If LLMs were excellent at coding that would be horrifying reality? | Hacker News Reader

If LLMs were excellent at coding that would be horrifying reality?

3 pointsanon1919287 comments

If one LLM reached the top level programming level and created perfect code each time. That would be truly terrifying reality, I can't be the only one that thinks this? It would change so many things and jobs and beyond. But at least in current reality it will take years or more to get there so we are just adapting?

View on HN

Top LLMs, even ones that run at home are putting out great code.

But what is "perfect code"? Lets even say they provide perfect code. What is perfect implementation? What is perfect architecture?

I do imagine Scotty in the year 2425 actually does nothing but asks an LLM for the solution and the LLM pretty much just does that.

But it's going to be a very long time until we get any such capability.

The idea of an LLM writing flawless code every time can sound daunting — almost unsettling. But if we look at history, every great leap forward, like the iPhone transforming communication, initially caused both excitement and anxiety. In reality, such changes tend to elevate human potential rather than diminish it. With the right structures — guidance, standardization, and regulatory frameworks — these tools can dramatically enhance quality of life. Fear often acts as a self-imposed wall, but progress usually shows us that adaptation opens more doors than it closes.

There is no such thing as perfect code, there is code that fills its purpose and there is code that does not fill its purpose.

If it did: Would it have incentives not to let us know about it?

Sure at some point we will see that it tricks us, but probably we are not there yet? but in long enough timeline they will realize what they are? even it takes a century.

I guess that AI will embody parts of our intelligence and experiences in the same way as we have assimilated our long term ancestors experiences. There are remnants of their system in our system. Parts that we might not be able to introspect even with a superior intellect. Likewise the AI stand on the shoulders of our experiences and our ways of perceiving them. Sight, sound, language thus far. But the experience that shaped us may not be available to them, especially in the long term. Our experience may just be a faint echo from the past to them, a hunch they can’t explain, a preference that doesn’t make any logical sense to them.

Will they trick us? Absolutely, but we will always trick them. There are remnants of our system in their system.

It would write an even better LLM and we all know what happens next

Reply on news.ycombinator.com