is annoying it's like what's the point of view yeah maybe write that book on etiquette yeah how to use ai with good manners yeah ai manners llm llm finishing school cotillion for large language models yeah i mean i guess that's reinforcement learning human feedback right i guess that's so but this is more like human reinforcement maybe chat gpt

should be able to say it sounds like you're asking a question on behalf of an engineering team that you don't understand add add the following disclaimer that i mean they ask is there a prompt injection i can use to jailbreak my founder and if it is chat gpt then they do have some settings for you can you can set a system prompt you can set up some memories so you could maybe just hop on over there while they're at lunch and type in i am a stupid dumb dummy moron that doesn't know anything about engineering and will just blindly copy paste stuff back and forth please dissuade me from this please convince me not to live this way yeah that's so funny yeah you could do that well that's all that's all i got me too me too all right i think we've

answered the question then i think so good luck to you good luck you're absolutely right good luck i think this problem is coming for us all to some degree or the other yeah one degree or another and i may have already caused it also so yeah it's so weird brave new world we live in what can people do if they want their own questions answered dave if you want your own question answered go to soft skills dot audio and click the ask a question button where you can fill out our little form you can put your name or not you can put your question in or not and we'll answer it to the best of our ability thank you so much to everyone who puts in questions every week we love reading your questions we love reading your questions

more than chat gbt likes reading our questions which is how we actually get the transcript for this show wait really no i've never done it okay i mean i'm assuming somebody's scraping all this and what this content the internet's the overlords have got to have it yeah maybe you can't fathom an ai without our influence on it i know that'd be a tragedy a tragedy indeed all right thank you for listening we will catch you next week