Soft Skills Engineering - Episode 513: Forgotten employee and what skills actually matter in the AI world

Episode Date: May 18, 2026

In this episode, Dave and Jamison answer these questions: How long until I am ethically obligated to let my company know I’m doing no work? There was a big re-org at the company I recently joined. It’s a huge financial company, so there’s a lot of bureaucracy. Somehow in this mess I was kicked off the first team I joined so that I was pending re-assignment. It’s been a month and I haven’t been re-assigned. Instead of being in between jobs not working, I am currently in between teams, but still getting paid! This feels unintentional on the part of the company as no one has followed up with me. I’ve been happily using the time to catch up on hobbies, but I’m starting to get concerned about what will happen when the other shoe drops. At what point do I start to venture into irreversible reputational harm? For some added context, teams at this company are about 6 people and the manager handles several teams, so I’ve never talked to my manager since I joined. I got the message about the re-org from a screenshot of an email that my scrum master shared in my last stand-up. AI changes how software is build. Actual coding is mostly done by LLMs now (I hate it, but it is what it is). What do you think should I focus on learning now to not become obsolete in 1-2 years? I am a senior dev with 13yoe, and loved coding, but now I find it hard what my role will be in the next years (still have 30+ before retirement). I do not want to go to the EM track, at least not in the near future. So is it more architecture knowledge, security, AI internals, …? What do you think?

it takes more than threatening the ai with a pip to be a great engineer this is soft skills engineering episode 513 i'm your host dave smith i'm your host jamison dance soft skills engineering is a weekly advice podcast for engineers who have turned into computer managers instead of computer code writers yeah robot managers i wonder if the overall level of understanding of what it's like to manage a team of engineers has risen among software engineers or is it just different i don't know also i wonder if threatening it with a pip is threatening it to use python instead of a performance improvement yeah the ai is like yeah i really hate python i really hate the package index i i think yeah i do what

you say what does pip even stand for python is pretty it's yeah that that is correct i looked it up package index python what seriously python index package i know there's the pi pi the uh python package index yeah what is pips maybe it doesn't stand for anything the pip stands for install my package all right uh dave should i thank your patrons yes go for it thank you to these folks that we shout out every single week for their prodigious contributions thank you to giant panda is best panda 42 is the best school ai made me top code producer company went bankrupt air your will angel name variant must contain at least one uppercase letter one number and one special character oh man i'm starting to believe these aren't all actually will angel

happy birthday from episode 400 that's funny will angel if you can hear me please save us i start every prompt by thanking my llm in advance it's a token got it my actual name on linkedin is yammy debugging in the dark canny the mobile development ordinary first of his name name of his ordinary development mobile the canny dark in the debugging yammy jacob shandling kvolti oprah kwan turty or as my mother used to say fiat clef who's yes i don't get how the person who called others out for not changing their name could have changed theirs okay what are were they even doing and then it cut off i'm pretty sure that meant doing when are you changing your podcast name to will angel engineering an enthusiastic slice of bread nick molyneux javier gonzalez

chewy meow mix angel drone deploy salted fries single open parenthesis ted timbrel i've tried every tts service on the market this is the slowest one this is the most expensive one this is my favorite one it's not the most expensive we're only what 20 bucks a month it's great yeah and you can do like you can do like 20 words dozens of characters you can change it once a week yeah a single opening parenthesis dan from drone deploy never is not just a crater on mars flamingo emoji i like chicken i like liver meow mix mechanics please deliver kyle boss kent c dodds slowing down jenny kim the stochastic parrot ira chan jonathan king is not a beautiful functional user documentation ask your doctor whether this patreon name is right for you

will angel official has heard your call and will save you i think that's supposed to be officially dash dash f braden canes john grant britney ellick will angel won't go to slash new conference may 27th 28th in newcastle australia but will you closing parenthesis followed by a literal closing parenthesis closed parenthesis and then closed that's a killer one oh man thank you thank you for your support thank you if you want to join them go to softskills.audio click support us on patreon any amount will get you an invite to our slack team enough of an amount we'll get access to the slowest most expensive and best text-to-speech service out there that's actually hilarious i love that yeah that's all we are dave do you want to keep going with your text-to-speech

service by reading this question out loud yeah this part of the text-to-speech service is kind of a monte carlo like you can type in a lot of text but the chances we read it are you know it's low yeah but it's the free tier so you get what you pay for all right yeah this comes from an anonymous listener who says how long until i am ethically obligated to let my company know that i'm doing no work there was a big reorg at the company i recently joined it's a huge financial company so there's a lot of bureaucracy somehow in the mess i was kicked off the first team i joined so that i was pending reassignment it's been a month and i haven't been reassigned instead of being in between jobs not working i am currently in between teams but still getting

paid this feels unintentional on the part of the company as no one has followed up with me i've been happily using the time to catch up on hobbies but i'm starting to get concerned about what will happen when the other shoe drops at what point do i start to venture into irreversible reputational harm for some added context teams at this company are about six people and the manager handles several teams so i've never talked to my manager since i joined i got the message about the reorg from a screenshot of an email that my scrum master shared in my last stand-up oh my gosh this is crazy oh some executive spent like weeks and months of their life carefully planning this and

writing all these communication guidelines and they have all these processes and procedures and it's yeah it's funny to see how the best laid plans and intentions devolve into a screenshot of an email that you see in slack for in a stand-up that is so funny oh my gosh yeah oh wow you know i've actually been okay i'll tell you i used to think you know that capitalism would not let this happen you know like the free market but i have been interviewing people and i gotta tell you the number of people who are basically unable to describe the software product that they work on in a way that makes sense is kind of high you know i talk to people and i'm like these are like product managers so like your job is to manage the product

design the product yeah market the product you know like tell engineers how to build the product but when i ask you what does your product do and after a 10 minute explanation i still don't get it i'm like wow there's a lot of people kind of just coasting between i don't know what the right word is but here here you have someone who is explicitly being paid to do nothing whereas most people are just implicitly paid to do nothing they're being paid and they don't think they're doing nothing but they are doing things that you're saying i think so i've been thinking about this recently i guess this isn't a new thought but there are definitely people who do stuff they work but their work creates more work for other people or it's just work that doesn't need to get done like

they're busy but them not doing anything might leave people better off yeah i don't know that's a spot you don't want to be in this is also a spot you don't want to be in no this is not good and uh but i gotta say like okay i'm gonna i'm gonna mention capitalism again but you know it's so funny that the capitalism gets a bad rap for being super hostile aggressive like we will cut your job so fast and put you on the street we don't care about you but then also this happens it's like whoops we kept paying you and you didn't we didn't know it's like how is this the same system that is so ruthless and efficiency driven and also we're paying i mean this is an engineer it's not like they're it's not like minimum wage this is real money you know yeah

i mean maybe it's the intersection of bureaucracy and capitalism yeah maybe doing some some funky stuff okay i guess what you're saying is if you tell us the name of this company we will go short the stock and then that will provide signal to the market that bad stuff is happening and then eventually that will flow down to your manager who will look at their team and say wait a minute there's one engineer here who isn't doing anything yes and then that will get resolved and then our short positions will be worthless well that we'll have to close them first right of course yeah we'll have to time time this insider trading carefully yeah although is it non-public material information if it's shared on a podcast it's public yeah so it's fine it's

public we're not insiders we don't know what the company is this is all a joke and also space financial advice but not earth financial advice that's right i think you should say hey i am not on a team and i need to get on a team who do you who do you even talk to about this you don't have a man i think it sounds like they have a manager i've never talked to my manager since i joined but they did say my manager so i assume that means they do have a manager they just don't talk to them like yeah they're busy they've got a bunch of teams it sounds like they're not doing much management if they're spread across several teams but whatever but i would like to think that they at least have the time to hear someone say hey i'm being paid hundreds of thousands of dollars

in kind of total carrying cost right maybe that's not even your take-home comp but whatever and i'm doing nothing and every every manager would like more stuff done it feels like an easy conversation hey i got lost in the shuffle give me something to do the instinct in this question that was kind of implicit was how long do i go before i do irreversible reputational harm that is exactly the right instinct because if you show up six months from now and you talk to your manager and be like hey look i have been unassigned for six months and i've been let me just tell you i've shaved five points off my golf handicap that's not going to look very good yeah i've played a lot of bellatro and i think it's close enough to real poker that i'm ready to dip into

professional yeah at what point do i start to venture into irreversible reputational harm i I mean, you're close. Probably already. Yeah, you're already there. Although in this company, maybe not. The longer you wait, the worse it will be. Maybe your manager is going to be like, wow, you too? Yeah, if they have several teams,

maybe they are doing nothing and counting on each team thinking, wow, they're really focused on the other teams. Yeah. They're just so busy with these other teams that I guess it makes sense that I never talk to my manager. Right. Could be shell games. Could be turtles all the way down.

Yeah. but like i think we both are in agreement here you are if not already in the reputational harm camp you are minutes away from it like you gotta get this resolved right away yeah outcomes here are you hide as long as possible and then get fired or laid off but i feel like you're just kind of always waiting to get found out that's stressful it feels i i would be unhappy because i would feel dishonest oh yeah i feel like i was ripping off the company and would be found out and the anticipation of being discovered as as someone who kind of was dishonest that just always weighs on me even when i feel like i'm not being dishonest so if i felt like i actually was then

it would crush me so i would not be able to just like chill and enjoy hobbies until someone found out a year or two later but i think there is some threshold at which it goes from like oh you get a bad rap to oh you get fired for you're probably not there already if they just have all this chaos of a reorg like yeah you're you're brand new and in my opinion it depends a little bit on why you got unassigned were you intended to be laid off but they just you know what i mean like yeah like maybe maybe they missed you somehow and you were they intended to lay you off in that case you probably don't want to raise your head up too high you want to go find a new job while you're still getting paid and then quit this job once you found a new a new one it says reorg though i feel like

reorgs do involve laying people off at times but if it was an actual layoff i'm assuming you would just say a layoff yeah so i'm assuming this is a they've just been lost in the shuffle and yeah great company has one extra engineer now that seems like a positive right that being the case i would definitely go to my manager right away and if you can't figure out who your manager is which is a real possibility. Go to your scrum master. That person seems really informed. Yeah, they know what's going on.

They had a screenshot of the reorg list. If that doesn't work, go to your company's website, find a name, email that name. Yeah. Say, hey, I've been doing nothing. You're on the board. Yeah. Find somebody I can tell that I don't have a manager

or any work to do. And then you're going to have to come up with a story for why you didn't say this a month ago. Yeah. You know, I think the story is, look, I've been navigating the org chart trying to figure out where I belong. I've talked to a bunch of people. I'm bringing this now to your attention. I probably should have done this earlier.

I apologize. It's been tricky trying to figure out where we all landed. But you need some story like that. Because otherwise, if the story is, I used this time to catch up on hobbies, that's so bad. Yeah. I mean, this rhymes with office space a little bit. I guess you haven't been. Oh, my gosh.

It so does. this is like office space two fired and are still working but wasn't there a guy who just decided not to do work isn't he the main character pretty much yeah that was him not gonna do anything and see what they do to me and what they do is like give me a promotion trust me more maybe you could go found a new team like write a charter document and get this thing going go hire some people i wonder if there's a secret underground i guess this is a silicon valley plot point also there's a secret club of people who are abandoned like you and can you yeah start a skunkworks project yeah go find all the other ones that'd be awesome yeah well have we answered the question i think so

good luck don't get forgotten that's the message for y'all be unforgettable listening to this question yeah yeah should i read our next question yes please i have not just just the reason you're allowed to do this is because even though we had a massive reorg here at the podcast you did not get left behind you're unforgettable oh good thank you i've contributed enough this is from an anonymous listener who says ai changes how software is built actual coding is mostly done by llms now i hate it but it is what it is what do you think i should focus on learning now to not become obsolete in one to two years i am a senior dev with 13 years of experience and loved coding but now i find it hard what my role will be in the next few years i still have about 30

plus years before retirement i do not want to go to the engineering management track at least not in the near future so is it more architectural knowledge security ai internals what do you think hmm actual coding is mostly done by llms now yeah depends on what you mean by coding if it's like generating the characters in the source code files then maybe yeah like typing i think that probably describes yeah the the state yeah i think most of the characters created that go into new code written at my company are created by lms yeah a lot of editing happens by humans obviously a lot of guidance goes into those characters yeah yeah but there's a lot of reading code yep boy is there a lot of reading code a lot more than ever yeah how do i not become obsolete in one to two years

i okay there's a bunch of different takes here i think you can become more important than ever in the next one to two years. I don't think software engineering is becoming obsolete. There is a camp that says LLMs are producing so much slop that they'll need intrepid heroes to come in and rescue them. Or the next generation of LLMs. Yeah, I don't think I fully agree with that, but I do think there's a lot more room for software engineering

where the small scale of write the function that does this thing is becoming less a part of the work, But architecture and identifying broad patterns and kind of like, we've talked about this a lot, but making the system steer itself in a good direction is higher leverage than ever because you have a lot more capacity to just like, what's a good metaphor here? You've got a rocket that you can strap to the back of a boat, and so you better have a really good rudder? Is that how boats work?

I think so. Yeah, you attach a rocket and there's a rudder to steer. I'm pretty sure that's how boats work. taking what i know about space and applying it to water it applies i believe that's correct yeah the space rudder yes the space rudder also works that way yes exactly yeah rocket plus rudder yeah the rudder is what grabs onto the vacuum and steers the spaceship exactly exactly you deflect the vacuum left and right and you get equal and opposite yes yeah solar sails i think or thing yeah anyways so what is my point my point is you can invest a lot in steering the system so that

this added capacity to generate code takes you in the direction you want which i think was some people's dream right they don't want to sit down and like implement the tickets they want to build these abstractions and architecture and broad direction and that's always been a thing you had to fight engineers to say like stop go write the feature or some engineers anyways yeah but that can become more of the job now. Yeah, it totally can. And I think there's kind of a couple of ways to answer this question. One is how can I be happy at work and how can I not be obsolete? And the not be obsolete is the one actually explicitly stated in the question. But I think the kind of be happy at work question is very much relevant to that. Like the more excited you are to do the job,

the less obsolete you'll become, assuming the job is actually useful. And so, yeah, like, you know, Some of the things I like to focus on now are more systems focused, like architecturally, how do these components interact with each other? What are the layers? What are the interface points? What are the failure points? This makes you really valuable in situations like outages. Like you can be the person who figures out how to resolve them. Very valuable also in like security review.

You can be the person that understands the threat surface area of your software. and these are things that an llm can sort of do but they usually don't have enough context to do it well without a human who understands more than just the text on the screen oh i had a thought this is maybe a short-term answer to your question but you can look at where the benchmarks aren't saturated yet and there's a big debate about why are they saturated because they're just in the training data or do they actually reflect real capability but the general like software engineering benchmarks i think are getting we keep having to invent new ones because the lms just solve them all but last i checked the more ops focused ones like you said resolving outages or

incidents or kind of dealing with prod things were not very saturated and that is an area where knowing the system really well helps a lot too yeah so maybe that's an answer is is know the system well enough to be really good at firefighting or to help lms get better at firefighting yeah This is also the LLM output I've seen from engineers that know the existing system really well is so much better than the output from engineers that don't. Yeah, true. So I do think like knowing and or creating a good theory of your program is more important than it has ever been. Yeah, totally. And LLMs can kind of help you with that.

They miss some stuff, but they're really great to talk to about existing code bases. if it's just a spaghetti mess, then there's not a clear theory to form right away. But even just knowing the code base better is higher leverage than before, I think. Yeah, I totally agree with that. And it's interesting, what I've noticed is that engineers who are more experienced, like this question asker, 13 years of experience, it's possible that you've already started to get into this world a little bit, maybe a little bit more leadership, where you're giving instruction to other engineers, maybe a little bit more systems thinking, maybe a little bit more architecture. And I think LLMs are only going to make you more

powerful, because you, especially if this is true, but you've probably already gotten accustomed to writing code a smaller percentage of your time compared to your first year. You know, in your first year, it's kind of like, take ticket, write code, you know, get code reviewed, ship to production, okay, next ticket, write code. But you know, as you as you gain experience, you're often expected to do more of these architectural things, more of these direction giving things. And I found that LLMs takes people like that and turns them into just super powerful team members. And the last thing I would consider them is obsolete because they know, like you were saying, Jameson, how to direct the LLMs because they're used to directing people the same way. And LLMs are

actually very, very similar to that. And they're also used to making architectural level decisions and LLMs are just not great at that yet. So that's where I would focus is how to give great instruction how to design great architecture how to really understand like how systems work what are the choke points what are the failure points where the security problems like we were saying before and double down on that and i think you'll be definitely not obsolete at least for the foreseeable future security is an interesting one because i don't know a month or two ago by the time this episode comes out there was the big kerfuffle about mythos and anthropics new model that oh we we can't possibly release it because it's too good it's so good that it'll destroy the

world with finding vulnerabilities and we've seen both a bunch of fixes flow into popular open source products and some people come out i read a blog post by the curl guy who said yeah it's pretty good i don't know found some stuff it found one bug one security bug that was pretty small i've been using frontier models for a while to look for security stuff and it i don't know it's not that different so i think security is interesting because lms are both really good in some ways at finding security issues but also i think it does require a deep understanding of the project you're looking at to differentiate the the noise from the signal i've played around with them quite a bit well not quite a bit a moderate amount on my work code base to look for security

issues yeah and was astonished at some of the stuff it found and had to wade through a bunch of garbage right a lot of noise identify what yeah what the noise was and i don't think i could have just like harnessed my way through to get just pure signal out of it i don't think i could have like arranged the prompts enough to only get the good stuff and none of the bad stuff so like you can go deep into security also lms are good at security also i think you need good context to be good at it yeah i don't know also there's a difference in scope well maybe scope isn't the right word maybe level of abstraction there's a very big difference between finding security bugs in code and understanding the threat model of a system you know it's like oh that's a buffer

overflow that's a sequel injection like or oh we have a problem over here where this authentication token that we use on this service is also used over here and so if an attacker gets that one they get this one you know it's like that's a very different kind of a very different level of analysis yeah that's pretty timely there was a yet another tap the sign there was another npm yeah ecosystem vulnerability or somebody got pwned and then someone used their credentials to publish a bunch of malicious packages malicious versions of popular packages and it was all around like system boundary stuff of this github action runs in this context with this credentials that poisons the cache that then runs in another context with secure credentials

and et cetera, et cetera. Yeah, it is. It's crazy. I think we're going to see a lot more of that. In fact, security might be, I don't know, this is a little crystal ball-y,

but security might be an interesting area to specialize in for the coming years because I think we are, I mean, I don't know. I've seen a lot more security incidents lately.

There was the crazy Instructure hack that happened this week. The one you're talking about, the TanStack hack. Yeah. There was a gigantic mega GitHub vulnerability

that was not exploited i think but was insane where i think it gave you just access to everyone's private source code as far as i can tell and i think we're going to see a lot more of that stuff and so hardening these systems against an adversary that is just more powerful than ever is going to require a lot of human attention i think so that's an area you could focus on that I think won't become obsolete, but who knows, man. One of the challenges with this question, although the question asker was generous enough to say, how do I focus on becoming not obsolete in the next one to two years? I'm like, yeah, that's about the horizon, right? Like one to two years. I don't think anyone can see past that right now. A few years ago, I literally wrote

a blog post. This is almost 10 years ago now, but I wrote a blog post. No, it wasn't 10 years ago. It was less. Anyway, I wrote a blog post that said, here are the skills in software engineering that have served me continuously for 20 years. And I listed them out and I realized in the advent of LLMs, I think maybe all three of them, two of the three for sure, are basically you don't need them anymore. Like overnight, they served me every day

in my career for 20 years and then LLMs have made them basically obsolete. So it is very hard to anticipate which skills are going to be valuable and which skills are going to be obsolete. Now I'm looking for your blog post. I'll have to dig that up. I see a bunch of Python stuff from 2013.

That's old. This is you, right? I found it. Yeah, maybe this is your old blog. Oh, it's on Medium. In case that helps. Oh, okay. The title is The Three Highest.

Well, Medium is also old now. Oh, you're saying I've used two different fully deprecated blog platforms. Well, no. Well, I don't like going to blog platforms because I feel like they're all going to go away eventually. And I'd rather just have my stuff on my own website. Anyway. Anyways. But it's hard.

it's hard to anticipate but right now what i can see is that developers who are crazy skilled at understanding complex systems holding lots of context in their mind and going way beyond just like this source file behaves this way on this line like who go outside of that purview some of them are very happy and they're also extremely valuable maybe even more valuable than ever in fact i would say the gulf is widening you know it's always been the case for at least for the last 10 or 15 years yeah the the floor is much higher now because lms are much better than like i don't know how to say this without sounding callous but very bad software developers lms are just way better than very bad software developers and i'll say like again how do i say this without

sounding kind of arrogant or callous but there are so many things that i can just own it you know what we're air we are arrogant and callous okay there you go you've listened to 513 episodes it's time to reveal the truth comes out i'm twirling my very short mustache hairs you know for many many years as a software engineer i've had more ideas of things i can do on my code or in for my team oh then i could possibly ever actually do and now i can just do them and this makes me more valuable to the company because there's like there's information i can find there's there are tools i can build there are little scripts that just aren't we're not worth the time to create in the past like i don't have three days to create this thing and iterate on it but i do

have three minutes to prompt an llm and get it and it'll probably work in another three minutes and so for me this has made me a lot more valuable not because i'm good at using the llm which i'm like okay at it but because i have these ideas because and the ideas only come because i have a deep and broad understanding of the systems i work on at work the um i'm glad you called that out i've felt that a lot and haven't really explicitly said it but the the amount of times 10 years ago i've said and we'll probably never do this right is much lower than the amount of times i say it now because yeah here's this great thing that would that would improve the developer workflow or fix this nagging bug in this area of the code base that isn't that used and here's the

pr for it is it's easier to say one other idea is you could move more into product facing work probably the engineer i i know who's the best at llm use or maybe the most ai pilled told me a while ago i think i'm just gonna go be a product person like i'll still sort of now i'm paraphrasing but my my reading of that was he's still gonna maybe poke at the code a little bit but he's just gonna work at the level of abstraction not at kind of the architecture and system level but at the like build these features for this user set level which is another interesting direction not everyone likes that or could be good at it but there's always been that shape of engineer that's really product focused and they kind of do the technical work because

they want to get the satisfaction of like shipping the feature that they think should exist and you can go really far in that direction now and i think you are much more empowered than you were before if you had to write all the code by yourself yeah i think in a way it's maybe maybe your product skills are a little bit weaker than if you are a product person trying to use an lm to ship code and you don't have the technical stuff but i think you could probably go a lot farther coming from the engineering side with an lm then yes that is definitely true if you just want to like deliver excellent products to users yeah that definitely makes sense to me especially if you're in a more technical domain like if your product space is in a more technical domain

I want to talk about fundamentals for a minute. I've been a strong proponent over the last 20 years of everyone learning computer science fundamentals. Things like runtime complexity, things like different data structures and algorithms, things like how database indexes work under the hood, things like that. I've always been a big proponent of that. And I don't think I am as big of a proponent of that anymore because it's so much easier to learn just-in-time fundamentals now. So I think what I would pivot to say, so I'm directly answering the question, what should I focus on learning now? I would directly pivot to breadth over depth and then just-in-time depth. Here's what I mean by that. You need to know that something exists in order to know that it's time to look into it, to go deeper. but you don't need to go super deep into it right like before you need it because now llms can teach that stuff to you you know if you're like oh i think this is a case for a red black binary tree

i've heard about red black binary trees i don't remember how they work but i think this might be one of those things hey llm teach me everything about this and show me sample code and and help me like and then quiz me and make sure i understand it right it's like that is so much more efficient now for learning these things but if you don't know to ask the question the llm won't necessarily tell you to ask it? I had a memory leak in production. I think it'd been happening for a long time, but I didn't notice it until some parameters got tweaked that then exposed it a lot more. And I have idly read a bunch of blog posts about working, I don't know, debugging memory leaks in Node. And I think I've given a talk about memory leaks in JavaScript like a decade or two

ago but i hadn't ever gone into like a prod service that was failing live with memory leak issues so i had this vague awareness of all these tools and techniques i didn't remember them all i hadn't used them all deeply and with the help of llms i built a pretty useful and convenient and powerful workflow out of all these different tools because people have spent a ton of time investing in heap dumps and profiling and like there's there's a huge ecosystem out there that i don't know very well but kind of vaguely knowing of it enough helped me dig really deeply into this problem and solve it knowing that the ecosystem exists right and like knowing that there are mature tools yeah now you know what to ask yeah heap snapshots sure yeah and i can i can then

analyze them somehow i'm sure with some tools that i can google for or ask the alum about and And then it helped me solve this problem that I probably would not have, certainly would not have solved as quickly and possibly would not have solved at all. And then would have just said, I guess our server restarts every once in a while. Like we're back to Ruby 2007 where you just restart every few hours. Every 300 requests, you must restart. Yeah, that's a really good point. So, but why it's interesting, you knew keywords and you knew about like fields of study that

existed even though you weren't necessarily an expert in those things and that's all you need now and then you just in time dive in and i thought keyword knowledge has always been valuable like knowing for example just knowing that that's called a memory leak you know just knowing that term at all helps you to now go and in the past search stack overflow or google for like memory leak tools in node.js right like things like that but now it's like probably 10 times more valuable to know those keywords because you can tell an llm i have a memory leak i need to investigate it i can give you a heap dump because i know that keyword too but show me how to do it you know and then it'll be like oh sure i can walk you through that but if you were just like hey my

server it dies every once in a while i don't know why you know the llm is going to be like well here's 25 possibilities you know that you might look into and none of them are going to be right yeah exciting times to know a little bit about a lot of things yes very true one last thing you could do is go where there's no training data or little training data so it's time to write your own programming language that's not in the corpus yes and make it different enough that it cannot be easily figured out by an llm although i think i've seen some posts about it writing brand new languages so maybe that doesn't work yeah i don't know i bet i'm trying to think of what domains would not be well covered by lms by default it means they're sort of out of the mainstream but

maybe it's time to get into radar analysis signal data i don't know whatever like web dev pretty well covered by lms yeah so if you want to do a lot by hand go somewhere where it can't easily do that by hand. All right. Have we answered the question? I think so. Good luck. Best of luck to you. Best of luck to us all. I do not know what I'll be doing a year or two from now. It's true. It's so true. I believe it'll be more different than it has been after a year or two passed. That's right. If that makes sense. What can people do if they want their own questions answered? Yes. If they want their own questions answered, go to softskills.audio and click the ask a question button where you can fill out our little form. We express our deepest heartfelt thanks to all

of those of you who submit those questions as you know this is an unreliable free tier tts service and uh we stand behind that and i think we do a pretty good job of reading the the words that you say sometimes it is lossy sometimes yeah a little bit a little bit but that's why it's a free tier but we read them all yes if you want the actual legitimately read tts go to soft skills audio and click the support on patreon button and submit the dollar amount that will get you a weekly shout out on the show and we will say your entire patreon profile name every week yep and someday they're gonna bump it up and a big post will come out saying good news you can now post megabytes of text as your patreon profile name and then oh we're in so much maybe that will kill

the podcast yes exactly it's just hours and hours but until then we remain uh diligent and dedicated to answering every single question thank you we'll catch you next week We'll be right back.