Extract Web Data Easily with AI | Hacker News Reader

Extract Web Data Easily with AI

kadoa.com

29 pointst_a_v_i_s13 comments

I'm someone who regularly scrapes websites and uses bs4 to format the data. Admittedly it takes me some time to get what I want exactly, but the past year ChatGPT has _drastically_ sped up the creation of my scripts. Given that I am your target audience, why would I pay a pretty steep 39USD for something I can do myself in 30 minutes - and without limits? There is a high correlation between scraping websites and being tech-savvy enough to produce a (simple) script to do so. The only reason I'd consider buying your product is if I have to do this so often that it would save me the time of writing ONLY the bs4 part of my scraping script (which is often the easy part; websites are awfully incoherent and buggy), but that would only happen if I wanted to scrape vastly more than 25k "lines in a csv" (which is what I assume 1 credit gives you, but is never explained).

Perhaps they wish to create a new market?

Amazing. Pitched the same idea for a startup with my students back in 2020 (never got very far though as every member of the budding team got busy with other things). Was convinced we could turn the language models of the time into this back then and now a few years later it is done through API calls to a third party. We truly live in interesting times.

Nine months ago HN user genmon posted the results of his experiments to classify the BBC In Our Time podcast using ChatGPT: https://news.ycombinator.com/item?id=35073603

It's the same principle. Give the podcast text to ChatGPT in chunks, ask it to classify using Dewey decimal system, pull out guests and mentioned books, and summarise the episode.

It's funny that user nl posits that 'We are months away from being able to do this with images too' - we were indeed months away! End of September, to be precise https://openai.com/blog/chatgpt-can-now-see-hear-and-speak

One of the examples is scraping rei.com. The URLs extracted by the tool look like this:

    https://rei.comhttps://www.rei.com/product/207125/asolo-eldo-gv-approach-shoes-mens

Extract data easily, no mention of accurately.

Actually, yes there is.

>How accurate are the results? What about hallucination?

>Kadoa validates the data accuracy through multiple steps, ensuring reliable and accurate data extraction. For example, we verify that the extracted data truly exists on the source. Kadoa tries to strike a balance between limiting noise (precision) and including all valid parts (recall). While being robust, it also operates efficiently, processing millions of data records in production.

"What We Do

Kadoa is an AI-powered no-code platform that allows anyone to build complex data workflows effortlessly. We use AI to navigate, understand, and transform unstructured data from any source. The orchestrating AI agent chooses the best strategy for each task, such as where to go, what to extract, and how to format the data. We do this at scale and"

I especially liked the "We do this at scale and" (verbatim) bit.

/ Trusted by

Probably no-one.

/ Why Kadoa

Because founders need to eat, too!

/ Popular use cases

There aren't any, because this isn't actually an established business but a GPT-4 wrapper that didn't exist months ago.

The last point is just ridiculous. If you use the product for a few minutes it’s clear to see where value-adds exist such as scheduling, alerts, delegating, and having a real team to help with issues, etc.

I experience your take thus as quite cynical and missing the bigger picture.

How long have you been in business? What size is your "real" team?

Have we come round to AI-generated startups? Truly a revelation.

Some have substance, but most are simply thin wrappers with no real defensible moat. Best to let time tell what’s successful.

I tried to scrape a simple imdb actor page and it failed and required support to look at it. :/

Reply on news.ycombinator.com