This AI steals programming jobs... let's try it
2024 ж. 16 Мам.
3 744 Рет қаралды
GPT Pilot shows the future of software engineering.
Here is the app it built for me: typing-speed-gptpilot.vercel....
⚡️ Learn to build AI projects with me: lastcodebender.com/bootcamp
Use code YT30 for 30% off
For free stuff:
- my AI developer beginner roadmap: lastcodebender.com/ai-developer
- my twitter: / zaurbekstark
#gptpilot #devinai
Codebenders, some important notes: - Multi agent systems are slow, and GPT Pilot is no exception. It does take many hours to complete tasks just because of all the agents working in a chain. - The good part though is that it's not active time that you have to spend doing something. I just had GPT Pilot open on one of my monitors while I was doing my other tasks. It was there running in the background, so in terms of actual productivity, it didn't take me much time except a few distractions. - Cost is another consideration. The longer the agents run, the more expensive it will be. You are paying for each GPT-4 token it uses. Conceptually the way I look at it is like having a junior developer who works for you and you just need to review their PRs. - This means that it's not gonna be the fastest. - And on its own it will only be able to solve simple tasks for now, unless you start assisting it more which then based on how much you help, it will be able to solve more and more complex tasks. - It also means that you can't just use it to replace your job easily, right now it can replace some tasks of your job because you can give it an existing code base and it will work in it. But if the codebase is too big, that might not be an option. What I think is more viable right now is using it for some simple freelance gigs, building MVPs, etc. that you can complete in parallel to your regular tasks since it will be running in the background. This is where I see the future of software engineering going. We will basically be overseeing the work of many agents in parallel. This technology is evolving fast, agents are getting faster and more performant. It will be interesting to see where we are a year from now, especially depending on when GPT-5 comes out. Sorry should have added these points directly in the video and made this more clear, will try to do better next time! Thank you for watching!✊
i saw the price point and noped right out of it.
Yeah price goes high pretty quickly, that's the problem with these agents This is just the beginning though, like every tech, with time, price goes down and performance goes up
We'll need more "Devin Feeders" in the future (lol), directing what to do and finally testing the product.
If AI ends up automating that part too, then there is nothing left for us. We will just dream about an app and they wil build it for us🤣
@@thecodebendermaster - Well someone's gotta feed Devin when it gets hungry. Time to go back to farming. heh
These AI Bots know more than a Human... But dont know how to apply things to make it interactive until we command them
They need human assistance, its hard to build a very autonomous agent yet, especially when it comes to code. That's why the future of software engineering will most likely be to just oversee the work of many AI agents
Not worth the try for more complex stuff. It is constantly stuck in loops. Waste of time.
yeah for complex stuff it won't work well, too expensive for now. I think these agents will become really useful once GPT5 comes out
I have been a software engineer for about 9 years now. And this stuff doesn't worry me at all.
In the near future, I agree with you. Even Devin that got so much hype, is no where near to replace any engineer right now. However 5-10 years down the line, I think the conversation will be very different
I looked into the apps it created in their own website, the apps are very basic and took almost 6~7 hours to complete which is absolutely useless in this matter. I can see potential but you are giving the wrong advertising here for someone who knows some knowledge about sw engineering.
The time is one of the big drawbacks yeah, I should have mentioned it more, thanks for the call out. All the muli agents systems have that problem. The good part though is that it's not active time that you have to spend doing something, I just had GPT Pilot open on one of my monitors while I was doing my other tasks. It was there running in the background, so in terms of actual productivity, it didn't take me much time. The way I look at it: it's like having a junior developer who works for you and you just need to review their PRs. This means that it's not gonna be the fastest, and on its own it will only be able to solve simple tasks for now, unless you start assisting it more which then based on how much you help, it will be able to solve more and more complex tasks. This is where I see the future of software engineering going. We will basically be overseeing the work of many agents in parallel. I do sometimes get ahead of myself and imagine that it's all very close haha, but yeah right now the level it's at is closer to just doing some simple freelance gigs for you and anything more advanced you need to start spending more of your time assisting it/fixing some issues. A year from now, we might have a very different discussion though, especially depending on when GPT-5 comes out.
Just added a pinned comment with details about this btw, thanks for bringing this up 👍