For the best experience on desktop, install the Chrome extension to track your reading on news.ycombinator.com
Hacker Newsnew | past | comments | ask | show | jobs | submit | history | more karpatic's commentsregister

I did the same exact thing and then wanted to post it into the groupchat then see this.

1000 iterations too!

Why are we like this XD


What model are you using for this? I can't help buy imagine this gets expensive


Great! I wish there was a "bang to buck" value. Some way to know the cheapest model I could use for creating structured data from unstructured text, reliably. Using gpt4o-mini which is cheap but wouldn't know if anything cheaper could do the job too.


Take a look at Gemini Flash 1.5. I had videos I needed to turn into structured notes, and the result was satisfactory (even better than the Gemini 1.5 Pro, for some reason). https://jampauchoa.substack.com/i/151329856/ai-studio.

According to this website, the cost is half of the gpt4-o mini. 0.15 vs 0.07 per 1M token.


Seconding Gemini flash for structured outputs. Have had some quite large jobs I’ve been happy with.


I haven't found a model at the price point of GPT-4o mini that is as capable. Based on the hype surrounding Llama 3.3 70B, it might be that one though. On Deepinfra, input tokens are more expensive, but the output token is cheaper so I would say they are probably equivalent in price.

Also, best bang for the buck is very subjective, since one person might need it to work for one use case vs somebody else, who needs it for more.


I always plug openrouter.ai for making cross-model comparisons. It's my general goto for random stuff. (I am not affiliated, just a user)


I love the idea of openrouter. I hadn't realized until recently though that you don't necessarily know what quantization a certain provider is running. And of course context size can vary widely from provider to provider for the same model. This blog post had great food for thought https://aider.chat/2024/11/21/quantization.html


To expand a little, some providers may apply more aggressive optimization in periods of high load.


Fisker continues to widen my eyes. Unfortunately.


Any clue if Latex or html works better?


My guts say that only pdf and doc formats matter. I would imagine that it’d be more likely a dev is testing outputs from word or printing to pdf from a browser.

Of your options my thought is HTML to pdf via the browser is most tested.


The langchain docs are God awful, impossible to navigate with every page half baked. It was made even worse by their trying to push lg to the point you can barely find their lc docs. And, It's like they just shove jsdoc output into docusaurus pages for the api and call it a day. The api works well but thr api makes excessive calls imo and the output is still harder to grok then setting some of this stuff up manually.


Thanks man. Basically sounds like they haven’t gotten better.

We have an advisor that’s pushing us towards langgraph but he’s the CAIO at a LargeCo and they have an entire team that’s devoted to building tooling for working with it. Which in and of itself is insane - the whole point of using a library is to avoid doing it yourself - but is also damning against LG because as an early stage startup I can’t afford the time sink of needing to work around a weird abstraction.


Hi do you mind sharing more what you mean by "the api makes excessive calls"?


This needs to be brought up. Was looking for the demo and ended up on the contact form


Funny enough, the definition has nothing to do with ancestry. "Native: a person born in a specified place or associated with a place by birth, whether subsequently resident there or not."


Yeah, this is the exact sense in which I wrote it originally - but I sensed the screams of 1000 idiots and wrote "native" to try avoid that entire line of conversation. It seems though that my efforts went unnoticed - oh well!


It was an amazing statement to read. I'm glad someone else caught that


I am aware of Hispania, etc. The thing you're perhaps missing is that 'minignape began their comment with "as a white person...".

That is usually a characterization used by non-Hispanic white people; hence my reply referenced non-Hispanic whites. I also wanted to highlight that they would probably be tolerant of an unfamiliar Hispanic white in-group at their company, but weren't tolerant of a South Asian one.


Text makes communication hard. Have a good day bud.


> And that's just false. And rude.

Sorry, I'm just going to have to go with the consensus of the majority of scientific study on this one: https://en.wikipedia.org/wiki/Native_Americans_in_the_United....

I'm sure you'll find some like-minded people that share your worldview and are not "rude".


Chiming in like this makes me wounder what your view on abortion is. I'm not asking, though.


I'm not really sure if I follow how that relates?

I'm just saying that "X is legal" is not really a good counter to "X is not moral", which of course also applies in the reverse. Legality and morality are just not the same thing.


Oh lol. I was being serious. Been thinking about the nuances of it recently and I felt it kinda resides within that framework. Wasn't meant to be snide but also this threads not the place for it.


Consider applying for YC's Summer 2026 batch! Applications are open till May 4

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

Search:

HN For You