Software Engineers Discuss If Devin Will Replace Coders 

Join our free coding community :) - www.skool.com/software-developer-academy

The issue is not about the AI replacing coders, the issue is that it will give one engineer the ability to do the job of 50 coders, thus making 50 jobs redundant.

I've been chatting non stop on my channel and it's wild. A few things that I've gathered, is that the 14% is based off only a small population of Python repositories (if you look at the SWE bench paper originally posted from Cornell). They say "solving issues from 12 popular open source Python repositories". That's a huge red flag to me and does not represent real issue solving. I'm making this up, but I would imagine like 80% of repositories out there are private, and corporate code bases have not been explored. I feel like open source would have a better track record for documentation (because that's the easiest entry point for open source devs), so that would help Devin too. I think Devin has a super long way to go, and also there are many outside hurdles of whether it can be adopted, especially company buy-in that would satisfy security concerns, intellectual privacy concerns, and audibility. I think those alone would be issues that prevents real adoption by 5-10 years.

Hey guys, really enjoy the community, thank you for organising it!

You guys make me feel calmer about my future as a programmer. Thank you.

A common argument is that AI won’t replace _ALL_ software engineers. But it’s a silly argument. Of course it won’t. Yet. But if it even replaced 20% of software engineers, that is absolutely catastrophic for the job market. Think it’s bad now? It will be worse when the supply of 20% of jobs are permanently gone.

Okay now this vid is exactly what we needed in this time. Tim you've definitely been working out, good stuff.

Great conversation, guys. Lots of salient points, and interesting opinions.

This was great you guys gotta get together and do off the cuff stuff like this again

most fail to realize most people aren't CRUD programmers

a good conversation with good friends :')

8:55 "nobody else is fixated on AI replacing [their] job" Artists, Actors and Voice Actors, Animators, Photographers, Designers, VFX, Film Makers, Musicians, Writers and Copywriters, Editors, Content Creators in general, etc have entered the chat. (It's already happening). The outstanding litigation surrounding the training data from artists and writers? The SAG-AFTRA strike?

Honestly you guys should make a podcast... It would be absolutely future proof... Amazing and will inspire and attract a lot of talents to and for your future companies

More of these kind of podcasts please 👍

It is just like self-checkout in where one person is only supervising what maybe 8 or 10 people used to do!

Watching this now I feel silly sitting around in my bedroom because I don't have a little thing to hold

I’ve been using Aider to help me make simple shell scripts and other such tools that I use in my work and play, which normally I wouldn’t have the time or resources to make.

This was really cool. I love this format. 👍👍

Tech workers are their own worst enemies. You automated/coded yourselves out of a job. And Devin is just the first iteration of these things. More advanced versions will get developed

Okay why no one is questioning how they achieved 13% by running on pre existing model which just can do 4%. Answer is SWE bench mark dataset is made for AI models not for AI agents. That means its not meant for iterative improvement for a problem by multiple prompts and debugging. There is video by Volo. Check out his video. He explains it well.

damn Tim got guns! hugging them selves

Funny and great conversation!!

The point is the building blocks are there now. Each step in the loop will get better. In terms of productivity, It's somewhat like replacing hand tools with power tools when building a house. There's still a lot of other work to do to create and deploy a complete system, and that work will require humans. For now.

Intelligent group of guys. 👍

Someone say AI will replace everyone's job, that is mostly come from AI's creator, their main purpose is actually to get more funding, so they talked about this replacement in public in a purpose of attracting investors. Investors may be easily get influenced to pay more, to invest more on the AI product in order to get rid of developers that cost the majority of the investor's money. Because of the notion they get from AI's creator, they pay more on AI products and then they pay less on developers. Whatever the AI product produce whether it is a success or failure, they will think it later, but the first thing they want is to do cost cutting.

Hey @kennygunderman brother, more like these!

The 14% metric is worth more than its percentage value. The reason being I that the percentage will get higher, and on more complicated tasks. I don't think companies will fully trust tools like Devin when they consider their bottom line. It's too risky to have tools like Devin be running without supervision. (I believe)

I think it’s a commitment to learn software engineering. There are hundreds and thousands of students learning this skill as we speak. I don’t think Kelly who works in admin cares as much about robots

Many points, but about you not seeing anyone "replacing" programmers - you probably wont see it it directly in the future either, because what will happen is, that existing programmers simply become more productive using AI, which means companies don't need to hire MORE programmers even if the workload increases. AI capabilities increase at the same rate. So the "replacing" part becomes more like a "we dont need to hire you due to AI".

What you think, it would be good if you finish Devin course as junior?

i think design patters, coding skeleton code with API-names and objects/class definition with plain-english within the functions if given as requirement AI will get hold of it and code the flesh-part within the functions of what we want to do. This will speed up the Development cycles and out of 10-engineers only 2 or 3 will be required. Devin and other AI tools will cause industry shrinkage over period of 5 to 10 years... This seems unquestionable reality.

The real question is how fast is it going to go from solving 13% of problems to solving 95% of problems?

I have decades of very well paid experience of specifying, designing and coding advanced & complex software systems & products (non-web). Recently looking back over my career I was startled to realise that the current AI tools (under my supervision) would have been able to do most of my work in maybe 10% of the time I actually needed. This covers more than coding - the design, documentation, marketing materials, test code etc would also have been boosted. So what happens to the extra 80% or 90% of time I would have had available? Would I have been able to fill in in with second priority projects? Or would I have 'stolen' work from other developers? Is there enough work out there to saturate the available developers' capacity .. or will we run out of work for the developers, leading to headcount reduction? My guess is that maybe 50%+ of developers will have problems in the coming 5 years.

Why they dont invite Devin to chat and give his opinión?

Sounds like some hopium mixed with we cannot predict how jobs will evolve.

There was a new video on this channel right about daily life of programmers or something ?

You can tell they are software engineers just by how casually they are carrying out this discusion on a bed 😂

My mentor is here. Tim

The avengers assemble of RU-vid developers!😂🔥

27:17 😂🔥

True. More apps like Devin are already in the works.

We may predict that it wont replace but it will definetly improve productivity.

very good chat, but i think any design principal/guidelines/good practice can be modeled in a multi-agent AI coding solution. solving 99.99% of software issues is just a matter of time. it might take years to get massive adoption though.

I was waiting for your take on devin

Something that I have been questioning about the 14% benchmark is whether Devin had access to the internet during testing. Cognition AI told us that they had removed the remote branch before testing, but if it had internet access couldn't it find the PR, they are there and closed already? Perhaps I am missing something though. Guess I will not know until they decide to publish the paper.

Also the team at Cognition Labs is legendary. They have the chops to achieve their goals with Devin.

"The world would just completely change if everything was automated and I think everyone is just a little too concerned about that." Yes. Thats exactly correct and what the goal is. They call that "the Great Reset" or "Agenda 2030". You're thinking in terms of how things work TODAY. Which is where a lot of this coping comes from.

gyatdamn, Tim Daddy is jacked!

Where is Ben Awad?

This video is 🔥

The last part😂

Programmers arent the only industry worried about AI as it regards to job security. I work in the film biz (post production) and everyone is concerned. I also have a brother in law who is a lawyer. He is slightly concerned.

Open Interpreter which is FREE AND OPEN SOURCE can and does search for solutions on the internet by itself.

14% now, 30% soon

My theory is many people are going to have to move to either the manual labor industry or service industry. That will expand the lower class and shrink the middle class. I don’t know if companies would share their record profits from cheap software production with their employees

Can Devin out perform Claude or use other LLM's?

We are afraid of ai as developers (as a junior) because nobody is hiring juniors anymore. Because HR thinks they can replace us already.

A lot of the hype around AI reminds me about the hype around outsourcing software development to low cost locations in the early 2000s. Corporate managers trying to squeeze costs thought it was a panacea and retrospect it wasn't. Ultimately, in the long run, you get what you pay for. The same applies to AI...it will just take sometime for the market to realise this. In the same way that greedy outsourcing firms used too many junior engineers, leaving organisations heavily front-loaded with technical debt and unable to adapt to market conditions, AI generated code will have similar problems.

of all the AI videos popping up recently, this one describes scenario the best.

its like giving a non trained person to perform eg Appendix removal (appendectomy) surgery then give him tools and the procedure and you expect that person to deliver. That's AI on nominal things

YES

Two beds, four people ! Do you guys sleep in pairs ?

"at least not right now" isnt the future of ai and programming the whole topic of the video?

coding by humans will be reduced or expired

I mean if people don’t need to be coding then won’t the game be about who has the best product produced by these code machines?

I don't think that it will replace our job any time soon, as for a new project the client doesn't even know what they want, how will they input it in a prompt to create a solution. Can AI be a UX/UI engineer, be a project manager, Scrum Master, Project owner etc. I think we should just embrace AI to make our job easier my helping us as software engineer to create solutions for a specific problem

When you observe at 15:40 in the video nail bitting starts, something is wrong.

They seem to be saying we shouldn't be concerned based on what it can do today, and aren't really looking at how fast it has gotten better. I mean it's barely been a year since ChatGPT came out.

Devin cant provide critical solutions which is not in any database also cannot create new things as libraries or technologies but yes it can surely replace avg SWE

Dude said: I’m not gonna worry about it until it’s already happening and it’s too late. lol. Good move.

Thanks guys but why on bed??

Calculator, and computer didn't replace accountants. They were just tools for accountants. AI won't replace software engineers.

Base on the learning curve with AI tools You can learn faster. I remember the relatively wasted time doing research by the time I developed my code. I forget how I resolved a specific issue 1 month ago. 😆😅

Melkey Should be Here.

Ah yeaa its only fear mongering.... These AIs need someone smart with software experience to use, I definitely agree. But us entry level people going into the industry need to gain experince somehow, if these AIs need people in the future to tell them exactly what to do, and we don't get any experience then what is whole point? The assembly line got automated decades ago, and the only way to fight it was to become more educated. Now what?

0:39 Do you f*cking homework. 14% of a bench mark!

nice video but why are you in a hotel room why not do this at a park or somewhere

Soooooo it's only a few months old and it already is better than many of you guys in your 1st few months? ( you don't compare it now to senior engineers, that is ridiculous) At the root of it all, the higher level, human friendly code that you all have mastered has to be translated down to machine language. Intrinsically then , in the future the machines will know how to best execute tasks on themselves and find methods that currently elude us. Coupled with clear instructions about the business process/ value proposal as well as the ability to to tie in to user data in real time and accumulated knowledge of human behaviors and preferences, I see it eventually being the automated executive stack no different than 3D printers can replace material carving and machining.

Programmers are already automated

these guys are all juniors, sw eng going to google for solutions?

Who's going to debug -- or even understand -- all the auto-generated code?

I think it also depends on us Software Engineers, if we decide to fight back against AI like artists we probably can slow the progress of AI in coding down even if it's theoretically possible to create an AI Software Engineers that are intelligent enough to replace most humans we can get into legal/copyright battle with the likes of open ai, nvidia,cognition that may take years even decades to settle, I personally think that there is no evidence that it's theoretically possible to build an AI with the current technologies that is intelligent enough to steal the jobs of human Software Engineers but that doesn't mean we need to sit down and not fight back against the likes of open ai and nvidia like the artists are doing, cause they're attracting a lots of investments from foolish clueless rich investors and are putting a lots of junior/graduate/student Software Engineers in an unnecessary trauma, we need to fight against these blood sucking parasites and deprive them from their most important source: Training data, we have the power to do that, their stupid chatbots are nothing without our code that are using for training, we can stop it if we really want to.

Software engineering people, would you send your kids to college to study computer science?

There is a reason why many countries have been trialling UBI for the past few years - so many people are going to be out of work.

I mean to say ai wont replace software engineers is ridiculous honestly. It is true that it will replace most other jobs as well. But I don't remember the last time I wrote a full line of code myself. I only write comments and method and class signatures and tab tab. Of course I review all of it and there is more to our job than just coding but I am 100x more productive eith ai in my ide. The big question for the industry is if we are all using ai and use natural language to code going forward, what will happen to the quality of products. It is easy for us experienced engineers to review and find faults in generated code but I can totally see an engineer in the future who never went through actual coding and having to figure things out on their own, and is faced with a problem llm is simply not giving the solution to. Are they expected to keep prompting until it works? I agree most of actual learning we do happens in our jobs and at least to me, seems like that part will go away. So my prediction, ai definitely replaces most programmers, juniors at first. Seniors will stay for a while because they will be the most efficient using AI. Workforce is cut by over 50% and over time seniors die out and the quality of software goes way down

Suggesting that "SPENDING MILLIONS IN CODING" is "Normal" and "Should be expected", will sound INSANE in the near future.

Good luck with QA. In the end it all comes to user experience and quality control. Even with tools like devon you need people to make sure everything fits together as it should. I dont think companies would take the risk and give all that quality resposibility to a couple of folks and ai coding machine. I wouldn as i do not trust ai. You will never see me in ai controlled machine

Anything can be automated. Fear is dumb, just do what you want to do. There is a kid somewhere building and working on AI right now and they will be the next Elon. Just do what you want, life is short.

Conclusion: Cognition should hire a jr. developer to create a more professional looking website for themselves.

Something Devin didn’t let us know was it was 14% of 2,000 GitHub issues Devin solved

11:50 very bad take , everyone is a junior in a different area , if i started Rust today i would be a junior i would be able to pick it quickly than someone who just started but still in the world of AI that won't mean much

Programs are like math equations. They are precise and tell the machine exactly what to do. This is not the case with Humans as the language we use doesn't convey the exact same thing which we were thinking in our head. Machines would face the same issue.

Well these people are python engineers which was designed in a way that anyone could code same as AI, so just one ster before prompt engineering. So we had software engineers > python coders > prompt engineers. It's just the next simplification.😂

It's not like programmers lost their job directly because of AI but because fewer devs can handle the workload so some were let go and juniors are not hired. There you go. No-one will rub it in your face.

Tren With Tim. Bro got the Guns!!

"You know, you could ask all those questions to an AI, and it will provide you with answers. 🤪🤪"

So the real question is. I love coding and I want to do it for a living… but I also have 3 kids. So do I pick something harder to replace like continue to be a mechanic or fully invest in software development.

TBH a "developer who is really, really good" but who cannot communicate or understand business needs, is in reality NOT "really really good". If you have the opportunity, check out the engineers at or close to the Cxx level of a high tech. They are coding gurus ... and know history, philosophy, and many other topics. They also have energy, vision, creativity, business awareness etc. They also have high IQs and are total self confident. THESE are the engineers who will survive - and in fact control - the arrival and use of AI.

All these videos miss an important take , coding is the first to get replaced because the people making the AI's know how to code so its easy for them to imagine creating an AI that can code , if lawyers where making these AI's it would replaces lawyers and so on

I don't want to be jobless due to AI and I don't want coders to be replaced