Gemini is the end of SWEs
Its so good in coding, explaining code, and finding bugs. It can parse and keep context for 100k+ tokens. This is insane.
What about parsing requirements and new designs? Things that SEs should be doing anyway beyond just coding.
Are you kidding ? It is no better than a person who just knows how to write hello world. Over time it will get better but not as of yet
Any good software is broken into small parts with as little complexity as possible. For an AI to assemble these parts into meaningful functional units is not difficult. The software engineering role is on the precipice of obsolescence.
@Bloomberg, Who is going to break it down to the smallest part?
So who will ask it what to code, who will validate the code, who will deploy the code, who will fix the live site issue, who will figure how to take the current code in production and generate new code to make the next version of the product?
Some sort of meta-engineer. But I suspect many if not most current “engineers” (code monkeys, really) are not well-suited to this more abstracted role.
GPT-4 is better
Its not for big codebase
Yep. And the remaining work—breaking down requirements to adequately-sized and scoped prompts—is monkey work. The future of software development involves just a handful of low-skill humans.
Have you used Gemini before? It is so bad at X86 Assembly and debugging C++
You write x86 at Google?
Good at Rust though
How big are your $GOOG losses?
it will probably just expose work that is actually easy figuring out what it actually is that needs to be coded and communicating that is usually an iterative process without a clear “correct” solution.
Tpm now: 🤩 😎 Tpm future: 😭 😑
Okay. Good. I'll have more time to do other stuff.
Youll have more time unemployed
Any luck, AIs doing more stuff will cause some deflation and my "unemployment"/"retirement" will cost less.