Internet of BugsThe primary function of a Software Engineer is not just writing the code. Our primary function CANNOT be replaced by "AI" code generators (if it could, we wouldn't have needed all the Agile/ProjectManagement overhead for all these decades).
I said what I said.
The video I mentioned at 5:49 about how software is getting worse (and explaining the graph towards the end of this video) is here: Software is bad, and it's been getting worse. But it's still worth doing. youtu.be/U-IhIqmCHlc
NOTE: (Not that anyone will read this, but) When this video uses the term "AI." it (like most of the current discourse these days) means current and near future "Generative AI" a.k.a. Large Language Models. If you're thinking about the Agents from the Matrix or the Robots from Terminator, that's DIFFERENT. That's the OTHER AI, the Human-Intelligence General AI, Sometimes called "AGI." That DOES NOT EXIST. For the time being it's PURE FICTION. It may get here soon, although I (and a lot of people smarter than me) seriously doubt it, but we'll see.
Software Engineer jobs (Mid-Senior) can NEVER be replaced by a Coding AI like DevinInternet of Bugs2024-03-16 | The primary function of a Software Engineer is not just writing the code. Our primary function CANNOT be replaced by "AI" code generators (if it could, we wouldn't have needed all the Agile/ProjectManagement overhead for all these decades).
I said what I said.
The video I mentioned at 5:49 about how software is getting worse (and explaining the graph towards the end of this video) is here: Software is bad, and it's been getting worse. But it's still worth doing. youtu.be/U-IhIqmCHlc
NOTE: (Not that anyone will read this, but) When this video uses the term "AI." it (like most of the current discourse these days) means current and near future "Generative AI" a.k.a. Large Language Models. If you're thinking about the Agents from the Matrix or the Robots from Terminator, that's DIFFERENT. That's the OTHER AI, the Human-Intelligence General AI, Sometimes called "AGI." That DOES NOT EXIST. For the time being it's PURE FICTION. It may get here soon, although I (and a lot of people smarter than me) seriously doubt it, but we'll see.How Programmer Jobs get created and destroyed, & how AI might change that.Internet of Bugs2024-11-14 | The economics of how tech jobs get created and how layoffs happen is worth understanding if you're not sure how AI might fit into it.
00:00 Previously on @InternetOfBugs vs AI 01:50 Caveats: US Only, not Gaming 02:33 Building vs Maintaining 03:46 How Projects get funded 05:08 Hiring process 07:42 How the last decade or so has been unusual 10:26 How AI might change that 11:19 Enter the Stock Market 12:47 How Layoffs get decided on 15:58 How to ride out the apparent downturn 21:37 Bad advice from people who have never experienced a downturn 24:38 Resources on How to look for a job
It still doesn't have the judgement a programmer needs (or anything close to it), and it can't do the communication parts of the job (and is still awful at writing tests). But it's starting to live up to the capabilities of this form of Machine Learning.
If your only programming experience is homework assignments, or If you're programming workflow primarily consists of copying and pasting from Stack Overflow, then I've got some bad news for you.
00:00 ChatGPT-O1 is different 02:00 What I've been testing 03:03 Coding RegExes is Rust 05:19 Incorrect Assumptions and Expectations 06:22 Rechecking ChatGPT-4o. Still bad 06:59 The specific case where it might replace a programmer job 07:11 But, it's a stupid job that shouldn't exist 08:03 Many C.S. programs don't prepare for useful skills 09:24 O1 Still has a lot of limitations 10:48 Just as bad at tests as Claude 3.5 11:16 Still bad at picking between alternatives 11:51 O1 couldn't even make this compile 13:40 "Human-like reasoning" claim is nonsense 14:22 Don't panic - tools obsolete skills every few years 15:23 But, things are going to be bumpy 15:35 Advantage to smaller, nimble companies for a while 16:01 Still going to be plenty of programmer jobs 17:35 But you need to learn how to use these tools, and well
References:
CodeCrafters (affiliate link) if you want to see how you measure up to 4o: https://iob.fyi/CodeCrafters
My video on Software Testing in an AI world: youtu.be/02FD59whT5A?si=yJZukC5R7W95s5FcChatGPT O1 Preliminary test comparison with previous model test videosInternet of Bugs2024-09-21 | I ran ChatGPT O1 through the same test I ran on other models in
00:00 Intro 00:30 Comparison to tests I ran with @Proof_news 01:10 ECM Prime Number Attempt 01:56 CodeCrafters HTTP Server challenge 02:47 Still can't do a developer job 03:14 But, an increase in code quality 03:40 More to come 03:55 Usability issues 04:14 About comparisons to a "smart PhD"Starting a side-project: Why you (and most developers) should, and how AI can help.Internet of Bugs2024-09-16 | Info at youtu.be/l9Z7KPnP0-c on my ideas about where the channel is going, including channel memberships.
# Sites/Communities where other people are doing this - WARNING: be wary of rip-offs!! indiehackers.com microconf.com
00:00 Not a get-rich-quick scheme 01:20 Who this is for - or not. (Hint: for Developers) 02:01 Why I'm making this video 03:53 Why you should consider it now 07:24 Pivoting: how ideas become projects 08:28 Where AI comes in 09:24 Why this isn't quick: trading time for money 09:35 Don't quit your day job 10:57 More videos on this topic to come 11:24 Books to get you started 11:37 Learn about A/B Testing 12:04 Example online communitiesAI Coding Comparison Challenge. 4 AIs build an HTTP Server in Python. See how they do.Internet of Bugs2024-09-06 | Company provided demo videos are awful, and I'm tired of them lying, glossing over errors and hiding what the AIs got wrong.
So I made my own. The full, unredacted video is at youtube.com/watch?v=mikK5MoApQM where you can watch everything each AI did right and wrong.
This video is a summarization of what happened, what I learned, what the different AIs did well or did poorly, who "won" and what recommendations I have based on these results.
If you want to learn more about how Internet Software works, sharpen your skills, and have some fun, you can use this link: https://iob.fyi/CodeCrafters to sign up for CodeCrafters free and if you do decide to upgrade later, you'll get 40% off.
00:00 Why this AI demo video is different 02:24 The competitors 03:13 Caveat: Current Point in Time. Current versions of products Sept 2024 03:33 The Challenge 06:27 Problem One: "200 OK" 07:43 Problem Two: "404 Not Found" 09:33 Problem Three: Echo/Request Body 10:30 Summary So Far 11:25 Problem Four: User-Agent/Headers 13:02 Problem Five: Adding a Loop 13:15 Problem Six: Returning a File 14:10 Discussion of Results 15:14 Recommendations 15:19 Observations on Companies 16:19 Thanks to CodeCrafters.io for the playing field 16:34 Closing
My video on Software Testing in an AI world, mentioned in this video: youtu.be/02FD59whT5A
My video on the Claude 3.5 Sonnet coding demo: youtu.be/x0y1JWKSUp0Code Generation (Not the AI kind): The D.R.Y. meets K.I.S.S. Tool that Time ForgotInternet of Bugs2024-08-29 | Thanks to @BookOverflowPod for inspiring this video!
00:00 The Secret Technique THEY don't want you to know about 00:18 K.I.S.S. meets D.R.Y. 02:04 Go check out @BookOverflowPod 02:19 The Chapter Removed from The Practical Programmer's 20th anniversary edition 02:52 Why THEY (wrongly) think you don't need it anymore 04:42 Why this is worth learning 06:21 Not getting into Code so I won't lose half the viewers 06:46 The Concept 08:08 Where to find more Information
Open Source Code Generator Projects: [Not an Endorsement - I haven't used these, but they'll give you an idea of how Code Generation should be implemented. But if they work for you - cool] yellicode.com telosys.orgAI Coding Crap: More Examples. Claude 3.5 Sonnet Demo & more - with @Proof_newsInternet of Bugs2024-08-13 | AI Generated code is bad. But the demos do their best to hide that fact. Even Claude 3.5 Sonnet, which the company says "raises the industry bar for intelligence" But in their demo video, they gave Claude a REALLY easy problem to fix, it did a lousy job, and their video glossed over the bad parts.
But, it's not just Claude. In conjunction with @Proof_news and a new AI tool they've built for creators, I've broken down some different problems that human Software Developers can do, but - despite being trained on all the information and having near total recall - AI's just cannot get right.
00:00 AI Code is still not good 00:45 Work with @Proof_news 01:39 Intro to the Claude 3.5 Sonnet Demo video 02:26 Transferring code from video to an editor so I can demo it 02:50 What Claude thinks the bug is - and why it's wrong 03:45 The 3-character fix Claude missed 04:03 But Claude's test code is even worse 04:07 Hey, Claude - that's a square, not a circle 04:34 Don't test by starting with the answer 05:24 Bad demo practices contribute to the hype cycle 05:33 Collaborating with @Proof_news 06:06 Generalizing from "Claude got this wrong" to "AI's in general get this wrong" 06:58 Other kinds of Software problems AI does a poor job of 07:51 A classic problem, with a modern solution the AI's can't code 09:47 When AI's replace human programmers, this is what is lost 10:17 Planning and Estimation - or the lack thereof 11:16 AI coders fail at even small stuff, but the AI companies talk big 11:58 Companies are probably going to get better at not getting caught in poor demos 12:13 More videos about things AI's can't seem to get right 12:27 Wrap up
There are real people being harmed every day by AI. And the companies should be held responsible.
Don't let them fool you that "AI Safety" should be about SkyNet. It should be about preventing DeepFake porn and artists having their work, their voices and their likenesses stolen. It's about resume filtering extortion and AI autopilots crashing Teslas and Boeing 737Max's. Don't ever let them forget it.
00:00 Is AI *Really* as serious as nuclear war? 00:26 Changing the subject 01:36 Topic Statement 01:43 Disclaimer - the future is impossible to predict perfectly 02:48 Perverse Incentives and Moral Hazard 04:06 US Government missing the point on AI Safety 04:40 The points they missed, that we need to pay attention to 05:56 AI is here to stay (in some form) 06:19 We can't keep letting them do this 07:33 The dumbest way to do what they're claiming they're doing 07:57 The Software Development Life Cycles models as they apply to AI 10:02 OpenAI doubles down on "Only we can save the world!!!" 10:42 The dystopian hellscape we're Actually on track for 11:27 Wrap Up.
# The new, even worse supposed existential AI risk axios.com/2024/07/25/axios-interview-altman-urges-us-action-to-beat-china-in-ai-raceCrowdstrike Internet Outage: development techniques that could have prevented it.Internet of Bugs2024-07-22 | In my last video. early last week, we talked about how a study showed that 90% of all catastrophic software failures were caused by poor error handling code.
This week let's see if we might be able to think about some event that we might be able to use as an example of poor error handling, and talk about the kinds of endemic problems in our industry that allow such things to happen, and talk about a number of strategies (Seventeen of them, to be specific) that should have been used to prevent this kind of nightmare from happening, but obviously weren't.
00:00 Clown (Crowd?) Strike fail 00:45 After Channel Intro 01:22 What we know happened 03:28 Business reasons this can happen 06:28 Technical reasons this can happen 07:12 Who this video is for 07:56 Error Handling techniques 08:23 "Works on MY machine!" 09:11 Unit Tests are USELESS 10:01 Web programmers can learn from this, too 11:09 How do you test for this kind of thing? 12:12 What if those tests don't catch it? 12:59 Phased/Slow Roll-out 13:33 Centralized software log collection 14:48 Sanity check files before you execute them 15:17 Pay attention to weird behavior, don't just "see if it happens again" 15:58 Separate high-risk code like parsing new drivers files into a different process 16:18 Minimize the code that runs in kernel space, at boot, or with elevated permissions 18:15 Why didn't the update get rolled back when the boot failed? 19:54 Recap 22:26 These skills can't be replaced by A.I.
00:00 A Philosophy of Software Design 01:16 Chapter 1 & 2 02:58 RE: Google and VMWare - LOLOL 03:17 Abstractions and what's "important" 04:03 Unix and TCP/IP 04:55 Classitis 05:14 Chapter 10: Error Handling 😡 06:34 Second System Syndrome 07:02 On comments 07:22 Naming and Renaming 08:46 More about comments 09:17 Chapter 19: Why even bother? 10:04 Chapter 20: Design vs Performance 11:29 What the book doesn't cover 11:51 An industry wide problem 12:01 Don't be these guys 13:27 Delusions programmers want to believe 15:50 The vicious cycle of bugs, willful ignorance and doom 16:50 Should you read this book? 18:13 Prioritizing your timeAI Hype is completely out of control - especially since ChatGPT-4oInternet of Bugs2024-06-10 | Now that ChatGPT4o is out, it's time to revisit my previous thoughts on AI, and discuss what might be happening in the future, and what evidence we have of what might happen.
00:00 Current brief status of AI 03:21 ChatGPT-4o is faster. at least 03:30 The Multi-modal interface isn't new 04:00 What matters for disruption? 04:48 Science and measurement 05:46 What counts as evidence? 07:54 ChatGPT-4o benchmarks 10:39 Psychology of Humans and AI 11:28 Dark Patterns 12:35 History of AI demos 18:40 New article about AI Hype 19:54 Conclusion
Sources from Video have been moved to a pinned comment, since the description field ran out of characters.
youtube.com/watch?v=VctsqOo8wsc&lc=UgxfeL871kHWrc5-S154AaABAgLearning AI, PyTorch and How to Program with Github Copilot, but I HATE Python!Internet of Bugs2024-05-22 | Being a professional programmer is all about learning new things and keeping up with the changes in technology. Periodically, you have to stop and look around and decide what to learn next.
I've been through a dozen or so phases of my career so far, and at present, I've decided to learn more about PyTorch and LLMs. In this video, I'll explain why, what resources I'm using, how having a coding LLM copilot changes the process, how advice I got decades ago makes working with Github CoPilot much more efficient, and thoughts on the future of programming languages.
Also some thoughts on how web searching is changing for me (much to the better), how that's impacting my thinking on AI (with help from Steve Jobs) and initial impressions of ChatGPT-4o.
00:00 Introduction 02:36 Various tech I've used (much abandoned) 06:22 How I've decided what to study 07:07 Recommendations here not for working on existing codebases 07:41 Advice I got on being interruptible vs "In the Zone" 10:04 "Literate Programming" is a more formal way of doing it 10:45 Literate Programming technique is perfect for working with CoPilot 11:16 Clarification - this is NOT sponsored - just stuff I really use 11:42 Pros and Cons of using CoPilot/LLM/AI to help you code 14:32 Choice of Programming languages (or lack thereof) 16:32 Programming Languages are tools - don't get attached 16:57 Soon, I expect languages to matter much less 17:56 Resources I'm using to learn PyTorch/LLMs 18:49 How I recommend learning a new language/framework 21:09 My current side project (if you care) 22:28 Coursera course on Machine Learning 22:33 YouTube Channels I'm learning a lot from: 22:43 @3blue1brown 22:58 @AndrejKarpathy 23:14 @borismeinardus 23:46 @ArtemKirsanov 24:06 @MachineLearningStreetTalk 24:34 @Computerphile and @davidbombal (for Dr Mike Pound) 24:55 New Web "Search" with perplexity.ai (not sponsored - I just like it) 25:34 Changing my mindset on Search Agents vs Search Engines 26:17 First impressions of ChatGPT-4o 26:26 ChatGPT-4o vs Perplexity.ai 26:36 ChatGPT-4o multimodal (disappointing, but at least it's fast) 27:45 Wrap up
# Search tool I've been using lately (not sponsored): perplexity.ai
# Source for Steve Jobs Quote youtu.be/ptXJjI-l2mk?t=710Clean Code is bad. What makes code maintainable? part 1 of nInternet of Bugs2024-05-13 | In my "Top 10 Software Developer Books" video, there was a lot of discussion about "Clean Code."
It's horrible. It's based on flawed assumptions, and it makes code harder to maintain. In this video I'm going to talk about some specific examples that Clean Code gets wrong, the workflows it prevents, and a bug that took me days to find that Clean Code's STUPID recommendations can cause.
I'll also discuss what "maintainable" really should mean.
There's no way I can cover everything wrong with this book (or everything you need to know to write maintainable code) in one video - so expect more of these.
00:00 Most "clean coding" advice is bad 01:24 "Clean Code" is trash 02:13 Thing like "Clean Code" only serve to create arguments 02:42 "Maintainable" is judged by people other than the programmer writing it 03:25 "Maintainable" code is useful when you do something else for a while and then come back 03:35 Code is not read top to bottom like a book 05:13 Real programmers read code from the bottom up 07:00 "Clean" codebases tend to obfuscate bottom-up reading 08:43 Vertical slices of code 09:40 Root of much programming advice 10:09 Welcome to Whack-A-Mole 11:38 Specific Example - Real Bug (details changed, yada yada) 14:26 The bad assumption in most coding advice: Bugs are preventable 15:28 A bug that's hard to reproduce means the code is bad 16:43 The real point of maintainable code 17:11 Rant 17:55 Wrap upHow Developers might stop worrying about AI taking software jobs and Learn to Profit from LLMsInternet of Bugs2024-05-06 | Right now, the software industry is kind of stuck, where no one wants to hire non-AI developers because the hype is claiming that all the jobs are about to get replaced by garbage like Devin.
But new evidence indicates that the pace of AI growth is slowing down, and that two years of GitHub CoPilot has been creating a 'Downward Pressure on Code Quality.'
The future is never certain, but it looks like there's a path for the next few years to be a new boom in software development, and it might start soon.
00:00 Intro 03:49 Ezra Klein's interview of Anthropic's CEO from April 12th 04:06 There are no Exponentials in the real world 05:10 Research showing LLMs are reaching the point of diminishing returns 06:28 Article on the stagnation of LLM growth 06:57 Stanford AI Index Report 07:18 Research showing that LLMs are running out of training data 07:28 Research on "Model Collapse" 08:13 Research showing AI reducing overall Code Quality 09:04 Quick Recap 09:27 Implications to Software Developers 10:56 Parallels to 2008/2009 and the App boom 11:44 How and when we might know 12:07 Wrap up
# Release Hype about GPT-5 tech.co/news/gpt-5-preview-release-rumorsFreelance Developer Books for building your software consulting businessInternet of Bugs2024-04-30 | By Popular demand, I'm doing a video on Software Freelancing - which may become a series if enough people want that.
This video - in order to give you the most information as quickly as possible, is a series of book recommendations for books I've found helpful in the freelancing parts of my career.
NOTE: This is NOT a video about software or programming books. This books about the freelancing part (leads, process, business, communication, etc) that you need to know to do that kind of work effectively. These books aren't software/programming specific, and are likely useful to other kinds of freelancers.
This video also explains some about how I got into freelancing, and how I lucked into some experiences that won't be available to everyone.
00:00 What this video is about 01:21 Caveats - Freelancing is difficult 02:17 How I got into freelancing - short version 03:59 Different kinds of freelancing 04:58 Project Based Work 05:38 Subcontracting is much like being an employee 06:34 Upwork and Fiverr and such sites 08:52 Getting freelance jobs is like selling yourself 09:45 Most freelancing comes from leads from your network 10:18 Books on networking (people/business, not TCP/IP) 11:46 Free Agent Nation - about the "Gig Economy" 12:36 Two "How-To" Consulting books 13:24 You need a book on fiance/taxes for your country 14:06 Books about Project Management - estimating for bids 18:20 "Back of the Napkin" - client communication 19:16 "We Got Fired" - Turning unemployement to your benefit 20:42 Wrap up
Alternatives to Microsoft Project for planning and estimation: OmniPlan for the Mac (What I use and Prefer these days): omnigroup.com/omniplan Free/OpenSource/CrossPlatform: ProjectLibre: projectlibre.com/product/1-alternative-microsoft-project-open-source OpenProject Community Edition: openproject.org/community-editionTop 10 Programming Books-Dead Tree Edition: Internet of Bugs Book Club + I prove(?) Im not AI!!Internet of Bugs2024-04-26 | As requested: This is volume one of my programming book recommendations: Dead Tree Edition: The 10 books (or book categories) that have been the most valuable to me over the last 35 years:
00:00 Intro 00:43 Channel Intro 00:53 Book Relocation and proof(?) I'm not an AI... 02:34 The Pragmatic Programmer by Andrew Hunt and Bob Thomas 03:40 The Mythical Man-Month by Fred Brooks 04:20 Working Effectively with Legacy Code by Michael Feathers 04:42 SQL for Smarties by Joe Celko 05:39 Get a book on Assembler for your processor of choice 06:44 Get a textbook on Algorithms you can look stuff up in 07:46 Transaction Processing by Jim Gray and Andreas Reuter 09:05 TCP/IP Illustrated Volume 1 by W Richard Stevens 10:35 Advanced Programming in the Unix Environment by W Richard Stevens 11:56 Firewalls and Internet Security by Cheswick and Bellovin 13:39 Find the new technology (LLMs?) for your time that Firewalls were for me, and learn it. 14:49 The theme: Learn the underlying tech your code lives on, not just the surface level 16:52 Sign off
There's about to be a war. Not a "Robots Kill All Humans" war, but an "iPhone vs Android" war: and it's for who gets to have the winning AI Agent. And what those Agents will be able to do... They might not be as scary as the Agents from The Matrix, but they'll be bad enough, and they're a billion billion times more likely.
And if the Agent's capabilities weren't sufficiently frightening already - in a race to be first to market, the winner is *always* the Bugs. Careless people write exploitable code, rushing makes people careless, and incentives make people rush. If we're not careful, this is going to be a nightmare...
00:00 Intro 01:37 Channel Intro 01:58 List of components 03:52 Phase 1: Al Agents 05:21 Phase 2: Add Video 07:15 Phase 2b: JOI 08:04 Phase 2c: Checkpoint 10:17 Phase 3: Don't Google This 11:33 Phase 3b: Enter Biometrics 12:16 Now with even less reality 13:14 Viral Conspiracy Hallucinations 13:54 The two corners 15:22 In the race to be first, the Bugs Always Win!! 15:53 A bright side? Maybe not?
UPDATE: I did't see this until after I posted this video, but this dropped in the last couple of days (assuming it isn't faked): microsoft.com/en-us/research/project/vasa-1
# CAREFUL WITH THESE: You can't unsee or unlearn what you might see or learn here. # GREAT reporting - disturbing topics: vice.com/en/contributor/samantha-cole 404media.co/author/samantha-coleMyth of the 10x Developer: Technical Interviews are Broken, (part 2 of n)Internet of Bugs2024-04-16 | One of the reasons the code-test LeetCode HackerRank style interview process started was because of the myth that there were 10x developers, and that, by interviewing well, you could fill your big tech company with just those top performers.
And that's how we got here.
And, chances are, most companies will stay here.
But, if your company interviews differently, then in my experience, you can get a huge productivity advantage.
00:00 Video Intro 00:20 Channel Intro 00:34 Categories of programmers for our purposes 04:16 The Original. Flawed, 1960s Experiment 06:00 A few truths underlying the myth 07:48 Context for the argument in the video 08:29 My argument against 10x developers in corp settings 09:49 Not a skill issue, and not something LeetCode can detect 12:00 What LeetCode might select for at many companies 13:03 The advantage companies can have without LeetCode 14:21 A better focus, according to me, not that it will matter 16:03 This one neat trick for better productivity 17:28 Back full circle to where we started
Links from the video: The Mythical Man Month: en.wikipedia.org/wiki/The_Mythical_Man-Month Great book - highly recommended. I didn't talk much about it in this video, but i'm sure it will come up in the future.
Original Paper that used toy programs to invent the 10x myth: https://web.eecs.umich.edu/~weimerw/481/readings/productivity-performance.pdf
Modern paper dispelling the 10x myth: https://insights.sei.cmu.edu/blog/programmer-moneyball-challenging-the-myth-of-individual-programmer-productivity/Tech Interviewing is Broken (Part 1 of n), How we got here and why were still stuck here.Internet of Bugs2024-04-14 | Tech Interviewing is so broken. My first dry run at this topic clocked in at 73 minutes, so it's getting chopped into several videos.
This video talks about how interviewing is broken, and some of the history about how we got here, why companies KEEP interviewing like we do even though they should know better, and a bit of job hunting advice - to go along with the last jobs video I made at youtu.be/NQmN6xSorus
Remember - no politics in the comments, or I'll delete them.
00:00 Intro 00:18 I am not trying to sell you anything related to this video 02:00 Advice for job seekers on how not to get ripped off 03:30 Book recommendation 04:29 What I do spend money on in between jobs 05:38 My Amazon Interviewing experience 10:04 My history/experience with 35 years of professional interviewing 12:43 My opinion on the effectiveness of companies that use LeetCode interviews 14:36 Live whiteboard coding and take home tests have roughly the same effectiveness 15:24 How did we get here? 20:32 Why do companies keep doing this? 25:52 Consequences for companies of this poor interview process 28:43 Somewhat off-topic rant I couldn't bring myself to cut out 29:47 Choices while job hunting 30:36 Signoff
# Note: I do not believe in the concept of a Super Star candidate. Subscribe and stay tuned for my upcoming video, tentatively titled "The Myth of the 10x Developer"
The book I would actually recommend for job seekers (you can probably find it used): parachutebook.comDebunking Devin: First AI Software Engineer Upwork lie exposed!Internet of Bugs2024-04-09 | Recently, Devin the supposed "First AI Software Engineer" was announced. The company lied and said that their video showed Devin completing and getting paid for freelance jobs on Upwork, but it didn't show that at all.
UPDATE!!: The original job poster of the Upwork task has made his own video!! Go watch it here: youtube.com/watch?v=xE2fxcETP5E ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ I'm not kidding! Go watch the original poster's video!
I broke down the Devin Upwork video frame by frame, and here I show what Devin was supposed to do, what it actually managed to do instead, and how bad a job of that it did.
On the whole that's not surprising given the current state of Generative AI, and I wouldn't be bothering to debunk it, except: 1) The company lied about what Devin could do in the video description, and 2) a lot of people uncritically parroted the lie all over the Internet, and 3) That caused a lot of non-technical people to believe that AI might replace programmers soon.
00:00 Intro 00:30 The claim and the problem 03:47 What the job actually would have required 05:50 Requirements that needed to be determined 07:32 How a human compensates for Upwork's lack of RFP process 10:11 What Devin did instead, and how poorly 13:03 Devin seems to be fixing code from Github 14:19 But Devin is actually making up errors, and then fixing them 15:40 All Devin had to do was run the command from the README 16:23 But Devin couldn't figure that out 16:26 So it created this nightmare 'C'-style low-level buffer append loop in Python 17:25 My replication of what Devin tried to do 18:15 It took me about 36 minutes 20:16 It took Devin at least six hours, and maybe more than a day 20:48 more bad Devin code 22:08 List of useless things that make Devin look competent 23:48 Conclusion, and a Plea
Video falsely claiming "Watch Devin make money taking on messy Upwork tasks!" that I'm debunking: youtube.com/watch?v=UTS2Hz96HYQ
Raw, uncut, boring video of me replicating Devin's results in 1/12th the time: youtu.be/TMl_82eavHo
My video about communication being the most important part of software work: youtu.be/7-f7rPdj6tI
My video about why Devin isn't really a "Software Engineer": youtu.be/Mj6zQUWw6j4How Testing Software will need to change in an AI-driven WorldInternet of Bugs2024-04-02 | As the use of Generative AI as a tool proliferates, the amount of code that will need tests will dramatically increase, and the quality of that code will go down. Generative AIs like Devin or ChatGPT won't be able to replace Software Engineers. They will however, both make us more productive and make more work for us.
The best book I know on Software testing is "Working Effectively with Legacy Code" by Michael Feathers (informit.com/store/working-effectively-with-legacy-code-9780131177055). Unfortunately, it was written in 2005, and the techniques that were best practices then just aren't going to cut it in a world writing code at an AI-driven pace.
00:00 Intro 04:16 Flashback to NetSpend -2000 06:31 Book recommendation, with reservations 08:37 These Days We Can Do Better 11:41 Developers Need to Do Better, too 15:35 What about Ul Tests? 16:27 The new minimum standard 20:55 The Challenge 25:07 Recap 27:09 Until next time...AGILE & Scrum Failures stuck us with AI hype like DevinInternet of Bugs2024-03-28 | How did we get to the point where so many people believe that an "AI" like Devin will completely replace the jobs of most, if not all, Software Engineers? Why is the work of Software Development so poorly valued?
The question here isn't "is it true Devin could replace Software Engineers?" Because it can't, and I already made a video on that.
The question here is: "Why is it so easy for people do believe it could?" and "What does that mean for the future employment of Software Developers in a pervasive-LLM world?"
I think I have an answer, and, just like much of the problems we've had in the last 20 years, it all starts with AGILE's failures.
00:00 Intro & Question 00:30 The 1990s and Gantt Charts 00:55 Extreme Programming 01:44 Branding FAIL 01:53 AGILE Manifesto 02:15 Scrum Alliance 03:14 Perception of term "Engineer" 03:55 Perception of term "Software Engineer" 04:34 What do people think Software Engineers do? 05:21 What can we start doing about it?
NOTE: (Not that anyone will read this, but) When this video uses the term "AI." it (like most of the current discourse these days) means current and near future "Generative AI" a.k.a. Large Language Models. If you're thinking about the Agents from the Matrix or the Robots from Terminator, that's DIFFERENT. That's the OTHER AI, the Human-Intelligence General AI, Sometimes called "AGI." That DOES NOT EXIST. For the time being it's PURE FICTION. It may get here soon, although I (and a lot of people smarter than me) seriously doubt it, but we'll see.5 new Software Engineer Jobs that AI will make (that arent for ML specialists)Internet of Bugs2024-03-22 | Not only with "AI" LLMs like Devin not take all the Developer Jobs, but there will be new Software Engineering opportunities made by "AI" in the future. And i'm not talking about "AI" Jobs like Machine Learning Experts. These are jobs for Existing Developers and Software Engineers living in a world full or ChatGPT and Claude and Gemini.
Don't believe the hype about there being no human programmers left in five years. "AI" Models are software, and ALL SOFTWARE HAS BUGS, and with bugs come opportunities.
Chapters: 00:00 Intro 01:57 Number One: New Startups 02:02 Number One Part 1: No-Code Web History 03:45 Number One Part 2: LLM Startups, Likewise 05:24 Number Two: Automation created more products 08:20 Number Three: Customizing LLM Models 10:15 Lora and ControlNet Demo 11:19 Number Four: Debugging LLM Generated Code 12:23 Number Four Part 1: What was it thinking? 15:23 Number Four part 2: Regression Bugs 16:37 Number Four part 3: LLMs have Bugs in their own code 17:30 Part Five: Adversarial Attacks 20:12 Adversarial Attack Examples 20:31 Hope that wasn't depressing, too. 20:39 SignOff
Links from the video: For more on ComfyUI/StableDiffusion (This guy is great): youtube.com/@sedetweiler
NOTE: (Not that anyone will read this, but) When this video uses the term "AI." it (like most of the current discourse these days) means current and near future "Generative AI" a.k.a. Large Language Models. If you're thinking about the Agents from the Matrix or the Robots from Terminator, that's DIFFERENT. That's the OTHER AI, the Human-Intelligence General AI, Sometimes called "AGI." That DOES NOT EXIST. For the time being it's PURE FICTION. It may get here soon, although I (and a lot of people smarter than me) seriously doubt it, but we'll see.Software Engineers REAL problem with AI and JobsInternet of Bugs2024-03-19 | "AI" isn't going to replace all developer, programming or software engineer jobs. Devin is not an "AI Software Engineer" at all, much less the first one.
But "AI" is destroying developer jobs in another way. It's making it almost impossible for competent software engineers to have their resumes seen by hiring managers by flooding all the programming job postings with ChatPGT-generated spam cover letters and auto-submitted resumes with buzzwords added to match the job posting list. And it's doing that hundreds of times an hour.
You can't beat the "AI" at job postings, so the best thing to do is go back to the old ways. At least until the industry figures out how to get around the job posting nightmare AI has created for us.
Let me tell you how we used to do it, and what I'm guessing we'll fall back on for the time being, and I'll give you some advice from someone who has been doing this a long time.
NOTE: (Not that anyone will read this, but) When this video uses the term "AI." it (like most of the current discourse these days) means current and near future "Generative AI" a.k.a. Large Language Models. If you're thinking about the Agents from the Matrix or the Robots from Terminator, that's DIFFERENT. That's the OTHER AI, the Human-Intelligence General AI, Sometimes called "AGI." That DOES NOT EXIST. For the time being it's PURE FICTION. It may get here soon, although I (and a lot of people smarter than me) seriously doubt it, but we'll see.No, Devin is not a Software Engineer. but it *is* really cool.Internet of Bugs2024-03-13 | An experienced software engineer's take on @Cognition-Labs's new "Devin" offering. It's not really a Software Engineer, for reason's I'll explain. It's a tool, but it's a really cool ai tool, and I look forward to using it and other LLM code generation and problems solving tools like it.
But I'm not worried about it replacing all software engineering jobs. And you shouldn't worry about the question of "can ai replace software engineers?" either.
NOTE: (Not that anyone will read this, but) When this video uses the term "AI." it (like most of the current discourse these days) means current and near future "Generative AI" a.k.a. Large Language Models. If you're thinking about the Agents from the Matrix or the Robots from Terminator, that's DIFFERENT. That's the OTHER AI, the Human-Intelligence General AI, Sometimes called "AGI." That DOES NOT EXIST. For the time being it's PURE FICTION. It may get here soon, although I (and a lot of people smarter than me) seriously doubt it, but we'll see.Could AI replace all programmers in 5 years?Internet of Bugs2024-03-09 | * Educated, professional explanation of the state of the benchmarks used to test Generative AI code generation. * What those benchmarks do (or don't) tell us about the gap between LLMs and programmer jobs. * Advice about what traditional programmers should focus on in the next year or two.
MIT Intro to Python class (for comparison: https://ocw.mit.edu/courses/6-189-a-gentle-introduction-to-programming-using-python-january-iap-2011/ https://ocw.mit.edu/courses/6-189-a-gentle-introduction-to-programming-using-python-january-iap-2011/pages/assignments/ https://ocw.mit.edu/courses/6-189-a-gentle-introduction-to-programming-using-python-january-iap-2011/resources/mit6_189iap11_final_proj/
NOTE: (Not that anyone will read this, but) When this video uses the term "AI." it (like most of the current discourse these days) means current and near future "Generative AI" a.k.a. Large Language Models. If you're thinking about the Agents from the Matrix or the Robots from Terminator, that's DIFFERENT. That's the OTHER AI, the Human-Intelligence General AI, Sometimes called "AGI." That DOES NOT EXIST. For the time being it's PURE FICTION. It may get here soon, although I (and a lot of people smarter than me) seriously doubt it, but we'll see.Software is bad, and its been getting worse. But its still worth doing.Internet of Bugs2024-03-04 | Welcome to the Internet of Bugs.
I've been a software professional for 35 years, and I'm the father of a child who is now a College Freshman majoring in Computer Science.
I started this channel to answer the questions I've been getting recently from College CS students about what a career in computing is like.
Maybe someone else will find it useful, too.
Link for data for bug count graph (Retrieved Feb 2024): cve.org/About/MetricsAI wont replace all programmers. Weve heard this before.Internet of Bugs2024-03-04 | Generative AI won't replace all programmers. These panics happen every few years, and Generative AI isn't any more likely to replace all the programmer jobs than the many previous things.
Here's the story of what was supposed to kill the IT industry in 2000-2004, what actually happened, why that situation is similar to today's, and why surviving that one can give us confidence we'll survive this one.
NOTE: (Not that anyone will read this, but) When this video uses the term "AI." it (like most of the current discourse these days) means current and near future "Generative AI" a.k.a. Large Language Models. If you're thinking about the Agents from the Matrix or the Robots from Terminator, that's DIFFERENT. That's the OTHER AI, the Human-Intelligence General AI, Sometimes called "AGI." That DOES NOT EXIST. For the time being it's PURE FICTION. It may get here soon, although I (and a lot of people smarter than me) seriously doubt it, but we'll see.Programmers Worst Problem in 2024 (for professional Software Engineers)Internet of Bugs2024-03-04 | This is the biggest problem with being a software professional in early 2024. It's not "AI", it's not layoffs, and it's not even Microsoft (although I use them as an illustration in this video).