Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

The one thing that LLMs are good at is prototyping super fast. For accelerators/incubators etc this is a game changer since more ideas can be realized and tested for market fit.

Now if the market fit indeed exists then someone needs to rearchitect and rewrite the thing. But wasn’t it always the case? The POC always was a hacked together solution with no real viability to be used as the final product.



LLMs are good at prototyping something that is similar to something that already exists and is open sourced.

It may be that there are such projects which can be monetized or need better marketing.

Innovative it is not, however.


I suspect the vast majority of innovations are a combination of a small number of things that already exist.

Almost nobody is making things that are so unlike everything that previously existed that LLMs can't help.


I don't understand this position.

LLMs are good at prototyping using data across _all_ similar projects that exist.

It is not a 1-1 copy.

Most frontend is a dozen components. Most backend again is a handful of architectures when it comes to DB/business logic, CRUD.

It goes to say that if you can guide the LLM to build something innovative you can think of, it will put those components together in a reasonable way - good enough to start off with.


Exactly. One could build the new video platform with revolutionary customer facing features. The tech stack will likely be the same. Some frontend, some backend maybe some calls to an endpoint that happens to be an LLM.

Startups that typically end up in incubators etc are not about new fundamental systems (languages, frameworks, theoretical methodologies etc), but rather about new products.


there's only so many times (in fact, 3!) I want to implement oauth from absolute scratch


This seems to suggest a failure in our model of software. We were supposed to have reusable components. Writing the same thing more than once was not supposed to be necessary.

I recognize that in reality this hasn't always worked out. But I also don't think that the answer is a black box that can churn out questionable boiler-plate.


think about the early days of Facebook, when it was called TheFacebook.com

a bunch of PHP scripts :-D


>The POC always was a hacked together solution with no real viability to be used as the final product.

well that's the different. It is the final product, now. In many cases, you're selling the prototype itself, becuse that's what investors like seeing. User engagement be damned.

And people still seem to doubt that we're in a massive bubble.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: