CZM Rewind: The Truth About Software Development with Carl Brown

Summary

---

Better Offline – CZM Rewind: The Truth About Software Development with Carl Brown

Date: November 26, 2025
Host: Ed Zitron (CZ Media)
Guest: Carl Brown (Veteran Software Engineer, YouTuber @ InternetOfBugs)

---

Overview

In this episode, Ed Zitron speaks with veteran software engineer Carl Brown to dig into the realities of software development, the myths perpetuated by tech industry hype, and the current and future impacts of Large Language Models (LLMs) and generative AI on software engineering. The conversation unpacks the actual work of software developers, misconceptions about AI "replacing" programmers, the practical limitations of AI-generated code, and deeper industry and management issues that are eroding long-term software quality and professional culture.

---

Key Discussion Points & Insights

What Does a Software Developer Actually Do?

• Definition and Core Tasks

  • Software developers primarily solve problems by writing code that instructs computers to perform tasks ([02:46]).
  • However, coding is a minority of the job—often between 10% and 25% of the actual work for experienced engineers ([03:25]).
    • Quote:
      “As a general rule, I would say maybe between 10% and 25%.” – Carl Brown [03:25]
  • Senior devs spend much more time communicating, understanding requirements, and planning than just coding.

• Early Career Differences

  • New graduates spend more time coding because they're handed specific, smaller tasks ([03:51]).

---

The Hype and Reality of AI and LLMs in Software Development

• Can LLMs Replace Coders?

  • At the junior level, LLMs can approximate the work by breaking tasks into small chunks, but LLMs don't improve over time, while humans do ([04:29]).

  • LLMs lack long-term thinking and context, which are crucial to real software development, which is about building systems incrementally and maintaining context over many days and iterations ([04:46]).

  • Quote:
    “But past that, it doesn’t do a good job of being able to do any kind of long-term thinking. And that’s largely the job.” – Carl Brown [04:46]

• Measuring AI’s Real Contribution

  • Claims like "30% of code at Meta/Google is written by AI" are misleading; most "AI-written" code is accepted autocomplete suggestions (typos, standard code blocks), not creative work ([05:50]).
    • Quote:
      “If all that counts as AI, then what percentage of your stuff is written by AI?” – Carl Brown [05:50]

• Problems With AI-Generated Code

  • LLMs tend to create redundant, inconsistent code, making maintenance difficult ([07:07]).
    • Example:
      • Multiple ways to fetch data from a server are generated, instead of a reusable block.
  • This leads to security headaches and debugging nightmares—finding which duplicated code block is the problem ([07:56]).

• Security Concerns

  • LLMs don’t make all code similar; their randomness means diverse implementations, sometimes increasing unpredictability ([08:11]).

• What Are LLMs Actually Good For?

  • Great for small, well-defined, self-contained tasks or when the developer needs a quick syntax reminder in an unfamiliar language ([08:52]–[09:43]).

    • Quote:
      “So they are decent at when you know what you want and what you want is a fairly simple self-contained thing... it can type it faster than you can.” – Carl Brown [08:52]

  • Not Good For:

    • Building complex systems, where context, design, maintainability, and security matter ([09:47]).
      • Quote:
        “Oh, not at all.” – Carl Brown [09:47] (on trusting LLMs to build a full package)

  • Executives vs. Engineers

    • AI tools are often being pushed by executives, not engineers, as cost-saving or “innovation” measures, regardless of actual engineering value ([10:38]).
      • Quote:
        “The executives are all really excited about it and none of us are.” – Carl Brown [10:58]

---

Software Engineering: Misconceptions and Challenges

• Misunderstood Job

  • Most of the job is understanding and defining the problem, not just coding ([12:08]).
    • Debugging and long-term thinking are core, and LLMs/GPTs cannot replicate this process.
    • Quote:
      “I talk about it like LLMs or generative AI is good at solving riddles, but actual software development is more like solving a murder.” – Carl Brown [13:08]

• The Calamity of AI-Generated Code for the Next Generation

  • Younger engineers may grow up not learning the underlying "why" and the infrastructure side of software, instead just “filling in blocks” ([13:19]–[14:48]).
    • Quote:
      “...a generation pumping the Internet and organizations with sloppier code.” – Ed Zitron [13:19]
  • Code is buggier and weirder, making debugging harder for future engineers ([15:16]).

• Hiring Crisis & Broken On-Ramp

  • Entry-level hiring is stagnating as management assumes AI can replace junior devs ([14:03]), worsening the long-term talent pipeline.

---

Empirical Evidence: Declining Code Quality

• Code Churn Up, Quality Down

  • Studies show “code churn” (lines that are quickly rewritten after being introduced) has increased since AI coding tools like GitHub Copilot proliferated ([16:32]–[17:44]).
    • Indicates more code is being written that needs immediate revision—sign of poorer quality.

• Maintenance Nightmares

  • AI-generated code encourages proliferation of inconsistent approaches to the same problems rather than refactoring and code reuse ([19:07]).

---

Inside GitHub and Modern Developer Workflow

• How developers use GitHub ([23:20])
  • Developers create and test code locally, then “push” changes to GitHub, open pull requests, and go through code review and automated checks.
  • LLM industry obfuscates these terms and workflows, contributing to hype and confusion.

---

The Devaluation of Software Engineering

• AI hype minimizes the entire discipline

  • Management treats coding as “just typing,” ignoring the critical work of design, requirements clarification, architecture, maintenance, and long-term thinking ([25:03]).
    • Quote:
      “They're devaluing all of the stuff that's not just hacking code.” – Carl Brown [25:03]

• Rise of “Vibe Coding”

  • The trend of prompting an LLM for code instead of understanding what is being built or the consequences of changes ([25:40]–[27:08]).
    • Good for prototypes or one-off tools, bad for production or code that will be maintained.
    • Quote:
      “Vibe coding is great for a thing that you’re going to do once and then throw away.” – Carl Brown [26:49]

---

Security and Quality: A Growing Disaster

• Adversarial Internet, Security Risks

  • Unlike a generated story, code placed online is attacked by malicious actors ([27:49]–[28:17]).
  • Quality assurance (QA) has been deprioritized or defunded in some large companies—creating risk ([28:33]).

• Accumulating “Scar Tissue”

  • As AI-generated code proliferates with bugs and vulnerabilities, a massive remediation effort looms in the future when the cracks start showing ([29:40]).

  • Known Risks and Secure Coding

    • Secure code requires careful control of user inputs, file access, and data permissions. AI code is likely to miss these checks, leading to exploits and data breaches ([30:55]–[31:54]).

---

Workflow and Culture: Agile, Churn and Management

• The “Agile” Shift and Short-Termism

  • Explains how Agile software management tries to manage uncertainty but encourages a focus on only short-term deliverables ([32:32]–[34:55]).
    • Leads to loss of long-term vision, planning, and architecture.

• Scaling Problems

  • As organizations grow, layers of management increase, diluting technical leadership and cross-team coordination ([36:15]–[38:06]).
    • This is as much a problem at Google/Amazon as elsewhere.
    • Recent “enshittification” (management over product/engineering) is degrading engineering culture ([37:21]).

• The AI “FOMO” Effect

  • Orgs adopt AI en masse out of fear of missing out, not because AI delivers real value ([39:12]).
    • Quote:
      “Might as well throw spaghetti at the wall and see if it sticks, because it might kind of mentality is pervasive at the moment.” – Carl Brown [39:12]

---

Management Delusions and Workplace Stress

• Layoffs, AI Productivity Promises, and Stress

  • AI is being used to justify layoffs and freeze new hires; existing engineers expected to increase output ([45:18]–[55:32]).

    • Hype justifies executive decisions from layoffs (Duolingo, Klarna, BP) to mandated AI tool use.

  • Impact on Working Engineers

    • Anxiety is rampant; even seasoned engineers wonder if their teams will be cut next ([54:34]).
    • More “tickets” (units of work) are being demanded per developer because of supposed AI efficiency ([55:32]).
      • Quote:
        “We expect more tickets closed per two week period... because we're giving you this AI now.” – Carl Brown [55:32]
    • Measurement systems (velocity graphs) now act as pressure tools.

• Executives vs. Reality

  • The people making strategic decisions about software and workforce are often not technical and don’t understand developer realities ([52:40]).

---

AI Agents Won’t Fix It — Or Replace Engineers

• Reasoning Models Haven’t Changed the Game

  • Newer “reasoning” models are mildly better (“don’t make as many stupid mistakes”) but not transformative ([49:57]).
  • AI does not possess the contextual or organizational intelligence for robust software ([46:12]).

• The Human Factor Is Irreplaceable For Now

  • AI coding tools can multiply productivity in narrow, specific cases. But building and maintaining real-world software — especially secure systems — fundamentally requires context, experience, and judgement ([57:17]–[58:19]).

    • Quote:
      “The best analogy I've always found to writing code is actually just writing, right?” – Carl Brown [57:17]

    • LLM-based “AI agents” (like Devin) introduce new risks by generating code that is not robust or secure; could expose more of the Internet to bugs and security attacks ([58:27]).

---

Advice for New Engineers

• What Should New Developers Focus On?

  • Learn to use AI-driven code editors: it’s an expected skill. Understand their strengths and limitations ([60:12]).
  • Learn to rigorously test and evaluate the software output — don’t trust autogenerated code blindly.
  • The hiring process is currently broken, dominated by bots and irrelevant interview gates.
  • Focus on skill areas AI cannot replace: long-term design, debugging, security, and context-sensitive problem-solving.
  • Use AI for disposable code or prototypes (“vibe coding”), but real, maintainable projects need intentional architecture.

• Where Can AI Actually Be Useful?

  • For tasks with little consequence or clear, testable goals (e.g., website layout, content variants, A/B testing).

  • Not for critical infrastructure or long-lived systems.

  • Quote:
    “The only people that are really going to be replaced anytime soon are people that either weren't doing a great job to start with or people whose bosses don't understand what they were doing to the point that the boss thought that what they were doing mattered.” – Carl Brown [65:20]

---

Notable Quotes & Memorable Moments

• "Being able to understand what the problem actually is and how it needs to work… all that work is basically not important [to those pushing AI tools]." – Carl Brown [25:03]
• "Actual software development is more like solving a murder." – Carl Brown [13:08]
• “A lot of this seems to be really, you know, we don't like dealing with the prima donna programmer.” – Carl Brown [48:54]
• “There's a lot of artificial productivity requirement increases... because we're giving you this AI now so you ought to be more productive.” – Carl Brown [55:32]
• "You can get ChatGPT to spit out a few paragraphs for you, right? But you end up with the legal briefs that have the story that's made up..." – Carl Brown [57:17]

---

Selected Timestamps for Key Segments

• [02:46] — What is software development, really?
• [03:25] — Only 10–25% of job is writing code
• [04:29] — Can LLMs replace coders? Why not?
• [05:50] — Myths about “30% of code being written by AI”
• [07:07] — AI causes copy/paste maintenance and debugging nightmares
• [13:08] — “Solving riddles vs. solving a murder” analogy
• [14:03] — Hiring crisis for junior devs, exacerbated by AI assumptions
• [16:32] — Code churn as a sign of declining code quality
• [25:03] — Devaluation of software engineering work by AI hype
• [26:49] — What is “vibe coding” and its limits?
• [28:33] — Companies ditching QA, adding risks
• [32:32] — How Agile shapes short-term thinking, for better or worse
• [39:12] — LLM “FOMO” and org chaos
• [45:18] — How management now plans headcount and projects with AI “savings” in mind
• [55:32] — The pressure from above: AI must increase output
• [57:17] — The analogy: code writing ≈ writing, AI ≈ basic autocomplete
• [60:12] — Advice to new engineers: learn AI but focus on depth, not just output

---

Tone & Language

The discussion maintains an accessible, frank, often skeptical tone—edged with sarcasm and dry humor. Both Ed and Carl are incisive, sometimes exasperated with hype and managerial ignorance, but rigorous about the limits and purposes of tools like LLMs in the real, messy world of software engineering.

---

Final Thoughts

Ed Zitron and Carl Brown’s deep dive is a reality check for anyone believing in the AI coding hype. While LLMs and generative tools are here to stay—and can help with specific, low-stakes tasks—the true skills of software engineers extend far beyond writing code. Context, critical thinking, design, maintenance, and especially security are irreplaceable. The episode warns that without a reckoning, management's tech-utopian dreams are likely to leave tech organizations, software culture, and ultimately all users, much worse off.

---

Guest links:

• Carl Brown: InternetOfBugs - YouTube

Host:

• Ed Zitron: Better Offline

---

Transcript

[Commercial Announcer]

This is an iHeart podcast. Guaranteed Human Black Friday at Abercrombie is here with 25 to 50% off everything. And iHeart listeners are getting an extra 15% off with code iHEARTAF. It's the sale you've been waiting for, made even better with an exclusive stackable code shop Abercrombie in the app online and in stores. 25 to 50% off everything valid in stores and online November 24, 2025 to December 1, 2025 in US and Canada. Excludes clearance and gift cards online. Price Reflects discount. Extra 15% off everything is valid through December 1, 2025 in US and Canada. Exclusions apply. See details online.

[Carl Brown]

Hey, Ryan Reynolds here wishing you a very happy half off holiday because right now Mint Mobile is offering you the gift of 50% off unlimited. To be clear, that's half price, not half the service. Mint is still premium unlimited wireless for a great price. So that means a half day.

[Sponsor Voice]

Yeah.

[Carl Brown]

Give it a [e-mail] Switch upfront payment.

[Commercial Announcer]

Of $45 for a free month plan equivalent to $15 per month required new custom for first three months only. Speed slow hacker 35 gigabytes of networks busy taxes and fees extra.

[Ed Zitron]

C mintmobile.com so as summer dies down in New York City and I get a burger from Hamburger America or ice cream from Van Leeuwen, I'm always happy to see Square as their stuff just works. And that's why I like it. And I think he will too. Square keeps you up so you don't have to slow down. Get everything you need to run and grow your business without any long term commitments. And why wait? Right now you can get up to $200 off square hardware at square.com go betteroffline that's S Q U-A-R-E.com go betteroffline run your business smarter with Square. Get started today.

[Sponsor Voice]

Running a business is hard enough. Don't make it harder with a dozen apps that don't talk to each other. One for sales, another for inventory, a separate one for accounting. That's software overload. Odoo is the all in one platform that replaces them all. CRM, Accounting, inventory, E Commerce, hr. Fully integrated, easy to use and built to grow with your thousands. Have already made the switch. Why not you try Odoo for free at o-oo.com that's odoo.com.

[Carl Brown]

Call zone media.

[Ed Zitron]

Hello and welcome to Better Offline. I'm your host Ed Zetron. Better Offline Today I'm joined by Carl Brown, a veteran Software engineer and host of the excellent YouTube channel internetofbugs. Carl, thank you for joining me.

[Carl Brown]

Thanks for having me.

[Ed Zitron]

So I'm gonna start with an easy one. What is a software developer like, what actually is that?

[Carl Brown]

So basically what we do is we take ideas about problems that people wanna solve generally, and we write software, we write code that tells computers instructions how to make the computer do the thing that needs to do to solve the problem the person asked us to solve.

[Ed Zitron]

Right.

[Carl Brown]

Gaming programming is a little bit different, but that's. Most software development is basically that.

[Ed Zitron]

And this is another quite silly question, but necessary. How much of that is actually writing code?

[Carl Brown]

It depends on how good the people that are asking for stuff is. As a general rule, I would say maybe between 10% and 25%.

[Ed Zitron]

Okay. Just really want to be 10 to 20. Even if we say 30% of the job, which is more than you said, that means the majority of this job is not actually writing code.

[Carl Brown]

Right. Now that's largely for folks that are farther up the chain. Right?

[Ed Zitron]

Right.

[Carl Brown]

So if you're fresh out of school and you don't really, you're not in the job, you don't understand how to manage requirements or any of that kind of stuff, yet someone's going to basically hand you a thing to do. And in that kind of case you're going to be spending a lot more time writing code than that. But for me, it's far, far more talking to people and stuff than actually writing code.

[Ed Zitron]

Right. The reason I ask that and the reason we're doing this as well is that there have been a lot of stories around LLMs replacing coders, LLMs replacing engineers, claiming that junior software engineers will be a thing of the past due to LLMs. How much validity is there in that?

[Carl Brown]

Well, when it comes to the really, really, really fresh out of school kids, right. That you have to basically break everything down and hand them little chunks of work. An LLM can kind of do that, although the kid will get better over time and the LLM is pretty much fixed. Right?

[Ed Zitron]

Right.

[Carl Brown]

But past that, it doesn't do a good job of being able to do any kind of long term thinking. And that's largely the job. Right. I mean, this set of, you know, I come in today, I do a thing today, I come in tomorrow having no understanding of what happened yesterday and do another self contained thing and so on and so forth. Right. That's not the job. The job is a long sequence of building up on things day after day after day after day until we get to the Point where the whole thing together works and does what it's supposed to do.

[Ed Zitron]

So I think that, I think that I've known. And one of the reasons I had you on as well is that really there are so many of these stories, they're claiming that like this software engineer's job is gone on, that these companies will be writing all of their code with AI, and it doesn't even seem like that is possible. One of your videos, you did a really good thing around like the 20 to 30%. I'll link to this in the notes, 20 to 30% of code that behind Meta and I think Google it was, is written by AI. Now again, how much validity is there to that?

[Carl Brown]

Well, I mean, so if one of the quotes was something to the effect of 30% of the code is suggestions that were given by autocomplete that a human accepted. Right, Right. Which could be as much as, you know, the, the, the thing said. Oh wait, you spelled this wrong. Let me give you a suggestion about how to spell it correctly. Right? I mean, how much of the, the actual text that you write is, you know, is corrected by a spell checker. Right. If all that counts as AI, then what percentage of your stuff is written by AI? Right?

[Ed Zitron]

Well, in my case, absolutely nothing. But that's just because I'm a freak. I'm just a complete freak. But no, I get your point. And it's. Without being a coder myself, it's something I've really noticed across these stories where people just kind of blindly push them out and they say, oh yes, 20 to 30% of the code is written by. But there's no verifying this. And also it feels like it might create a bigger problem, which is say we accept this idea, even though I don't. And it sounds like a pretty spurious one, kind of silly to do so at some point. Isn't code not just the series of things that you write to make a program work, it's connected to a bazillion other things, which if you don't know why that was written because you had something generate it, is that not a huge problem?

[Carl Brown]

Yes. But worse, what we're finding when code gets generated is that basically you end up doing the same thing in a bunch of different places, but in each one of those different places, you do it a different way.

[Ed Zitron]

Can you give me an example?

[Carl Brown]

So, for example, when you need to go fetch a thing from a server, right. Well, over here in this code, you fetch a thing from a server. Over here in the code, you fetch A different thing from the server. Normally you'd be able to use the same block of code to do that so that if there's a mistake in it, you can change it once and it's fixed everywhere. Right?

[Ed Zitron]

Right.

[Carl Brown]

But the way the LLMs work is you say, hey, I want to fetch a thing from the server and it says, cool, and it writes a whole thing for you that may or may not work the same way as the previous one. Right. So now you find, okay, under some circumstances we're having a problem fetching things from the server. I don't know which one of these 12 implementations that go fetch from the server is the one that's actually causing the problem.

[Ed Zitron]

Right. Also, isn't there a security issue of having large language models? Wouldn't all the code be quite similar or at least more similar, depending on if everyone's using Claude or everyone's using. Well, GitHub Copilot, I guess, is Claude now?

[Carl Brown]

No, not really. It basically kind of picks a random number at the beginning and goes, okay, so that's the, I think of it, kind of like you deal a deck of cards, right? Whichever deck of card gets turned over first, that's the beginning of the autocomplete that it starts. And so depending on which example it's. I don't want to say thinking of, but depending on which example represents that I'm drastically oversimplifying, but depending on which example is represented by that card, it's going to go down one path or another.

[Ed Zitron]

Right. And so what are they actually? What are these large language model coding tools actually good for? Because I get a lot of people who respond by saying this is proof that AI is a big deal. And I'm just kind of like, I'm not even looking for a particular answer, just truly what's useful about them.

[Carl Brown]

So they are decent at when you know what you want and what you want is a fairly simple self contained thing and you know how to tell whether or not the self contained thing does what you want. It can type it faster than you can.

[Ed Zitron]

Like autocorrect?

[Carl Brown]

Basically, yes. It's like autocomplete if you know exactly what you want. Yeah, I mean, so I use it a lot because I program in a bunch of different programming languages. Languages a lot. Right. On different projects at the same time or on the same day or the same week. And it's really easy for me to go, okay, wait, which language am I in right now? Okay, how do I do this in this language? Right. So it's kind of you can actually.

[Ed Zitron]

Understand the generation though, when it comes.

[Carl Brown]

Yeah, it's like, I know what kind of loop I want, but I don't remember the syntax for this particular language or I don't want to. So I use it kind of like a Google Translate kind of thing to go from one programming language to another sometimes.

[Ed Zitron]

But you wouldn't trust it to build a full software package?

[Carl Brown]

Oh, not at all.

[Ed Zitron]

Why not?

[Carl Brown]

Well, it wouldn't work to start with.

[Ed Zitron]

Why wouldn't it work?

[Carl Brown]

Well, I mean, so I've done some experimentation on that where I've taken fairly complicated challenges, challenges that were intended for programmers to basically get better at their craft and that kind of thing. And I've run AI, you know, told it step by step. Okay. The challenge says this is your next step. Do this. The challenge says this is your next step, do this. On really simple challenges in programming languages like Python, that it's got a lot, a lot of examples for. It does. Okay. Past the point where you're in the really simple kind of language things, they sometimes get to the point where they can't even create anything that builds at all.

[Ed Zitron]

Huh. Why is there, why do so many engineers swear by it then?

[Carl Brown]

Honestly, I'm not sure to what extent the engineers are swearing by it. I've talked to a lot of folks who are like, you know, my group, you know, this big bank, you know, friend of mine, my group is getting co pilot jammed on our throats whether we want it or not. And the executives are all really excited about it and none of us are interesting.

[Ed Zitron]

So it's executive. I've, I've personally had this theory that it's like executive pushed and that it's all about what the bosses want to see rather than even do. Sorry, now time to move my cat out the way.

[Carl Brown]

There's a lot of wish fulfillment. There's a lot of like, we want to not have to deal with these programmers anymore, so we would rather deal with the AI thing. And we're just going to hope that the AI thing is going to be, you know, just as good as the programmers or close to, is just as good as the programmers and not nearly as annoying.

[Ed Zitron]

Seems like a definitional. Well, maybe that's not the right word. Seems like the difference between a software engineer and a software developer almost, because it's not just about flopping code out, it's about making sure the code does stuff.

[Carl Brown]

Yeah, I mean, those terms get.

[Ed Zitron]

Mashed together.

[Carl Brown]

Yeah. I mean, so part of the problem is that I live in Texas and in Texas you're not allowed to call yourself an engineer unless you pass the engineering exam. Right. So I literally can't call myself a software engineer legally in Texas, as I understand it. I'm not a lawyer, but that's my understanding. So it's like the terms get all confused.

[Ed Zitron]

Right. So somewhat related, what is it that people misunderstand about the job then?

[Carl Brown]

Well, I mean, so one of it is what you said earlier, which is that a very small percentage of the job is actually slinging code. A lot of it is basically trying to figure out what it is the code should do based on what you've been told, that the problem, you know, the solution to the problem that you're trying to solve. Another thing is that a lot of the problem with the job is that every little decision builds up over time and at some point a bug is going to happen, they're inevitable. And when that happens, basically there's this process where what you need to do, if you're being competent, is roll back through those series of decisions, figure out what caused that bug, and then figure out what other bugs are likely to have been caused by that same set of decisions, and then fix not just the bug you're working on, but the bugs that you know, not just the bug that's been reported, but the bugs that might have also been caused by the same problem. Right, right. And that kind of long term thinking is not a thing I've ever seen LLM exhibit at all. I talk about it like LLMs or, you know, generative AI is good at solving riddles, but actual software development is more like solving a murder.

[Ed Zitron]

Yes, you said that in that wonderful video. Yeah, I. And it almost feels as if we are building towards an actual calamity of sorts, maybe not an immediate one, maybe it will be kind of sectioned off into areas because you've got a new generation of young people coming into software engineering or what have you, learning to use AI tools rather than your videos. Definitely talk about this as well, actually. How to develop software and make sure it works and make sure that it has the infrastructural side in line and also that you're building it with the long term thinking of someone else might need to understand how this works and they're not learning that. So you've just got a generation of kind of pumping the Internet and organizations with sloppier code.

[Carl Brown]

Yes, although, I mean, one of the problems we're having at the moment is that the hiring process for really junior engineers is actually pretty broken at the moment. And a lot of people are not hiring people that are fresh out of school because they're expecting that the AI will be able to do basically a senior or a mid level developer, with the benefit of AI, with the benefit of AI that's in air quotes, will be able to do the work of that person, plus a couple of fresh outs that they normally would have hired, but they're not hiring at the moment. There's some statistics about how the people that are fresh out of school these days are historically underemployed relative to the general population, at least in the US where I live.

[Ed Zitron]

It also feels like there's no intention behind the code. Like it's just, if you're just generating it, you don't really know why you made any particular. You could say I chose these lines. But is that at some point if you have large amounts of software developers using it, however large, but the young people in an organization using it to generate their code, they're neither learning to write better code nor are they learning how to develop. They're just learning how to fill in blocks they'll never grow within their job.

[Carl Brown]

Yeah, I mean, the trick is that those of us that have spent a whole lot of time debugging software, right, and like finding the problems and digging into them and trying to figure out what's going on, that kind of stuff, it's going to be really hard for younger folks to get hired into those jobs so that they have time to build the experience to be able to do that. And I'm afraid we're going to end up with basically an older generation or generations retiring and a newer generation that hasn't had the experience of doing that kind of debugging. And then it's going to be a real mess. Especially since from what I can tell, the code that the AIs generated are a lot buggier and buggier in weirder, like randomish kind of ways. Stuff just kind of comes out of nowhere in a way that I don't. I mean, I've debugged code from people that don't speak the same languages as I do and all that kind of stuff. AI code is different. It's just like, okay, why would anyone want to put that block there that doesn't have anything to do with what we're trying to do at the.

[Ed Zitron]

And why, why is that? Is it just because it's probabilistic?

[Carl Brown]

I guess so. I mean, it's hard to say why. I mean, the idea of why an LLM does what it does is kind of a, you know, anybody's guess.

[Ed Zitron]

Yeah, it's just I keep thinking of the word calamity because you sent me these studies as well about how they found like a downward pressure on the quality of code on GitHub. Would you mind walking me through what that means?

[Carl Brown]

Yeah. So basically what that study found, there have been a couple of them, but what that particular study found is that there's what they call Code churn has gone up. And Code churn is basically when you push something, you add a line of code, you push it into test or to production, and then in a short period of time, like, I don't remember exactly what the definition was, like in a month or two months, that line of code changes.

[Ed Zitron]

Right.

[Carl Brown]

So basically what that means is that the line of code that got created, somebody decided after it got put in, oh, wait, no, that doesn't work. Right. We're not happy with that. We're going to change it to be something else. Right, right. And the percentage of lines or the number of lines that get changed fairly quickly after they get submitted has gone way up since the, since the implementation of GitHub Copilot. And this is across like most of the giant, you know, millions of lines of codes on GitHub.

[Ed Zitron]

And for simpleton me, why does it being changed? Why is changing it so often bad?

[Carl Brown]

Like, well, I mean, so I mean, if you do it right the first time, you can move on to the next thing. Ah, right. If it's like, you know, if you're writing a document and you put, put the document in there and then you like, you're in, get in, you're in, in Google Docs and you're like tracking changes and it's like, okay, this sentence has changed times. Obviously the person isn't happy with the way that sentence is. Right.

[Ed Zitron]

So the generative code isn't good.

[Carl Brown]

Right.

[Ed Zitron]

And so people see need to change it.

[Carl Brown]

That's the presumption. Yes.

[Ed Zitron]

And so I. It also said the code quality itself, is that the only way they. Is that the only way they measured it or is it. There are other things as well.

[Carl Brown]

So they measured that, they measured.

[Ed Zitron]

Like moved code. Yeah, less moved code.

[Carl Brown]

The thing I was talking about earlier, where you've got a bunch of different places in the code that all do the same, try to do the same function, but they do it in different ways. Normally what would happen is you do a thing here and then at some point in the future, you need to do that thing again in a different place. What you do is you would move that original block that does the thing someplace else and then you would call that block from both place. Right.

[Ed Zitron]

Because it already works.

[Carl Brown]

Right. And then that way you've got, you know, however you go fetch stuff from the server, you're fetching it the same way but with this thing. Basically instead of doing that, you've got copy paste. Okay, let me put another one here, let me put another one here. Let me put another one here. And it's, it's a maintenance nightmare.

[Commercial Announcer]

Did you know Tide has been upgraded to provide an even better clean in cold water. Tide is specifically designed to fight any stain you throw at it. Even in cold butter. Yep. Chocolate ice cream.

[Ed Zitron]

Sure thing.

[Commercial Announcer]

Barbecue sauce. Tide's got you covered. You don't need to use warm water. Additionally, Tide pods let you confidently fight tough stains with new coldzyme technology. Just remember, if it's gotta be clean, it's gotta be Tide.

[Sponsor Voice]

Running a business is hard enough, so why make it harder? With a dozen different apps that don't talk to each other. One for sales, another for inventory, a separate one for accounting. Before you know it, you are drowning in software Instead of growing your business. This is where Odoo comes in. Odoo is the only business software you'll ever need. It's an all in one fully integrated platform that handles everything. CRM, accounting, inventory, e commerce, HR and more. No more app overload, no more juggling logins. Just one seamless system that makes work easier. And the best part, Odoo replaces multiple expensive platforms for a fraction of the cost. It's built to grow with your business, whether you are just starting out or already scaling up. Plus, it's easy to use, customizable and designed to streamline every process so you can focus on what really matters running your business. Thousands of businesses have made the switch, so why not you try Odoo for [e-mail] that's o d o o dot com.

[Ed Zitron]

Once again you're hearing me talk about quints in an ad because I like the clothes they make and the way they make them. I recently picked up a bunch of nice fitting T shirts along with a leather cafe racer jacket which fits incredibly and looks like a high end one. But it only cost 250 bucks, which quince paid part of as part of this ad.

[Carl Brown]

But I'm very happy with it.

[Ed Zitron]

I've paid significantly more for lower quality jackets. So why Quince? Well, by partnering directly with ethical factories and top artisans, Quince cuts out the middleman to deliver premium quality at half the cost of other high end brands so you can give luxury quality pieces without the luxury price tag. I was a customer before and I'll be a customer afterward. Everything I've got from quints I really really like and I look forward to getting more in the future. Give and get timeless holiday staples that last this season with quints go to quints.com better for free shipping on your order and 365 day returns. Now available in Canada too. That's Q-U-I-N-C-E.com better free shipping and 365 day returns.

[Commercial Announcer]

Quince.comBetta Time for a sofa Upgrade Introducing Anabe sofas where designer style meets budget friendly prices. Every anibe sofa is modular allowing you to rearrange your space effortlessly. Perfect for both small and large spaces. Anabe is the only machine washable sofa sofa inside and out. Say goodbye to stains and messes with liquid and stain resistant fabrics that make cleaning easy. Liquids simply slide right off. Designed for custom comfort, our high resilience foam lets you choose between a sink in feel or a supportive memory foam blend. Plus our pet friendly stain resistant fabrics ensure your sofa stays beautiful for years. Don't compromise quality for price. Visit washablesofas.com to upgrade your living space. Today. Sofas start at just $699 with no risk returns and a 30 day money back guarantee. Get early access to Black Friday now. The biggest sale of the year can save you up to 60 off plus free shipping and free returns. Shop now at washablesofas.com Offers are subject to change and certain restrictions may apply.

[Ed Zitron]

So for the, for the audience as well, how does a software developer actually use GitHub? Like really simple stuff I realized, but I think it's important for people to it just occurred to me that this may be something that most listeners don't know. Which is good. I think it's good.

[Carl Brown]

Yeah. So what we do is we basically make changes to code. We get to the point where we, the developer are happy with the way it's set up on our machine and then we do what's called a push and we basically send all that code, submit all that code up to GitHub, and then theoretically there can be automatic processes that kick in that check that code for particular things and run tests on it and that kind of stuff. And then at some point we have a thing called a pull request, which is basically a thing that says okay, I would like this to go into production now, or more or less I would like this to get promoted into the next phase now. And then someone theoretically will look at it and go, okay, that's fine. And then click the yes button or say, hey, you forgot about this. Go look at this. Or that kind of thing. Right? And the pull request is kind of the unit of work, kind of.

[Ed Zitron]

So with GitHub, you almost use it like an organizational code dump or where you centralize all the code. Sorry, just for the. For me and non coding as well. And I think it's. I think that the LLM industry has done a really good job of dancing around these terms and selling them to people like me. While they weren't selling. They didn't work on me. I too stupid. But it's where they've just like been like, okay, yeah, well, lots of people use copilot. That's good. And this is good because software's coding. But it kind of feels like, I don't know, all of this is taking the one thing like one major part out of software development and ruining it. And I don't even mean coding. I mean it's the intentionality behind software design and infrastructure and maintain. Like there's. It seems like they're removing intention in multiple parts.

[Carl Brown]

So the way I would say it is when they talk about the AI being able to do the work of a programmer, what they're doing is they're devaluing all of the stuff that's not just hacking code. Right. And so what they're saying is that basically the job of a developer is basically just typing, basically. And that all of the work that we do to understand what the problem actually is and how it needs to work and what other problems are likely to show up when we try to do that and how to avoid those things as we go and that kind of thing, all that work is basically not important.

[Ed Zitron]

And I mean, I'm going to say two words which will probably annoy you. This is. I feel like vibe coding is the other part of this. So if I'm correct, correct me if I'm wrong. Vibe coding is just typing stuff into an LLM and software comes out and hopefully it works.

[Carl Brown]

Yeah, Vibe coding is basically when you intentionally try. Well, I don't know about intentionally, but basically you make a point of not digging into the code and looking at what the LLM is doing and you basically say, okay, I would like something that does X, right? I would like a, a game where I fly airplanes around a city or something. Right? And then you get what it spits out and then you say, you know, okay, let me try it. Okay, well, can we have more airplanes? And. Okay, can we have some balloons with, you know, signs on them now? And can we do this kind of thing? And then you don't think about what the side effects are. You don't think about what things could go wrong. You don't think about air conditions, that kind of stuff. And you just hope that this, whatever you look at and has the right vibe and that, you know, if it. If it. If it looks like kind of what you wanted, that probably it's going to be fine or hopefully it's going to be fine.

[Ed Zitron]

How do you feel about vibe coding?

[Carl Brown]

So I do it. Sometimes vibe coding is great for a thing that you're going to do once and then throw away.

[Ed Zitron]

Yeah.

[Carl Brown]

Right. So if it's like, you know, okay, I want to do a thing, I want to translate this thing to, you know, I want to make this table go into this format over here or that kind of thing. You do it, you get the output you want, you throw the code away. No big deal. Right.

[Ed Zitron]

Like a prototype almost.

[Carl Brown]

Yeah, basically. And so we call them spikes or tracer bullets. Sometimes it's like, let me get a thing that works at all. Right. And then let me see what I can learn from that to move into my big maintainable project. But for anything that's like, this thing needs to run for a while. This thing needs to not get hacked. This thing needs to not crash. It's a really bad idea.

[Ed Zitron]

Yeah. And at some point, I feel like someone building a product that they don't really understand the workings of, it's kind of almost identical to generating a story with ChatGPT, except kind of more complex and more prone to errors.

[Carl Brown]

Yeah. And the other thing is that there's an adversarial component. Right. How so people will intentionally try to go hack that thing that's sitting on the Internet.

[Ed Zitron]

Oh, right.

[Carl Brown]

In a way that they don't intentionally try to go mess with the story that you wrote. Right, right. And so even if it works all by itself, that doesn't mean it's going to work when somebody starts pounding on it intentionally trying to break it. And if they can break it, then that's a whole other set of problems that you now have.

[Ed Zitron]

Feels like quality assurance is just never part. Oh, no. Are they claiming they're going to do quality assurance with large language models yet? They must.

[Carl Brown]

Some people are, yeah. I mean, to be honest, a lot of companies have just started been getting rid of quality assurance over the years. Right.

[Ed Zitron]

Oh really?

[Carl Brown]

When I worked at IBM we didn't have quality assurance at all. They would. No, seriously, they would do this. I was in IBM's Cloud Group and they would do these, these, what do they call them, Hackathon kind of things. They didn't call them that. I don't know what they called it. But basically everybody in all the other development groups would get together and basically bang on the code that was about to get released from some other group to try to see if they could break it. Right. But they didn't have dedicated testers anymore because they decided, I guess, that they didn't think they were worth the money. I don't know. But we had some issues because of that.

[Ed Zitron]

When did that movement happen?

[Carl Brown]

I was in. I don't know. So I was at IBM in like 2017, 2018.

[Ed Zitron]

Right.

[Carl Brown]

So it would have been sometime prior to that. When I got there, they didn't have any QA folks really.

[Ed Zitron]

Just feels like the. It's the management problem as well. It's the management guiding people, I would think. So it's a real shame as well. And forgive me if I'm forgetting exactly where you've mentioned as well that there is like compound scar tissue from AI generated code. A larger problem of lots of this code being generated with AI.

[Carl Brown]

Well, that's my expectation. Right?

[Ed Zitron]

Yeah. Just a potential worry.

[Carl Brown]

Right, Right. So the more of this we get and the more issues that we have, the more stuff we're going to have to dig out of. Right. And what I'm honest envisioning at some point in the. I don't know how long this will take. The crypto bubble took way longer to pop than I expected. So I don't know how long it's going to be before this one does, but I'm expecting that there's going to be this big push to try to clean up a bunch of this crap here in a few years once people realize that a lot of the code that's being written and generated right now has all of these vulnerabilities that nobody's bothering to check for at the moment.

[Ed Zitron]

Right. And those vulnerabilities again, non technical way. I read that it was like they call upon things on GitHub that don't exist. So bad actors create something that resembles what it's pulling from.

[Carl Brown]

So that's a more specific kind of one. I mean there are a lot of things. I mean, so there have been computer viruses since the 80s, right. The Morris worm and that kind of stuff. And basically there are known Ways that code, you have to write it in a particular way in order for it to be secure. Right? Right. And even then, sometimes people come up with novel ways of making something not secure.

[Ed Zitron]

And how do you have to write it to make it secure if it's possible to explain?

[Carl Brown]

Well, I mean, there's a big long list of rules, right? I mean, one thing you can do is you can use languages that are what they call safer. But still you have to make sure that any input that you get from the network, you're really, really careful to make sure that it doesn't get to overwrite parts of your program that actually execute things.

[Ed Zitron]

Right.

[Carl Brown]

You have to make sure that it doesn't have the opportunity to be able to write to places on your disk that it shouldn't be able to write to. You have to be able to make sure that it doesn't have access to read data that it shouldn't be able to read. You know, all that kind of stuff. And when those things don't happen, you end up with, you know, so and so got hacked. You know, turns out that somebody, we think maybe China is reading the email of the, you know, people in Congress. You get another letter in the mail that says your Social Security number has been, you know, leaked by, you know, some credit checking firm or something like that.

[Ed Zitron]

Even, even like I think it was what the big target data breach from a while back was through the H vac system. It's just. Except now we've got. And that was with humans writing the code, right? Imagine if we didn't know. Oh God, it really does feel like the, the young people are going like the. Actually now, I take it back, you were talking about Agile the other day. I'm going to ask you to explain that in a second. But it sounds like for almost decades they've been gnawing away at, management's been gnawing away at the sides of building good software and building good software culture.

[Carl Brown]

Yes. I mean, there's an argument that says we never got it right in the first place. But I mean, if you think about it, software has been a thing for what, 50 years, 60 years? 70 years. Right. I mean, compare that to like construction engineering or bridge building or that kind of stuff. Right. We're still, relatively speaking, in our infancy as a industry. It's been a constant evolution and a lot of times the things that we did to solve a problem that we had ended up causing other problems. Right. So going back to Agile, in the long, long ago we used to manage software projects, the Same way we managed like build bridge building and building, building progress, construction projects. And it turns out that when you're going to build a bridge, you know beforehand what you need to build the bridge to do. When you're building software, a lot of times people are changing their minds as you go, right? And you build a thing and you show it to them. They're like, oh, why don't we put this over here and why don't we change this and that kind of thing? Right, right. Because you don't have the same kind of constraints, physical constraints, that you do when you're trying to build a bridge. And so we got in this problem where you would create these project plans about how you were going to build this thing and you would never be anywhere close to on time because things would change. The who time. And so they created this thing called the Agile methodology. I'm drastically simplifying. There were steps in the middle, but basically. So this agile thing is where we, instead of saying, okay, so this is what the whole project's going to look like. We're going to be done in six months. And then things changing along the way, we basically block off a thing called a sprint. It's a week or two or a month maybe, it depends. And then everybody picks their own sprint length. And then you go, okay, I'm only going to talk about what's going to happen in the next sprint or two. And then you get to the end of that two weeks and you go, okay, cool, this is what we got done. What do we want to do next? And then, okay, that's what we got going. What do we want to do next? And that kind of thing. And that way, as you go, you have the opportunity to change things, you have opportunity to roll changes into the process, that kind of thing. Right. The problem with that is kind of the same way that dates always ran out in waterfall land. Projects can go way, way longer than they were expected to at the beginning because everybody's focused on just two weeks at a time. And you never kind of take a big step back like you ought to and go, okay, wait, we were supposed to be done two months ago. When are we going to wrap this up?

[Ed Zitron]

Right, and how has that led to things getting worse? Is it that just software culture? Software development culture has been focused on short term perpetually.

[Carl Brown]

The short term is part of it. Part of it is there are unscrupulous developers out there that basically want to extend the length of the project so they can get more money out of it.

[Ed Zitron]

Right, right.

[Carl Brown]

That's always the case. But the other thing is that you end up with a real. A lot of times you end up with a real lack of long term planning and long term understanding because everybody's same kind of thing. Companies are only worried about what happens next quarter. If you're only worried about what's going to happen the next week or the next four weeks, the things that you look at tend not to have the longer term implications that sometimes you need. Right. And there are times you get close to the end and you're like, oh, we didn't think about this, what do we do?

[Ed Zitron]

And also if you're in a two or three week thing, you're probably not thinking even what you did last. Sprint.

[Carl Brown]

Maybe last one, but not two or three, two or three ones ago.

[Ed Zitron]

Is this a problem throughout organizations of all sizes? Is this a consultancy problem? Is it everywhere?

[Carl Brown]

It's most places. There are some places that are usually in startups. We're a lot more ad hoc and we're a lot more focused on trying to get things done. Basically the idea is the larger you get as an organization and the more money you're throwing at it and the more management control you want, the more of this overhead you put in place and the more complicated things get. Just as a management structure kind of thing.

[Ed Zitron]

And in the big. So this is something you'd seen like a Google and an Amazon as well?

[Carl Brown]

Oh, absolutely.

[Ed Zitron]

So do you think it has the same organizational effects or.

[Carl Brown]

Largely. Yes. So those organizations tend to be. Well, those organizations historically have tended to be before the recent inshidification wave. Those, I'm assuming. I can swear on this.

[Ed Zitron]

Yeah, yeah, yeah.

[Sponsor Voice]

Fuck.

[Carl Brown]

Shit. It's all fine. Those organizations have historically been fairly more engineering driven, which means that you have. You typically have people higher in the organization that are technical and have been programmers and who understand some of the implications. And so they tend to try, at least we try, to run interference with management and to try to make sure everybody's on the same page and that kind of stuff. A lot. Not all, but a lot of problems can get lessened if you have people in the organization that are at higher level whose job is not to manage people, but whose job is basically to keep track and coordinate. Coordinate between different groups that are doing different technical things.

[Ed Zitron]

Right. To make sure people aren't building the same thing. I'm guessing. Or are building the right thing in the right way.

[Carl Brown]

Yeah. And that what this group is building is going to impact what this group is building at some point in the future. And making sure that when you get to the point where those two things need to talk to each other, they're both aware enough of what the other one is doing that the two things hook together correctly.

[Ed Zitron]

Yes. So based on my analyses of these companies, that's definitely gone out there the window. I mean, even with LLM integration. So there was a Johnson and Johnson story that went out Wall Street Journal a couple weeks ago where it was like they had 890 LLM generative AI projects of which, like the Pareto principle wins again, 10, 15% of them were actually useful. And the thing that stunned me about that, other than the fact it confirmed my biases, which I love, was the fact that there were 890 of the fucking things and no one was like, should we have this many? That there was no software engineering culture that was like, hey, are we all chasing our tails? Is this useless? But it sounds like they were all focused on their little boxes.

[Sponsor Voice]

Yeah.

[Carl Brown]

I mean, so the other thing. So understand that. Again, greatly oversimplifying a lot of the new stuff that's happened with large language machines, large language models and, and generative AI people didn't expect. Right. It was kind of a surprise when you throw a whole bunch more data at a large language model and it started spitting out text in a way that nobody really. There was no like, mathematical reason to expect it to be able to be as good at generating autocomplete stuff as it was. Right. And so there's this belief that if we did the thing and we unexpectedly got more than we asked for, if we do more of the thing, maybe we'll unexpectedly get more of what we wanted. Right. That. That hasn't seemed to really pan out the last couple of years from what I can see. But that we don't really understand enough about this to know whether it's going to work. So we might as well throw spaghetti at the wall and see if it sticks, because it might kind of mentality is kind of pervasive at the moment. And there's a lot of fomo. There's a lot of like, well, our competitors are probably doing this and so we don't want to get left behind. It kind of reminds me of the rumors that they talked about back in the 80s when the C was doing all this psychic research, because supposedly the Russians were doing psychic research and it was all complete crap. But both sides were convinced that the other side was making some progress and so everybody was dumping a ton of money into it.

[Ed Zitron]

LL MK Ultra.

[Carl Brown]

Exactly.

[Ed Zitron]

Yes.

[Carl Brown]

Title of the episode.

[Sponsor Voice]

Running a business is hard enough so why make it harder with a dozen different apps that don't talk to each other. One for sales, another for inventory, a separate one for accounting. Before you know it, you are drowning in software Instead of growing your business. This is where Odoo comes in. Odoo is the only business software you'll ever need. It's an all in one fully integrated platform that handles everything CRM, accounting, inventory, E commerce, HR and more. No more app overload, no more juggling logins, just seamless system that makes work easier. And the best part? Odoo replaces multiple expensive platforms for a fraction of the cost. It's built to grow with your business whether you are just starting out or already scaling up. Plus it's easy to use, customizable and designed to streamline every process so you can focus on what really matters running your business. Thousands of businesses have made the switch, so why not you? Try Odoo for [e-mail] that's o d.

[Commercial Announcer]

O o.com time for a Sofa Upgrade Introducing Anabe sofas where designer style meets budget friendly prices. Every anibe sofa is modular allowing you to rearrange your space effortlessly. Perfect for both small and large spaces, Anabe is the only machine washable sofa inside and out. Say goodbye to stains and messes with liquid and stain resistant fabrics that make cleaning easy. Liquid simply slice slide right off. Designed for custom comfort, our high resilience foam lets you choose between a sink in feel or a supportive memory foam blend. Plus our pet friendly stain resistant fabrics ensure your sofa stays beautiful for years. Don't compromise quality for price. Visit washablesofas.com to upgrade your living space. Today sofas start at just $699 with no risk returns and a 30 day money back guarantee. Get early access Access to Black Friday now. The biggest sale of the year can save you up to 60% off plus free shipping and free returns. Shop now at washablesofas.com offers are subject to change and certain restrictions may apply. The wait is over. The NYX Black Friday sale is on [e-mail] shop early and save up to 60% off site wide plus all kinds of limited time daily deals from the number one leak proof brand in North America. Don't miss your chance to save big on innovative intimates like leak proof underwear, wireless bras, shapewear and more. Everything is on sale. Millions have made the switch to NYX's revolutionary period underwear and there's never been a better time for you to try them too. During the Black Friday sale, save up to 60% on super comfy machine, washable and stylish leak proof undies. Plus shop the best deals of the year on NYX's best selling assortment of wireless bras. Don't miss this chance to stock up on your Knicks favorites or try something new during the Knicks Black Friday event. That's knix.com the sale ends December 2nd and sizes will sell out. Go to knicks.com that's knix.com as New.

[Ed Zitron]

York cools off and I begin actually going outside again, I'm always happy to pay for a sandwich at Court Street Grocers or a coffee from Birdie in Williamsburg. With Square, the tech just works. It's easy to set up for businesses, giving them instant access to earnings. With Square Banking built in marketing and loyalty tools and powers over 100,000 different food and beverage sellers from boutique bakeries to national coffee chains. It's incredibly easy to use for customers and businesses alike. And you don't need any special training or elaborate setup. And in my personal experience, I don't think I've ever had trouble with a square terminal. You just tap it and go. Or you slide a credit card in in person, it works. Online it works. I just really like Square and I'm happy that they're advertising with the shop. Square keeps you up so you don't have to slow down. Get everything you need to run and grow your business without any long term commitments. And why wait? Right now you can get up to $200 off square hardware at square.com go betteroffline that's S Q U-A-R-E.com go betteroffline run your business smarter with Square. Get started today. So, okay, LMK Ultra aside, is this something you're seeing in software development though? Because I know I've seen it in management where it's just gonna like shove the shit in there. This seems like it's an important thing, right? Or is this. Are you seeing it within software development?

[Carl Brown]

So I am seeing it within software planning, right? So when, when managers are sitting down and saying, okay, we need to build this new thing, we need to create a new group, we need to split this group apart, we need to decide what our headcount's gonna be for next year. There's a lot of okay, and what do we think the AI is going to do next year and how many headcount do we think that's going to save us and that kind of thing, right? There are some companies duolingo is one, Klarna is one. Bp, the former British Petroleum, what last year had a thing where they said they were cutting 70% of their contract software developers.

[Ed Zitron]

And in most of these they've kind of rolled them back as well.

[Carl Brown]

I don't think Duolingo has yet.

[Ed Zitron]

This is just being unfair to you. They like a day ago, really just like, ah, we're kind of. It's so fun. It's so funny. It's so funny that this just keeps happening in exactly the same way. It's like, oh, what a surprise. Human beings do stuff.

[Carl Brown]

Yeah.

[Ed Zitron]

But it kind of gets back to, I think what you've said about everything with LLMs. It's like you can teach something to say. Yeah, I think the right. The thing you're looking for is this, but you can't teach it context. And that's been a point you've made again and again. Like it seems the job of a software engineer is, is highly contextual unless you're like in the earlier days.

[Carl Brown]

Yeah. And I liken it sometimes to the memento guy from the Memento movie. Right. Where like can't form long term memories. And do you really want the memento guy to be the person that's building the software that makes the 737 Max be able to compensate for its control input?

[Ed Zitron]

Yeah. The thing is though, with that argument, they would argue, and I know that there is a better argument here, they would argue, well, what if we just give it everything that's ever happened? What if we just show every single thing we've ever done on GitHub, surely then it would understand.

[Carl Brown]

So what I have seen from the papers that I have read is that LLMs have a basically squishy middle context problem kind of the way that you do. Right. So if somebody gives you a big document to read or a big long documentary to watch or something and then they ask you questions, what they're going to find is that you remember, remember a lot more from the beginning of that and the end of that than you do from the middle of that. Right. And LLMs have the same kind of problem. Right. And the other problem that the LLMs seem to have is that when you give them a whole bunch of instructions, just instructions piled on instructions piled on instructions, they can either get confused and forget some of the instructions or they deadlock or they just start going, okay, I can't satisfy all of these, so I'm not even going to bother to satisfy any of them. Or they'll pick one or two. The fact that you can take a million tokens and you can stick that in the memory block that the GPU is going to process doesn't necessarily mean that all of the tokens in that memory block are actually going to be treated equally and going to be understood. Right. In theory, maybe if you could train your. If you could like custom train an LLM and modify all of its weights based on exactly what your stuff was and do that like day after day after day after day as things changed, you would theoretically get better. I still don't think it would be, you know, I still don't think it would understand the context as well, but that would be ridiculously expensive.

[Ed Zitron]

Yeah. And at that point you could train a person.

[Carl Brown]

Yes. I mean, the person would probably be more annoying. I mean, the point. I mean, a lot of this seems to be really, you know, we don't like dealing with the prima donna programmer kind of thing. Right. There's this, you know, I mean, not just programmers. Right. We don't also don't want to deal with the prima donna reporters or the prima donna illustrators or the prima donna.

[Ed Zitron]

Get rid of these people.

[Carl Brown]

Right.

[Ed Zitron]

He's annoying. They ask for stuff. They want money.

[Carl Brown]

Yeah. And days off and sick leave and, you know, health care and this is disgusting. How dare It's.

[Ed Zitron]

It's so. It's frustrating as well because it across software development and everything, but especially with software developers, feels just very insulting because it doesn't seem like this stuff. Actually, here's a better question. Have you seen much of an improvement with like 0103, like these reasoning? Do you think the reasoning models change things for the better? And if so, how so?

[Carl Brown]

A little. They don't make as many stupid mistakes is basically what it boils down to. Going back to your first thing though, right? I mean, so there was a piece that actually a couple pieces recently, one of them was about tech workers are just like the rest of us. They're miserable. I'll give you links to these. The other one was a Cory Doctorow piece that was like, the future of Amazon coders is the present of Amazon warehouse workers, or vice versa. There has been a lot of deference given to software developers over time because we have been kind of the engine that's made a lot of the last 20, 30 years work. And there's a desire to make that not so anymore. Right. And to make us just as interchangeable as everybody else. I guess from a economic standpoint, I kind of don't blame them. I understand why they're trying to do what they're doing. I mean, I don't think that the warehouse workers should be treated the way the warehouse workers are treated. Much less everybody else gets treated that way. Way. And it's been a lot worse since the giant layoffs at Twitter. Now X when that happened and the thing didn't crash and completely burn like everybody was, or not everybody, but a lot of people were expecting it to, the sentiment became, well, maybe all these software developers aren't as important as they, you know, we've always thought they were. And you know, we will see over time what the end result of that is. My guess is it's going to end up being a mess. But you know, I'm a software developer, right? I'm gonna, it behooves me for it to be a mess, right? So it might just be my bias that's getting in the way.

[Ed Zitron]

I actually, I think that you're right though because I remember back in 2021 and onwards the kind of post remote work, the remote work, there was the whole anti remote work push but there was this, the whole quiet quitting and things like that. 2022. Whereas like software engineering, they're just, they're, they expect to be treated so well because 2021 saw the insane hiring, right? You saw tech companies like parking software workers. I think that played into it as well where all of these companies who chose to pay these software engineers, they, they were the ones that made the offer offers, got pissed off that they'd done so and thought we should cut all labor down to size and then along comes coding. Almost makes me wonder if most of these venture capitalists talking about this don't know how to code themselves. You gotta wonder.

[Carl Brown]

I don't know many that do. Yeah, I know some that have at some point, but that's the thing, at.

[Ed Zitron]

Some point it's like they're not part of modern software development culture, which I know sounds kind of of wanky but I mean just how an organization builds software feels like something they should know. But then again they don't know how to build a real organization either. So who the fuck.

[Carl Brown]

Yeah, well, I mean, honestly a lot of it I've been in organizations that VCs basically killed because you know, we built a thing, that thing was, you know, a reasonable business. But VCs don't want a reasonable business. They want either 100x return or they want a tax write off and they don't want anything in between. Right?

[Ed Zitron]

Yeah.

[Carl Brown]

So I mean what, what they're looking for is really, I Mean, they're not trying to run a regular business, right? They're not trying to do the normal process. They're trying to, to either, you know, hit one out of the park or throw it away and move on. And so they're, they're, the rules for them are different because what they're trying to accomplish is not what the rest of us are trying to accomplish as a general rule.

[Ed Zitron]

Constant theme of the fucking show. It's just like, it's just like you have these people that don't code saying how coders should code, like Dario Amade the other day saying that this year we're going to have the first one person software company with a billion dollars revenue or something like that. And it's just, I feel like there are some people who should not be allowed to speak as much sometimes. But it's just frustrating and insulting and it's. But now that you've got me thinking about it, it does feel like this is an attempt to reset the labor market finally coming for software developers. And I don't mean finally in a good way way.

[Carl Brown]

Right. I mean it feels like that being in the, being in that organ, being in that industry at the moment, it really feels like that.

[Ed Zitron]

Is it scary right now? Is it scary right now?

[Carl Brown]

Not for me, because I'm old enough to be semi retired. Right, Right. But I mean, I've been talking to a lot of folks, I've been having a. Interviewing a bunch of folks that are, that are listeners from my channel and kind of trying to get a feel for what's going on. And I've talked to folks that are, like I said, I talked to some folks that were like, I work for a big bank. They're cramming copilot down our throat whether we want it or not. I've talked to some folks that are like, every time I sit down with my boss, I'm thinking that this is going to be the day that I'm going to find out that my group is getting cut the way the other three groups in the company is getting cut. There's a lot of artificial productivity requirement increases kind of thing which is like want just, you know, we, you know, we expect more tickets closed per two week period than, you know, we've had before because we're giving you this AI now so you ought to be more productive, that kind of thing.

[Ed Zitron]

Would a ticket necessarily be something that you just write code for or is it more than just that?

[Carl Brown]

Well, so generally it's more than just that, but generally the ticket that's kind of the way that we track the work that we do in a lot of organizations. Right. And some tickets are like, I'm building a new thing. And those are kind of easier to predict. And some tickets are, this thing isn't behaving right. Go figure out where the bug is. And those are a lot harder to predict. But they have these things. Agile has this thing called a velocity graph where basically you see how many tickets per person get closed over time and people want to see the slope of that line change because they're giving you AI.

[Ed Zitron]

And I'm guessing the people telling you to change that don't know what they're talking about.

[Carl Brown]

About that seems to be the case.

[Sponsor Voice]

Great.

[Carl Brown]

The good news, in theory, right? I don't know to what extent this is going to happen, but in theory, if they keep telling people the slope of that line should be changing because you have AI now over time, if we see the slope of that line not changing, then theoretically it will be proof that the AI is not providing the return that people expected.

[Ed Zitron]

Or you're not using it, right?

[Commercial Announcer]

Right.

[Carl Brown]

Well, yes, there's always that you're not prompting it, right.

[Ed Zitron]

That is. That is basically what I am. People, the con. One of the many reasons I want you on is like, I want to have people that actually code on to talk about this stuff because it's really easy as a layman myself and for others to just be like, oh, but this does replace coding, right? And it does. It sounds like it really doesn't. Like it can help. It can be like a force multiplier to an extent, but. But even past the initial steps, it just isn't there.

[Carl Brown]

Well, I mean, so the best analogy I've always found to writing code is actually just writing, right? I mean, you can get ChatGPT to spit out a few paragraphs for you, right? But you end up with the legal briefs that have the story that's made up or the just things that aren't connected to reality or stuff that when people read them, they're like, I mean, you can tell the difference between AI slop generated, you know, like the stupid, the insert from the Chicago Suntime, the Philadelphia Inquirer, all the things you can do this summer, right, that like made up books and all that kind of. I mean, like. But even the articles that weren't the ones that were making up stuff, you read the, you know, this is what's going to be happening this summer. This is what the weather's going to be like or whatever. And you're reading and you're like, like this, this. There's no, like, insight here. There's no thought here. There's no, you know, there's nothing in here that I get to the end of this. I've read the whole thing, I understand the whole thing, but I don't have anything I can walk away with.

[Ed Zitron]

Right. And AI agents aren't coming along to replace software. You're not scared of Devin?

[Carl Brown]

I am not scared of Devin. So I. Well, actually I kind of am. I am scared that Devin is going to make a mess of things and then more things are going to get hacked and that's going to end up being worse for everybod on the Internet. Right?

[Ed Zitron]

How would it do that?

[Carl Brown]

Like we were talking about before, right? So when you write code that isn't secure, right? And you write code that uses a library that's got an old version of a thing, that there's a known bug in it, but you don't bother to check to see if there's a fix for that bug, or you don't use best practices when it comes to writing code and that kind of thing, or you don't think about the kinds of maintainability issues that you're going to have. And you do things like you ship out code in an Internet of Things thing, a light bulb, right. Or a Internet WI fi router that cannot be patched over the Internet that has a bug in it, right? And now it's like that thing is going to have a bug in it forever and you're going to have to find all of the ones on the, on the earth and turn them off before someone's not going to be able to take them and be able to hack them and use them to attack somebody else from there.

[Ed Zitron]

I mean, Iot is a huge problem.

[Carl Brown]

Oh yeah.

[Ed Zitron]

But the cheaper ones have like the spyware stuff and crypto mining it just.

[Carl Brown]

But yeah, the ones, the ones that have, they have like really nasty vulnerabilities and they have no way of being updated once they leave the factory, Right? And it's just as long as they're out there, they're going to be a problem literally for everybody on the Internet.

[Sponsor Voice]

Jesus.

[Ed Zitron]

Well, what can. To wrap us up, what can a new engineer, someone new to software development, what can they learn? Right now you've kind of done a video on this, but I think it's a good place to wrap up us up. What can they start learning to actually get ahead, to actually prepare for all of this?

[Carl Brown]

That's a really good question. So you can't. These days, you can't really be able to be an engineer. You can't get hired as an engineer without some ability to talk about being able to do prompts and use some kind of AI code editor or that kind of thing. It's just an expectation of the job. Now, whether it should be or not is a different thing. I mean, like I said before, there are situations where you tell it what you want and it will type faster than you possibly can. So that's not necessarily bad. You need to understand that. You need to figure out, well, okay, I'll get back to something else. You need to figure out basically how to test the thing, right? So how do you make sure that the code that it spits out does what you meant it to do? And what I'm expecting is that we're going to spend more time thinking about testing and thinking more about trying to find exceptions and that kind of thing than we have in the past, because the code that's actually being generated is going to be less likely to be quality than it was in the past. The problem is it has become the case in the programming industry that the things you need to do to get through the interview to get hired have very little resemblance to the things that you actually do on the job, that you need to actually do a good job. And so that's a whole different. We could probably have a whole other podcast episode just about the interviewing problem. But. But the main thing right now, so right now, the whole hiring thing, and this is, I don't think true for just programmers, but it's especially true for programmers, is all bots that customize your resume and write a custom cover letter and then send them over to submit the thing to the bot that's screening the resume and screening the COVID letter. Right. And that getting it to the point where you can actually talk to a human is a nightmare right now. So the whole hiring system is kind of broken. So the actually getting to the point where you can get hired is a nightmare at the moment. But the thing that you can do is figure out what kinds of things that AI are good at is good at. And one of the things that AI is pretty good at is things that don't matter as much, right? So like AI can pick the layout of a site, potentially, right. And you could have it pick two or three of them, and you can basically do what's called an A B test, and you can randomly assign people to it. You can figure out which one of them performs better, and you can throw the rest of them away.

[Ed Zitron]

And even then, at some point, you will probably want the design customized.

[Carl Brown]

Yeah. I mean, but.

[Ed Zitron]

But.

[Carl Brown]

I think there will be a lot of things where people can kind of get something that's kind of good enough to get started. Right, right. And I think that to some extent this is going to be kind of a boon for the industry in the longer term, where somebody who can't program right now, but who has some idea of kind of what they want, can do like a vibe cod. They can validate that the market that they want to try to attack exists. Right. And that people want to use the kind of thing that they built and then they can bring in somebody to actually build it. Right. You know what I mean? And those kinds of things wouldn't necessarily have been able to happen in the complete absence of AI. So it's not, I don't think, completely useless. And there are times when as a developer, there are things that we're not good at, like writing marketing copy and that kind of stuff that if we're trying to do a project for ourselves, you know, a lot of that stuff we can just outsource to the AI because it's not the thing that keeps the project from actually breaking and getting hacked and that kind of thing. Right. So it's kind of like there's this concept where you need to keep the things that are part of your competitive advantage in house and everything else you can kind of outsource to somebody else. The kinds of things you can outsource to somebody else are the kinds of things that you potentially you could throw an AI at because they're not.

[Ed Zitron]

But even then, it's like, it doesn't seem like that's a ton of things right now or will be again.

[Carl Brown]

It's basically two things. It's things to where the quality of the thing doesn't matter really, which every business has those kinds of things. Right. And they're the kinds of things where you can define a metric that you can test the AI against and let it try over and over and over and over and over again until it gets to the point where it's good enough. So if your metric is more people. People click on this button than the button before. Right. Then you can have the AI create a whole bunch of different ways to skin that button. And then you can say, okay, so the one that tested best is the one we're going to keep. That's the thing you can throw an AI at. Right. Because you've got a well defined way of checking and no telling how long it's going to take. But you have a well defined way of checking to see if it's working right or not.

[Ed Zitron]

Yeah. I mean, for years I've had the theory that this industry was a 20 to 25 billion dollars total addressable market pretending to be a trillion dollar one. And everything you're saying really suggests, it's like you're describing things like platform as a service, things that you use in tandem with very real people in intentional ideas.

[Carl Brown]

Yeah, this is, I don't see a world in which this is a, we replace all the humans. You know, the whole like, you know, know this is going to displace 80% of the white collar workers in the world. I just, you know, the only, the only people that are really going to be replaced anytime soon are people that either weren't doing a great job to start with or people whose bosses don't understand what they were doing to the point that the boss thought that what they were doing mattered. And my guess is that there's going to be regret at that point and that at some point they're going to have to bring those people back.

[Ed Zitron]

Well, Carl, this has been such a wonderful conversation. Where can people find you?

[Carl Brown]

I am. InternetOfBugsoutube is probably the easiest place to find me. And then there are links on that channel to point at other things.

[Ed Zitron]

And you've been listening to me at Zitron. You've been listening to Better Offline. Thank you everyone for listening and yeah, we'll catch you next week. Thank you for listening to Better Offline. The editor and composer of the Better Offline theme song is Matosowski. You can check out more of his music and audio [e-mail] m a t t o s o w s k-I.com you can email me at ez betteroffline.com or visit betteroffline.com to find more podcast links and of course my new newsletter. I also really recommend you go to chat where's your ed to visit the Discord and go to R betteroffline to check out our Reddit. Thank you so much for listening.

[Carl Brown]

Better Offline is a production of Cool Zone Media.

[Ed Zitron]

For more from Cool Zone Media, Visit our website coolzonemedia.com or check us out.

[Carl Brown]

On the iHeartRadio app, Apple Podcasts, or wherever you get your podcast. This podcast is brought to you by FedEx. The new power move. Hey, you know those people in your office who are always pulling old school corporate power moves? Like the guy who weaponizes eye contact. He's confident, he's engaged, he's often creepy. It's an old school power move. But this alpha dog laser gaze won't keep your supply chain moving across borders. The real power move? Having a smart platform that keeps up with the changing trade landscape. That's why smart businesses partner with FedEx and use the power of digital intelligence to navigate around supply chain issues. Issues before they happen. Set your sights on something that will actually improve your business. FedEx. The new power move. So, usually on okay story time, our audience will send in their relationship problems. And the okay storytime squad gives some good advice. Goofy. But today we're not giving out our usual advice. Our producer Riley says we're giving something else. So what are we doing today, Riley?

[Ed Zitron]

Today we're playing a little game game.

[Carl Brown]

Says the man about the special gifts for you guys from ebay.

[Ed Zitron]

Each one picked with one of you in mind.

[Carl Brown]

Yeah, Dakota, if you want to guess.

[Ed Zitron]

All right.

[Carl Brown]

There is a gift at my feet. Open that thing. And now it is in my hands. Oh, I feel like it's got to be our resident gamer kiosk. This is the rectangle of childhood.

[Ed Zitron]

It's a portable game console. I used to have this as a kid.

[Carl Brown]

This game console.

[Ed Zitron]

I used to play all the time.

[Carl Brown]

And you know when your mom came into the room when you were a.

[Ed Zitron]

Kid and like, you're pretending to sleep.

[Carl Brown]

Yeah, yeah, yeah. But Riley, what a thoughtful gift.

[Ed Zitron]

Yeah, right. Thank you so much, Riley.

[Carl Brown]

You're crushing it.

[Ed Zitron]

But we have one more gift.

[Carl Brown]

Yeah. Open it.

[Ed Zitron]

Boom. Oh, camera. Yeah, an old timey camera.

[Carl Brown]

That's right. Classic. This is awesome. Yeah, because you know how I love to take pictures of my travels. Yeah, you're always somewhere, Whether it's in Kyrgyzstan with some nomads or just New.

[Ed Zitron]

York, you know, with a nice little.

[Carl Brown]

Piece of trash or a rat taking pictures with the birds. So, Riley, you got all of this from eBay, dude, eBay. It was really fun finding it with you guys. Like, I had very specific things for each one of you. Yeah, it was all there. Thanks, Riley.

[Ed Zitron]

And thank you, ebay.

[Carl Brown]

And guys, shop ebay for millions of finds, each with a story. EBay.

[Ed Zitron]

Things people love.

[Commercial Announcer]

It's finally here. Black Friday at ulta Beauty. Take 30% off Mac foundations and lip colors. Save $100 on the iconic Dyson airwrap. And starting Tuesday, shop mascaras for $14 and lip colors for $12. Perfect for stocking up. To make it super easy, shop online and pick up in store or get same day delivery. Head into Ulta Beauty and shop Black Friday deals now through November 29th. Ulta Beauty Gifting happens here.

[Carl Brown]

Ford was built on the belief that the world doesn't get to decide what you're capable of. You do. So ask. Ask yourself. Can you or can't you? Can you load up a Ford F150 and build your dream with sweat and steel? Can you chase thrills and conquer curves in a Mustang? Can you take a Bronco to where the map ends and adventure begins? Whether you think you can or think you can't, you're right. Ready, Set. Ford.

[Commercial Announcer]

This is an iHeart podcast. Guaranteed Human.