Ask HN: What do you use LLMs for?

I've come up with a few minor ideas, mostly for around the home.

I created a little web app (that I just run locally for now) where I can give it a recipe url and it will fetch the content and feed it to the LLM and adjust it to my cooking preferences.

For example, a lot of recipes will mask things as ingredients that are really steps, like "1 bell pepper, diced". The LLM can separate it into 1 bell pepper as an ingredient and dicing it as a step.

Another thing that annoys me in recipes is that the amounts for ingredients, like spices, aren't included in the steps; I hate scrolling back up to check how much to add.

I threw it together as a weekend project and it works ok; I'd like to add a small chat component to it so that I can revise recipes during or after and save them for the future, because I make mental notes of how I should make it next time but never remember them.

I find Kagi's AI summarizer[1] very helpful.

If I just want to get a general gist of an article/video, I run it through the summarizer.

For media I was only slightly interested in, reading the summary is enough. For other topics, it helps me decide if I want to invest more time in reading the whole thing.

It is also useful for helping me remember the exact contents of a video. (Say I want to share a video, but I don't remember if this is the one I want to share.)

[1]: https://kagi.com/summarizer

Sometimes I use LLMs to find vocabulary/definitions in a foreign language.

Usually online translators like translate.google.com and papago.naver.com work 99% of the time. However, for that remaining 1% Bing Copilot has come in clutch.

The LLM has the advantage of being able to query about a specific context. Of course, I always double-check the answers.

Here is a specific query I did recently:

> 영돈, in terms of spending money in korean to english

> The Korean term "용돈" translates to "pocket money" or "spending money" in English¹. It refers to money that individuals can use freely without any specific purpose, often given by parents to their children⁴. So, whether you're saving up for a treat or just need some extra cash, "용돈" has got you covered!

(I could have sworn the original Copilot answer repeated the incorrect "영돈" in its answer, but I found the correct "용돈" by looking up "spending money" or "pocket money" in the online translators.)

A lot less than the hype suggests.

I've found them useful for automatic note-taking of meetings; summarization; and that's about it.

I use it for coding a few times per month. Yesterday I had it convert some jQuery to JavaScript, that was kind of nice.

I use Kagi for search and will use their Quick Answer feature, which also provides sources, which I’ll usually click through to. It helps get me to the right page when there is more context and a simple search leaves me hunting more than I’d like.

That’s about it. None of my projects are using LLMs, nor do I have plans to. I’m a consumer and the consumption is pretty limited.

Been using llm for following usecase Big llm like Claude and gpt for coding help Copilot for coding Local llm like mistral2 for 1. Parsing LinkedIn posts to find jobs and put them in Google sheets 2. Mental sparring by giving local llm a problem and asking it to ask important questions instead of answer. This helps me identify if I have thought it through or have i missed some things

It's good for stuff you can't google because you don't know the name of the thing you'd need to google.

E.g. Learning programming, when I see something I don't understand I ask it to explain - i.e. "What do the brackets represent in the following code?:"

Great for classifying items and extracting facts from text, at least compared to anything we had 5 years ago.

Writing bash wrapper scripts. I hate writing bash but it does the heavy lifting.

I use it for technical questions I have instead of Google now.

Nothing