As a self-taught coder who's trying to figure out what to do with my career, really appreciate your insight and perspective as a veteran of the industry. Thank you!
DUDE. Your speaking my language (no pun intended). I’ll jump from JS, TS, C# and Java in a single day sometimes. I really just don’t care about the damn function/method syntax. Same with iterating, or modeling an object. This is just busy work that needs to be automated. The hard part is architecting how the thing works which copilot cannot do.
Excellent video, the concept of "try something, then learn about it" really resonates with me, my best learning experiences have all followed that paradigm. When it comes to co-pilot, I have mixed feelings. I really enjoy having the state in my head, and maybe I'm being stubborn, but I really enjoy that flow state where things are not written down (I also do not have that hard of a time picking things up again). When I have tried co-pilot, I found that it was a little too interupt-y. It would constantly derail my train of thought with it's suggestion, and I would many times spend just as much time processing the suggestion and making sure it's what I want than it would have taken me to write it myself. My current thought is that the ideal co-pilot is one that spends the majority of the time observing, only interjecting when it thinks I've made a mistake or introduced a bug, almost like automated real-time PR comments in my code. There are times when creating certain known functionality would be useful, but many times I want to trigger that, so a special way of ending a comment where a few function body suggestions get created from some doc-string like comment. The constant interjection just feels like the wrong medium for the ideal assistant.
I have noticed lately some people giving the advice of "go build a thing some language/framework/technology, when that is done, you are ready to learn how to use that technology". I think this is really good advice and it mirrors a very popular technique in learning regular languages like Spanish or Russian. Just start using that language anyway you can then when you get to a point where you are somewhat proficient in it go back and learn the 'rules' like the grammar. You are in a much better place to understand at that point and it is very good at keeping you engaged.
+1 This is very important point. As we have a limited amount of energy, time and memory capacity, you CAN'T learn every little detail. Understanding how things work in the big picture and then moving down towards the details is the way to go with programming. Same with languages, the grammar rules are almost endless. If you focus on them you are actually becoming more familiar with the rules themselves rather than the thing that actually matters - the language. You spend your energy on the details that you get lost into them, having no energy for the big picture.
@@mantas9827 This has been my experience even with spoken languages. I 'studied' Chinese almost 10 years ago and understood a lot of the rules but could speak very little of the language and gave up. We all learn to speak as children by doing, then learning. I wonder how this holds up to learning any skill. This is a good mind puzzle to mull over.
Exactly same with music. Learn minor pentatonic, get to a “mom look at me, I’m a rock star” and only then start your music theory journey. You have to get to a level of understanding where you are genuinely interested and have a real need for knowledge. Look at kids who spent 9 years in music school - they know a ton of stuff, but don’t play😅
Here! You can HAVE my subscribe just for the flashbacks inspired by the mouse switching to the hourglass multiple times before settling. You even invoked the archaic sounds of my 800MB hard drive crunching away while I waited. 😂
When I was first forced to use Python a long time ago I hated it too. I was annoyed by the indentation and dynamic typing. Back then I struggled with issues like tabs & spaces ending up in the same code, but I can't recall running into any issues with indentation in a long time. And I like how that feature has been adopted into other languages. The code just looks neater. I was used to statically typed languages, so working with dynamic typing required a bit of mental adjustment, but the increased flexibility and expressiveness is sometimes quite handy. Nowadays Python is my language of choice for all quick&dirty jobs and maybe some small to medium size projects, though I probably wouldn't use it for anything really big.
A similar talk to the one you reference at the end is "stop drawing dead fish" by brett victor. It focuses very specifically on how to use the programming medium for learning/teaching. Its pre ai but very interesting.
It's a frustrating time for aspiring developers like myself... I'm sat here working my a$$ off to learn backend microservice architectures, different OAuth2 authorization servers, deployment with Kubernetes, Spring batch processing, CQRS with Kafka or Axon etc. and everyone I know who isn't a programmer is all "you NEED to learn AI 🤡🤡" Of course, a year or so ago everyone was telling me I "NEED" to learn blockchain stuff, and I did and made a few projects, and probably wasted my time doing so...Like you said in your video... Solidity ain't coming back
Thanks for this video.. really great. I am new to programming and i've been taking cs50 python for the last 4 months.... Having the LLM's at my disposal have helped me progress on the course.. I am probably not a good programer by any means as I relly so much still on that tech to solve the course problems, but the fact that i can ask even the most irritating question, and always have an answer that I can evaluate; is a great tool that I try to squize the most out of it by asking as many questions that often lead me to understand the concepts.
The part about doing a project first, being proud of it, etc. then going back and learning things under the hood is incredible advice. Resonates with me and is something I will recommend to any junior dev. It's advice so sage in fact that I can forgive you for disrespecting Python 😂
Turbo Pascal was my second programming language in high school ! Turbo Basic, DBase, IBM AS/400. Nobody talks about them anymore. Those were exiting days for programming.
I'm pretty close to entering my first 15 years of programming career. I've spent most of my time working for B2B companies, and I always had to keep up with the current tools and tech stack in my private time. Enterprise companies always had massive legacy tech, and since it would be a disaster for them, no one wanted to change or adapt to new things. I always felt overwhelmed and had a bit of FOMO for not checking the latest 999 JS frameworks, which claim to be making 1% per cent better than the framework that came out six months ago. Eventually, I stopped following these types of content creators, and their newest videos make their previous videos outdated. When I look back, I see that I've worked on a scale from Visual Basic 4.0 apps to event-driven microservices architectures with 3 layers of caching; your content and your experience are the most honest.
"Try something then learn it formally" is a technic i started applying since last year. I started to hate attending courses before getting started with something new. That linear way to learn while building a boring project not aligned with my interests, and depending upon the teacher speed, how many minutes he/she takes on explanations... All these things really desperates me. My solution was to started to gather the information by myself on demand to build a project im really interested in to get it done and it works! I ended up learning a lot. And it gets better because with this knowledge learned on demand and built on top of my research i have a deep understanding on these things i wanted to learn, not just the linear way to implement something but real knowledge i can apply depending upon the project needs. It is awesome.
3B1B and Artem's animations and content are king. As to the comments on 4o, I agree the speed is a massive leap, but I would also add that I think the shortfalls are temporary as this is a new paradigm from the models they've offered before. All those modalities are rolled into 1. I think it's actually 2 but collectively working as 1. A primary, narrow, well defined (but shallowly predictive) bottom-up model and a broad top-down model (that can assimilate more weights) offering solid predictions to unknowns being offloaded from the first model - all standing as an analogue to our brain's vision system. It just so happens to work with more modalities as ours does. We can hear an ambulance, and visualize exactly what it looks like, tell it's apprx velocity etc. or read a comment and can visualize an ambulance. To this, I firmly believe we've all already been presented with 'AGI', I think it's still just a shard of the capability they probably have back in the kitchen though.
Carl, love your stuff. The variety of work you’ve done in your career is really impressive. Question for you. I’m a few years into my career now and my experience is fairly niche (embedded). When you decide it’s time to do something new, how do you substantiate the skills you’re self-learning well enough to get a job working with them?
First off, I'll do a side project in that skill, so I feel comfortable with it, and I have something I can talk about. Then I try to find a project at the intersection of what I've done before and what I want to move to. So, for example, if I were just doing embedded, I might see if I could find a project where the embedded device was talking to a mobile app, and then, while I was on that project, I'd follow along with the mobile team (build their code onto my own device, debug the phone and the embedded code at the same time), and ideally contribute changes to the mobile app code if I could. Then, on my next project, I'd go after a mostly-mobile gig (maybe one at a place where they have some embedded something if I had to). Then, on that mobile project, I'd follow along with the Cloud code or something. Or, I'd find an embedded project where the embedded device had a WiFi chip and talked to some Cloud service directly, and I'd follow along with the Cloud group's code/project, contribute if I could, do my own project (doesn't have to be huge or intricate - just enough to talk intelligently about) with that Cloud server tech (Lambda, Flask, Node, whatever). And then try to get my next project doing Lambda/Flask/Node whatever. I'd also probably try to study for, take , and pass some certification tests (e.g. AWS Certified Solutions Architect, Google Professional Cloud Architect, Microsoft Certified: Azure Solutions Architect Expert etc).
I have to admit, as I heard you rhyme off all the various changes in tech you've had to cope with over the years I can't help but have my doubts if being a software engineer is worth it as a long term career choice, especially in parts of the world where it isn't well paid like it is in America. In your case it's perhaps worth it as you have reached a point where you are a highly paid consultant and/or you love continual (unpaid) learning on top of your job, but many do not and realisitically there will be other more secure occupations that frankly provide an easier way to earn a similar amount of money. I think the reason so many move from tech to some form of management before 30 is because they see the road ahead and have other things they want to do with their lives than constantly stay 'relevant'. (I say this all as a programmer of almost 20 years experience!)
If I didn't REALLY enjoy the feeling of accomplishment that I get when I fix a bug or finish a project, or craft a difficult solution to a design problem, I wouldn't have made it 10 years as a coder, and probably wouldn't have made it 5. It just wouldn't have been worth it. And if I was really motivated by money, I would have gone into Finance. I have some friends from school who got Physics degrees along side me who became Quants and probably made 20x what I did.
Python is a great tool provided you aren't making anything above a certain size. Once you're making anything approaching a complete software product, it becomes pretty nightmarish. (I've run into almost the exact same type of bug you mentioned) I work in R&D so a ton of my job involves gluing together disparate math and ML libraries. Python's strength is that you can start experimenting by actually writing code instead of spending a week getting cmake to compile your random point cloud processing library written by a post doc with poor communication skills.
LOL at cmake. I've been there. I agree about it being great at gluing things together. And Jupyter notebooks make that even better - maybe the best I've seen in any language for "get the answer and move on" kind of problems.
@@InternetOfBugs Definitely agree! In the spirit of balance, I can also think of times where python was used and was as the worst tool for the job. Once I was sent to a jobsite to help with implementing some factory automation software that was holding up the setup of massive factory line. They were months behind. The customers wanted a python-based data collection program to interface with the factory PLCs for loss tracking. This was the first python program most of these devs had written, and they were using 8000 lines of python to control the data flow and even some control logic through the factory machines (instead of just using the PLCs themselves). So every time they put an argument of the wrong type into one of several hundred functions (many with over 15 input arguments) the entire line could freeze up and every other technician on site would have to wait for the error to be fixed. In that situation, python was nothing but a nice long, strong rope to hang themselves with. After a couple weeks sloughing off huge chunks of redundant code, I finally convinced them to just do everything but the data logging in the PLC and miraculously, they started making progress again, lol.
Someone had this talk in RU-vid that was titled something like "The goal of programming Python is to use Python least amount as possible" i.e. use libraries coded in something else than Python.
Dear Americans, it is endearing that you feel the need to keep apologizing over and over about how you don't know how to pronounce non-English names. At least from my side, it is most often not necessary and perhaps a bit too much. Thank you. Apart from that, thank you for sharing your experiences with all of us! Entertaining, well-spoken, insightful and honest.
I have a physics degree as well, and it helps knowing partial derivatives and linear algebra which is the basis of machine learning, but I would still recommend diving into the math at least a little bit to get a better intuition of how neural networks work. Even if it's just programming a simple neural net from scratch that can classify different logic gates ( this is a good hello world machine learning example).
15:30 TBH, This is like saying that because someone put a statement outside an if block and caused the bug, you don't like bracket scopes. Just keep going, when I was learning Python I just got used to it after 6 months or so, and now I can see indentation scopes the same way I see bracket scopes. If it is too difficult, consider adding a blank line after for/if blocks.
I ran into fundamental copilot issues very recently. The issue is I’ve become relying and trusting copilot too much. I had three in one day very simple, not logical, but value related bugs sneaking into my code. Things like the wrong enum and const string being the wrong value. These use to be typing errors but now they are AI generated errors. I know that this is what tests are for, but when you move fast, and you trust the AI to fill in stuff for you, and you don’t triple check (which you don’t do, I have AI for that) these things are getting more frequent.
I definitely have never trusted it - but I'm pretty good at spotting its logic errors (and writing (or telling it to write) test code to cover the logic).
I really like your history. I finished my software engineering degree in 2013. That was before the silicon valley boom. So we primary learnt OO in second and third generation languages C++, JAVA. My first job left all left behind. The market for engineers was pretty small then, well in Australia. You got a government job primarily using the two above or you worked for a small marketing shop. All of which were minimal pay and in some kind of LAMP stack. So i cut my commercial teeth on PHP then back to JAVA then Python and CFML(stay away) and now a primarily Javascript/Typescript. Really just following the market for what people want out of their Web Engineers. I've also done a fair bit of mobile dev along the way. Always on languages and frameworks that where platforms agnostic. Touching many different languages along the way has taught me a few things. Dont get mixed up in which is better for x,y reasons. Concentrate on how productive i can be and are there patterns and skills ive learnt along the way transferable and how seamlessly i can slot into a new code base/team/technology without creating too many waves.
@magiclover9346 > 2013. That was before the silicon valley boom Sorry - had to LOL there. Not trying to make fun of you, but I've been there before. My "first" Silicon Valley boom was around the Founding of NetScape in 1994. Then there was another one that ended right about the so-called "Dot-Com" SuperB owl in 2000. The next one was the rise of Google/Facebook/RU-vid/etc. You'll understand in a few years when someone you're talking to casually refers to the "first Silicon Valley boom. You know - in 2023, when ChatGPT was released." You have my sympathies for being forced to deal with ColdFusion. That thing was a mess. But not every language is a winner, and sometimes it's hard to know until you've been working with one for a while. But I think you have the right idea (but I would, because I have had the same idea) of going where the market is. It's the easiest way I know to continue making a living.
I'm OK with Python: mostly because whatever you do, there's already a library for it and increasingly because GPT4 isn't half-bad at helping me out with the stuff that I hated most about Python when I was just on my own. I have worked in most of the languages mentioned in the video. I'm making baby steps into learning Swift, but it gives me "icks" for some reason even though some of the stuff is pretty nice. The language I have enjoyed the most and still like programming in is Lua. It is far less practically useful than Python, but it's just so much more elegant in its simplicity. I'm just coding for my own (and friends') needs right now. If Apple introduces a great Swift AI copilot this year at WWDC, I might well dive in and work on some iOS and MacOS apps in Swift. I was OK with Objective C, so I think I'll be OK with Swift too.
I also thought that there is no excuse for an AI to make syntax errors. But then I realized there are two approaches to this problem. One is to use a compiler. But a lot of the code examples and snippets need some kind of an environment to complile, unless they're doing something really simple and self-contained. Another one is to write a parser that only does a basic grammar check. I am not sure which type of mistakes the Copilot makes and the ideal solution would maybe something that's a custom average of a compiler and a grammar checker. But since the whole attraction of AI is in allowing you to brute force things that used to require thinking, the more you need to customize your solution, the less advantage there is to using an AI.
I found Introduction to statistical Learning funny. It's also from Stanford. Could not complete it but I really found the course to be hilarious and the book, jupyter notebooks really wholesome.
Python is my first proper programming language I used, coming from a Matlab background for my research work. I hated it at first, but that’s because I use it more for modelling and have a more functional programming style. But over time I learnt to adapt and now I really love it. I totally agree with how terrible it is to debug by design - indentation instead of end, how variables affect each other insidiously because setting them to be equal doesn’t just assign equal values but basically mean they are the same object as well. Zero based indexing… Urgh. But on the upside, the libraries, resources, tutorials more than make up for it, for my use cases.
Great video, as always. I've been doing python coding for living for the past 15 years, and I don't feel offended :) Don't forget that it's a programming language that struggles with for loops.
It might be too much to ask as it might be painful to you but ... how about a video about why you hate Python? I'd be curious about specific things you think Python or its ecosystem is doing wrong, or even less specific -- just getting to know more about your opinion. I write Python for living, and I like it (while there's always a handful of languages I'd be happy to do more coding in) but it's always good to learn about things that can annoy or cause problems to other people.
I'd also recommend Sebastian Raschka (On top of all the other great people like Andrej that you mentioned) for the nitty gritty details of language modelling. He has a few books like "Building a Transformer from scratch' on Manning and one that's more of a Q&A for big AI topics called "Machine Learning: Q and AI". I graduated with a bachelor's specific to ML and data science so I'll try to hunt down more resources I've saved over the years. To help me stay on top of things. Lately I've been doing a project of setting up and eventually fine tuning my own local "code pilot" model to help me with future projects. It's been super fun.
Cocoa and Objective-C had a really good reference counting system/memory pool system which I haven't seen in other languages. It was manual but I never had memory problems and rarely leaks.
Yeah. The problem with it (same problem Swift has now) is finding and breaking reference cycles (where two or more objects refer to each other, but are disconnected from the rest of the object graph.) A mark-and-sweep Garbage Collector could find and free them, but ARC can't. If you follow certain best practices, you can reduce that, but there's no way to get rid of it entirely. That's a killer for certain use cases, especially in Swift-on-Linux.
@@InternetOfBugs I'm using Swift now at my job and can report very few problems. It would be bad language for real time graphics and high performance processing but it's gone well for Apple UI apps.
I think the future of production python is fully typed, and compiled, it just makes so much sense, makes it 100x faster and way more debuggable and its still interpreted in development
I feel learned to code in python and thought I’d never need another language. Then I discovered the joys of strongly typed compiled languages and endured the pain of javascript and my perspective forever changed
16:32 Nice sentiment, but there was someone who reframed programming languages as toolboxes because most programming languages we care about in industry exhibit many different paradigms. This runs counter to the idea that programming languages are tools because libraries with specific licenses in specific languages are available for companies to use. This is weirdly unique because with patents, the mechanisms aren't usually tied down to specific ingredients in a very specific order (such as software libraries), but there are plenty of software patents regarding data formats and transformations (such as boring data transit stuff for the sake of ingestion). Most bugs come from poor requirement analysis and management, both things that aren't taught in university. And yet, most of the attention is on the syntax instead of rhetoric, which tells us whether we understand the problem well and whether this is the right problem to solve relative to the high level business objectives & core values.
I’d assume almost every IDE with linting will pick up incorrect indentation in Python nowadays. And if you’re writing code in a text editor with no linting, it wouldn’t pickup on the equivalently missing semicolons or unclosed brackets either.
The problem isn't "how well can you write python with or without an IDE?" it's "how well can you tell that there's a whitespace-sensitivity error in python code that was written at some previous time that you're trying to debug now?"
@@InternetOfBugs No - the question is "Know your tools", and being aware of builtins like "tabnanny" in Python itself, or "set list" in vim, or any python linter, or ...
Yeah, see, that doesn't help. Not even a little bit. You're talking about "how, when writing python, can I make it suck less?" - and I get that, but I don't care. My question is "how, when trying to debug python that was written by someone else, who I don't know and will never meet, who may or may not have used any given tools or combination of tools, do I try to figure out what the hell they were trying to do, why they were doing it that way, and how I need to make it do what it was supposed to do, when it's clearly not doing what the customer wants?" Like I said in my "Clean Code is Bad" video. I don't give a #$%&#$ how easy someone believes something is to write or to extend. The test of whether or not a language, framework, pattern, or architecture is maintainable is "how well it can be understood and debugged by someone unfamiliar with the codebase".
@@InternetOfBugs No - I'm not talking about "making it suck less"! I'm talking about finding the bug. One of the "features" of Python is indentation-importance. As a newbie to Python, especially coming from other languages, you should be aware of that, and you should find the tools to help you fix that bug. I'm also long enough in the tooth to not give a f**k about who wrote it, or why they did it this way or that way. It's my code now - because it's staring at me. So I need to find out how to fix it. But the other cliché I learned about Python is "batteries included", so the immediate question is "which battery?" and "how do I find it?" I suggest you extend the test slightly, to "how well it can be understood and debugged by someone unfamiliar with the codebase, but knows how to learn it" P.S. I admire your commitment to replies on comments, a good deal above-and-beyond by comparison with most creators I'm subscribed to
So you load in the poorly formatted code someone else wrote into your IDE and then what? How does that help? So, in my case, where the (pseudo)code was: ``` for each mailbox: for each unread message: do_thing_one() do_thing_two() do_thing_three() do_thing_four_but_accidentally_only_once_per_mailbox() ``` The IDE helps how, exactly? (Not that we had Python IDEs back in 1996 - but that's a different problem)
The funny thing about MLST is that the thumbnail you showed is my favorite Deep Learning textbook! “Dive into Deep Learning“ is also a great (practical) DL book 😅
The GPT4o also improved the syntax/grammar correctness of non-English languages (eg. Icelandic, which is the second priority for OpenAI). It was a massive improvement from the 4Turbo
I'll take your word for it (I mean that sincerely - I know that phrase is often sarcastic, but I mean it in this case). I'll have to think about how that falls on the scale from "the pseudo-reasoning abilities are better" to "the embeddings are better"
@@InternetOfBugs tbh, it does tend to drift into tourist information both in content and tone, and I haven’t thoroughly tested it to its limits. The pseudo reasoning vs embeddings scale: please make a video on that if you’re so inclined, I’d definitely be interested in hearing more!
Based on some of the comments I'm getting in the video I released today, I really do need to explain how to differentiate between embeddings and the model. That's going to take a lot of research so I don't end up saying something stupid. But it's in the queue.
@@InternetOfBugs awesome! I’m just finishing my BA thesis on limits on AI understanding, your perspective has helped me clarify some things for myself primarily. Thanks again 🖖
Hello. In the perplexity part of this video, you mention: "It gets me in the mindset of asking and not searching, and that's great", and then "perplexity is doing a really good job of helping me get over thinking about things in terms of web searching and thinking it in terms of what I wanted in the first place was to get the question answered". Maybe I missed it in the video, but could you describe why this is great? Is it more of a time save? Is it better when thinking of steps in terms of literal programming? I believe I understand the difference, but I don't understand the benefits.
See the graphics at: 26:31 (or ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-pnMx3fSXEEg.html if that timestamp doesn't work). I can (and did) ask Perplexity "Did Steve Jobs ever mention radio in his speeches?" and not only did it give me a link to audio of the speech I wanted, but pulled out the specific quote I was looking for. ChatGPT-4o said "there is no specific record of [Steve Jobs] addressing radio in any significant detail" in response to the same question, which is (obviously) not true. Google gives me a ton of Steve Jobs quotes, but most of them are SEO spam "Top `n` quotes from Steve Jobs" lists that are full of popups and ads, each result page has mostly the same quotes as every other result page, and none of them (on the first two pages of Google at least) contain the one I was looking for.
@@InternetOfBugs Thanks! You're right; no matter how I search it, that exact quote does not show. I guess SEO has gone too far and looped back into making Google not the best for specific searches at this point. I thought Google would work fine since the video only compared Perplexity and ChatGPT-4o, but I'm dead wrong on that.
@@-est6908 I would have expected Google to have found it, too. But a google search, for me, for ( "steve jobs" "radio shows with pictures" ) only turns up two links, one from "sweetstudy dot com" and one from "transtutors dot com" and they're both clickbait "stuff paragraph after paragraph of every speech you can find from every famous person you can think of onto a site full of ads and wait for the AdSense money to come rolling in" sites (that I refuse to link to, because they don't deserve the traffic).
I hope this comment get past your RU-vid pet project 😂😂😂 Thanks for introduction of the "literate programming" paradigm. For a hobby programmer like me, this is very relevant. And I found fascinating with Donald Knuth demonstration of literate programming in his book. Like he was talking to the computer himself. There is a kind of romantic nostalgic feeling in that. Considering it was in early days of personal computer. The man is a genius.
I'm stuck in JS/TS land for more than 14 years, being fearful I won't be able to learn a "real" language. This is a really interesting perspective and gives me motvation!
Stack Overflow users are deleting or messing up their answers upon the news they were used to feed an AI. Any alternative you know about? BTW, I am not a programmer but I need to know code to work (Pytjon being the main one, but I also have notions of Java)
I hate to tell you, but stuff like Stack Overflow is a bad idea - and every generation has to learn it for themselves. The way that Stack Overflow is stealing everyone's content to profit from it now has happened over and over before. In 1996 there was "Experts Exchange" (whose website could also be read "Expert Sex Change"). They provided a forum for lots of questions and answers over many years, and them locked them behind a paywall. Laughably, that was reportedly at least part of the inspiration for StackOverflow to get created in the first place. (source: en.wikipedia.org/wiki/Experts_Exchange ) So I don't know of a StackOverflow alternative at the moment. Someone (or several someones) will inevitably use this as a reason to make a new one. One of the new one(s) will become the most popular, it will slowly gather enough content to become useful, and then it will screw over its users, and the cycle will start again.
19:50 listen to that advice it`s the only thing that worked for me dealing with algorithms and i been crunching through kg of books to learn them , now when i have to use them only this small flashes are are what remained from the learning process yet they suffice to form a road map in a decent amount of time then i go back to the books refresh them a bit
With brain types like ours: You cannot be interrupted even for a moment when in the zone. 👍 Lex Fridman interviewed a guy who said something to the effect that, “no I cannot give you just five minutes of my time, real quick, et crtera”
i started programing with my comodore64 but i am not a programmer. I am actually working on an image set of UFOs to capture flying objects with a python script. I have the code for the script and images but my image set needs work. I can send it to you if you want to make a show about it and provide the UFO community with an ai signature to capture flying objects.
amazing talk, have been doing software for > 10 years and can also agree that the hardest part about it is figuring out what is the best to do, syntax is just a part of the tool. The only thing I do think is sort of different though are paradigms like let's say async vs concurrency/parallel programming. I'm not so sure you can easily translate A -> B with an AI tool when touching this topics. Anyway thanks for your thoughts man, also fuck python
Do you all think that one can be a good computer programmer / SWE without a stem degree? I feel like I don't have a hard science background without, at the bear minimum, having doing the calc sequence :(
Solidity as in the language for Ethereum's EVM? I am pretty sure that's the winning language in smart contract development still. What would replace it like a zkevm language?
The book was called "App Accomplished" It's 10 years old, and out of print. It will never be reprinted 'cause most tech books have a short shelf life. www.informit.com/store/app-accomplished-strategies-for-app-development-success-9780321961785 You can read the first chapter and ToC/Index for free here: ptgmedia.pearsoncmg.com/images/9780321961785/samplepages/9780321961785.pdf if you want to. Blog hasn't been updated in years, but it's here: escortmissions.com/
Can you please elaborate on your comment about how languages won’t exist in the future? Do you mean that we will purely be programming in natural language? If so aren’t coding languages meant to create abstractions that can’t be achieved by simple English?
I never said that computer languages wouldn't exist. I said that it won't matter which one you use. Computer languages are becoming interchangeable with each other - the way I explained that I sometimes write code in Ruby (a language I like) and ask the AI to convert it to Python (a language I don't like writing in).
I'm learning python for ml too and coming from ruby this language seems to have a lot of weird hacks. Like using double underscores to have "private methods", having to implement .__len__ instead of just .len, and the the fact that it is somehow acceptable to write combined methods like groupby instead of group_by. Now I have to remember what style of naming every particular method uses
Ha, I'm actually with you. For my Python is like a large version of the lua scripting language. I have a difficult time getting comfortable within it. I'm also getting deep into the whole ML scene these days, but decided to go with Rust for everything as well... I loave Rust. There's quite a few good crates out there nowadays for this stuff such as rust-bert, candle, burn and others that are rivaling what's available for Python nowadays if it helps any. Good luck on your journey!
Can you upload your .py to Github and share the link, pls? I would very much like to see how you programmed it to filter out comments on this platform, and how you utilize that to improve your videos. Thanks.
Probably not. Certainly at least not soon. It's just an experimental sandbox that wouldn't make any sense to anyone else at the moment. Maybe at some point it will be worth cleaning up and releasing, but I'm not going to promise that. Sorry.
this was so great, thank you :) 💯re: python debugging, a tradeoff for fast writing (pre AI I guess). Have you dug into rust at all? As an older programmer myself, it feels like a great blast from the past, like it would've been very natural to learn rust right after C (and to a lesser extent, C++).
