Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

GitHub Copilot Workspace: Technical Preview

github.blog

Post reply on HN
Print view View original

271–280 of 333 posts

Re: GitHub Copilot Workspace: Technical Preview

#271

Post by neom »

SushiHippie wrote: post #268
no_carrier wrote: post #265

Earlier quoted context omitted.

I'm just as confused... If you're copy/pasting the code into a new session, isn't that reading the code?

The way I understand it: First Variant: 1. User: Asks coding question 2. Ai: Outputs half functioning code 3. User: Asks to fix specific things 4. Ai: Creates buggy code 5. User: asks again to fix things 6. Ai: writes even more buggy code Proposed second variant with copying code: Until step 4 everything stays the same, but instead of asking it to fix the code again you copy it into another session, this way, you'll…

I dunno how you SWE's are doing it, but I have my ChatGPT output files and if multi, zip files, not code snippets (unless I want a code snippet), and then I re-upload those files to new session using the attach thinger. Also, in my experience just building marketing websites, I don't do step 3, I just do step 1 and 2 over and over in new sessions, it's longer because you have to figure out a flow through a bunch of work sessions, but it's faster because it makes wwwwaaaaayyyyyyy fewer mistakes. (You're basically just shaking off any additional context the GPT has at all about what you are doing when you put it in a brand-new session, so it can be more focused on the task, I guess?)

Re: GitHub Copilot Workspace: Technical Preview

#272

Post by faeriechangling »

sixhobbits wrote: post #20
studenthrow3831 wrote: post #3

Student here: I legitimately cannot understand how senior developers can dismiss these LLM tools when they've gone from barely stringing together a TODO app to structuring and executing large-scale changes in entire repositories in 3 years. I'm not a singulatarian, but this seems like a brutal S-curve we're heading into. I also have a hard time believing that there is enough software need to make such an extreme prod…

Accountants thought spreadsheets would kill their profession, instead demand for them exploded. Compilers made it much easier to code compared to writing everything in Assembly. Python made it much easier to code than writing C. Both increased the demand for coders. Code is a liability, not an asset. The fact that less technical people and people who are not trained engineers can now make useful apps by generating mi…

Hard to know without the benefit of hindsight if a productivity improvement is:

1: An ATM machine - which made banks more profitable, so banks opened up more of them and drew people into the bank with the machines then told them insurance and investments.

2: Online banking - which simply obsoleted the need to go to the bank at all.

My inclination that LLMs are the former, not the latter. I think the process of coding is an impediment to software development being financially viable, not job security.

Re: GitHub Copilot Workspace: Technical Preview

#273

Post by lenerdenator »

singluere wrote: post #112

While I've not used this product, I've created somewhat similar setup using open source LLMs that runs locally. After having used it for about three months, I can say that debugging LLM prompts was far more annoying than debugging code. Ultimately, I ended up abandoning my setup and going in favor of writing code the good old fashioned way. YMMV

Honestly, I've found using GH CoPilot chat to be the real value add. It's amazing for rubber ducking.

That being said, my employer pays for it. I am still on the fence about which LLM to subscribe to with my own money.

Re: GitHub Copilot Workspace: Technical Preview

#274

Post by atq2119 »

api wrote: post #41
sixhobbits wrote: post #20

Earlier quoted context omitted.

Accountants thought spreadsheets would kill their profession, instead demand for them exploded. Compilers made it much easier to code compared to writing everything in Assembly. Python made it much easier to code than writing C. Both increased the demand for coders. Code is a liability, not an asset. The fact that less technical people and people who are not trained engineers can now make useful apps by generating mi…

This is what I've been predicting for over a year: AI-assisted programming will increase demand for programmers. It may well change how they do their work though, just like spreadsheets did for accountants and compilers did for the earliest generation of hand-code-in-ASM developers. I can imagine a future where we do most of our coding at an even higher level than today and only dive down into the minutia when the AI…

A crucial difference to the other examples is this. Compilers and spreadsheets are deterministic and repeatable, and are designed to solve a very specific task correctly.

LLMs, certainly in their current form, aren't.

This doesn't necessarily contradict what you and GP are writing, but it does give a flavor to it that I expect to be important.

Re: GitHub Copilot Workspace: Technical Preview

#275

Post by lenerdenator »

PoignardAzur wrote: post #51

I think Microsoft is going the wrong direction with Copilot (though it's a reasonable direction given their incentives). Right now Copilot is terrible at large changes to complex codebases; the larger and more complex, the worse. But it's great at suggesting very short snippets that guess exactly what you were in the middle of writing and write it for you. I wish Copilot focused more on the user experience at the sma…

> But it's great at suggesting very short snippets that guess exactly what you were in the middle of writing and write it for you.

I found it reasonably good when I described what code should do through comments and let it generate based on that, evaluated the output, moved on to the next piece of logic, and repeated.

Basically, chaining the snippets together. Then again I'm doing Python back-end web development, so not something terrifically hard.

Re: GitHub Copilot Workspace: Technical Preview

#276

Post by arvinsim »

specproc wrote: post #146
paradite wrote: post #123

Earlier quoted context omitted.

Locally running LLMs in Apr 2024 are no where close to GPT-4 in terms of coding capabilities.

Depends what for, I find AI tools best for boilerplate or as a substitute for stackoverflow. For complex logic, even GPT-4 ends up sending me down the garden path more often than not. I got Llama 3 8B down over the weekend and it's alright. Not plugged it in to VSCode yet, but I could see it (or code specific derivatives) handling those first two use cases fine. I'd say close enough to be useful.

Agreed. Can even get specialized LLMs like Deepsync Coder for better results

Re: GitHub Copilot Workspace: Technical Preview

#277

Post by karmajunkie »

studenthrow3831 wrote: post #28
packetlost wrote: post #10

Earlier quoted context omitted.

The hardest part of software development is not writing code, full stop. It never has been and it never will be. The hard part is designing, understanding, verifying, and repairing complex systems. LLMs do not do this, even a little bit.

I guess my worst fear is not "no more jobs because AI can code" but "no more junior jobs because AI can code under the supervision of a senior". SWE jobs will exist, but only seniors will have them and juniors are never hired. Maybe the occasional "apprentice" will be brought on, but in nowhere near the same amount. Where my blind hope lies more specifically is in networking into one of those "apprentice" roles, or m…

I dropped out of school and went into startups with my first full-time gig in March, 2000. Managed to make that one last a few years, but whoooo boy that was a tough time to be a junior-to-mid developer looking for a job. I even went back to school with plans to go to medical school (yet I'm still a developer 20 years later.)

Being a junior is rough, landing those first few gigs, no doubt about it. It didn't get any better with the advent of code schools, which pretty much saturated the entry level market. But, if you stick it out long enough and keep working on learning, you'll acquire enough skills or network to land that first gig and build from there.

I wouldn't freak out about AIs—they're not going to take all the jobs. They're a tool (and a good one, sometimes.) Learn to use it that way. Learning a good tool can easily accelerate your personal development. Use it to understand by asking it to summarize unfamiliar code, to point you in the right direction when you're writing your own code, but don't have it write code you don't understand (and probably can't, because it doesn't work as written.)

Give it a few years, things will generally work out. Make a plan to be resilient in the meantime and keep learning and you'll be fine.

Re: GitHub Copilot Workspace: Technical Preview

#278

Post by packetlost »

freedomben wrote: post #240
packetlost wrote: post #52

Earlier quoted context omitted.

I think latency is the biggest reason I killed my Copilot sub after the first month. It was fine at doing busy-work, like 40%~ success rate for very very standard stuff, which is a net win of like... 3-5%. If it was local and nearly instant, I'd never turn it off. Bonus points if I could restrict the output to finishing the expression and nothing more. The success rate beyond finishing the first line drops dramatical…

Interesting, latency has always been great for me. If it's going to work, it usually has suggestions within a second or two. I use the neovim plugin though so not on the typical VS-code based path.

I also used the neovim plugin. I'm on a fiber connection in the Midwest, so that is likely a factor. Latency was on the order of 2-5s consistently, which is way more than enough to interrupt my flow.

Re: GitHub Copilot Workspace: Technical Preview

#279

Post by imiric »

lispisok wrote: post #239
imiric wrote: post #221

Earlier quoted context omitted.

> Performance improvements have flatlined. Multimodality, MoE, RAG, open source models, and robotics, have all been/seen massive improvements in the past year alone. OpenAI's Sora is a multi-generational leap over anything we've seen before (not released yet, granted, but it's a real product). This is hardly flatlining. I'm not even in the AI field, but I'm sure someone can provide more examples. > Current AI predict…

You make claims of massive improvements but as an end user I have not experienced such. With the amount of fake and cherrypicked demos in the AI space I dont believe anything until I experience it myself. >Ironically, Waymo's self-driving taxis were launched in several cities in 2023. Does this count? No because usage is limited to a tiny fraction of drive-able space. More cherrypicking.

Just because you haven't used text generation with practically unlimited context windows, insight extraction from personal data, massively improved text-to-image, image-to-image and video generation tools, and ridden in an autonomous vehicle, doesn't mean that the field has stagnated.

You're purposefully ignoring progress, and gating it behind some arbitrary ideals. That doesn't make your claims true.

Post reply on HN