https://gitclear-public.s3.us-west-2.amazonaws.com/Coding-on-Copilot-2024-Developer-Research.pdf """ ...GitHub's research finding on this point says "developers are 75% more fulfilled when using Copilot." To a first approximation, developers embrace the product. This doesnât reveal whether their near-term satisfaction will be shared by those who go on to maintain the code. Initial impressions from longtime code researcher Adam Tornhill (author, Your Code as a Crime Scene) are skeptical: Developer researchers are concerned by the impact of AI assisted programming. GitHub claims that code is written "55% faster" with Copilot. But what about code that shouldn't be written in the first place? The problem here is that code spends 10x more time being read than being written, according to Robert Martin, author of Clean Code: A Handbook of Agile Software Craftsmanship. Writing bad code faster implies considerable pains for the subsequent code readers. That is the first of many challenges facing developers who use an AI assistant. Others include: 1. Being inundated with suggestions for added code, but never suggestions for updating, moving, or deleting code. This is a user interface limitation of the text-based environments where code authoring occurs. 2. Time required to evaluate code suggestions can become costly. Especially when the developer works in an environment with multiple, competing auto-suggest mechanisms (this includes the popular JetBrains IDEs [11]) 3. Code suggestion is not optimized by the same incentives as code maintainers. Code suggestion algorithms are incentivized to propose suggestions most likely to be accepted. Code maintainers are incentivized to minimize the amount of code that needs to be read (I.e., to understand how to adapt an existing system). These drawbacks may explain the difference between the greater tendency of Junior Developers to accept code suggestions compared to their more experienced counterparts. According to GitHubâs research: GitHub's own data suggests that Junior Developers use Copilot around 20% more than experienced developers Experienced developers have the most informed understanding of how costly code will be to maintain over time. If they are more averse to using AI suggestions, it raises questions about the extra code that Junior developers are now contributing, faster than ever? """
Well yeah totally. Itâll be like farming. Embrace the technology or get pushed out. The few that are left HEAVILY embrace tech advances. The combines now will sow a field basically automatically. Older folks naturally think that âx or y could never replace the quality you get with handmadeâ. Edit: to be fair, they might actually be right. But, you donât need better⊠you need good enough⊠and you need cheaper. Think of it as the ultimate outsourcing.
Being able to add remixes of existing functionality isn't sufficient to replace an experienced developer on a project. Scaffolding for build and test is the best use-cases I've seen for this tech. Writing tests is the worst use-case, truly it's garbage here and unfortunately where juniors struggle the most. A novel feature or anything outside of the lines requires the experience of a senior who has spent time learning the problems associated to a well-posed question. Juniors don't pose the question well.
Right now⊠you arenât seeing the bigger picture. Theyâre getting hand-tuned and trained to be better. This technology is advancing exponentially.
This research makes me think that this fad will wreck tech and depress the swe market in the short term. In the long term the SWE market will be much more in demand because people aren't learning coding by using GPT and those apps will eventually collapse under bugs from generated add-only cruft.
What an utterly stupid conclusion to make from that article. Older developers are going to use AI to help code. And they'll be better at it than Junior developers. The key is to understand what AI is outputting which many Juniors don't. People who use AI to code and don't understand what it's outputting are the same types of people who copied and pasted from stack overflow.
I used to think like this, but honestly, itâs called artificial intelligence and itâs worth actually thinking about what that means I pay $20 a month, and it really does feel like Iâve bought some kind of cat-level IQ to work on things for me. And there are cases where I can just trust the output Copying SO was generally bad in the previous tech era because the suggestion wasnât actually for your problem. Yuge difference here
You sound so close to getting my perspective when you say that "the key is to understand what X is outputting which many juniors don't." Based on the data AI isn't turning juniors into seniors, it's giving seniors more work and polluting code.
^ good for our job security for two reasons * Always work to do * Young kids will never hit senior level skill because they need statistically driven, spoon fed answers.
Iâd love to use copilot but only get to use codewhisperer =(
Wait, GitHub's study found that GitHub Copilot makes devs happier? Incredible.
That explains all the Tesla drivers that can't drive.