is there going to be any real invention anymore?
>>108791661The remaining 10% of people will actually invent and contribute (unless they abandon for greener pastures - literally). That 10% figure nicely fits the square root law (Price's law).
>>108791661If 90 percent of programmers VibeCode, it will be the end of quality and probably will cause a massive blackout.
>>108791661Yes, by leveraging ai hallucinations
>>108791661>think of invention prompt>the ai invents somethingwow amazing concept so hard to understand
Even before AI most people in most industries were not inventing.You need a fanatic that has passion and actually cares to make a difference and they are less than 1%.
>>108791661ideally it makes implementing papers etc. easier so opensource programs finally have the man power they need to be massively enhanced. Side effects are a problem so I guess functional programming will gain more value. So yes real invention can happen if you leave the programming to the machine and deal with the formulation of the algorithm and system design yourself. So far I have found it to be rather disappointing because none of the actual useful stuff for scientists have been implemented by vibe coders.
Using AI to write boring ass shit doesn't mean AI build it. It's like auto complete on steroids.
>>108791661you can invent 99% of things by recombining already existing parts and concepts
>>108793213commonly repeated falsehood from redditors
>>108793273Can't imagine someone who knows what he is doing just vibeslopping things without checking and guiding the ai
>>108791661There will be a flood of slop apps made by the Product Manager types. Low-level optimizations will slowly disappear because nobody will understand how anything works in the low-level anymore to even consider. Humanity is doomed if the trend doesn't swing back.
>>108793318Yes in theory but in reality it generates so much text that the "senior who knows what he's doing" is inevitably going to be lazy and just skim through it.
>>108793340Jonathan Blow talked about this in his Preventing the Collapse of Civilization speech. The general consensus is that the problem won't matter because the AI will worry about all the low level stuff so we don't have to. Just tell it, "Okay good now do a lot of low level optimizations to make it run faster."Sounds dumb, but already if you present a working custom software that you wrote to an LLM and then give it that optimization prompt, it might suggest working optimizations you never considered, if you've never given much thought to optimization before. I am of course playing devil's advocate right now.
>>108793842>the AI will worry about all the low level stuff so we don't have to.This is true though. Let the computer worry about computer stuff. It's faster, smarter, and more accurate, than all of us put together.
>>1087916612016:10% of programmers actually know how to do shit2026:10% of programmers actually know how to do shit
>>108792195coding agents are better than 90-95% of coders
>>108793842this plus rustliterally all you have to do is tell an agent to do profiling and it will make optimizations 90% of programmers will miss and if you have it spend a lot of time on it it will get ones that 99.99% will miss
>>108793842>JBLOWSHe also says that LLMs are not the AI for this, but machine learning. He uses chessbots as an example.Even then, maximally optimised code will be blackbox direct to binary and you'll have no fucking idea what it is doing, which is also bad. Then you'll have to trust some LLM to explain what it is doing, which it will hallucinate. Turns out low level human written shit was the king all along, we just don't have the opportunity to do it, which is where degeneracy and slop thrives.
>>108791661there have already barely been any real inventions in the last decade anywaysaccelerate towards the end
