Subscribe
Sign in
Hacker-AI
p/hacker-aiA/B test your HackerNews titles with AI before publishing
Start new thread
Kimmo Ihanus

Hacker-AI - A/B test your HackerNews titles with AI before publishing

by
hacker-ai.com uses machine learning to predict the success of Hacker News post titles. If you're uncertain about what title to use, trust statistical mathematics! Pre-test your "Show HN" titles and increase your chances of hitting high points in Hacker News.

Add a comment

Replies

Best
Uvictor
It would be nice if it suggested titles....😉 ...I will be launching tomorrow ...I will let you know how it goes....if you want
Kimmo Ihanus
Maker
@uvic got that right ;) maybe on next iteration.. good luck on your launch and let me know how it went
Uvictor
@uvic @ihmissuti unfortuantely it didn't get any traction..that might be bcoz of the product though..cool product will be using it again...
Kimmo Ihanus
Maker
@uvic sorry to hear that. I use this tool whenever I just can't make up my mind which title to use. It's better to use data than to rely on bare guessing. btw. I like the idea in PRA. Coding is not my strongest skill so technical implementation is a bottle-neck for me always. Hope you get crowd to that platform
Uvictor
@ihmissuti bro thnx...
Kimmo Ihanus
Maker
We always use this method to optimize our HN posts when we are unsure about which title to choose. Now we decided to create a UI so others can use it as well. The algorithm is currently able to guess 60% of the times correctly. We are continuously developing the algorithm, but it is already better than a random guess. Try it out and tells us what you think!
Abishek Muthian
@ihmissuti Congratulations on the launch! You might be also interested in this need gap - 'A/B test social media posts without posting' - https://needgap.com/problems/174... posted on my problem validation platform.
Kimmo Ihanus
Maker
@ihmissuti @heavyinfo definitely! Commented that one in your platform.
Hamza Nouali
Nice product Kimmo, I'll give it a real test on HN. Thank you!
Kimmo Ihanus
Maker
@hamza__nouali Thanks! Let me know how it works. There's already good feedback on improving the prediction model by eg. adding readability scores to the algorithm. So I'm extremely happy to hear any experiences and then tweak it accordingly.
menajem benchimol
TITLE A: Someone should sell this- Montessori toys [4% better] TITLE B: How you can earn $46k/M selling Montessori toys Tried it for one of our ProdcutByte products, lets see what happens!
Kimmo Ihanus
Maker
@mbenchi10 coool!! Thanks. Let me know the outcome then :)
Alejandro Cantarero
Will definitely give it a try the next time I post to HN. Tried out a few variations on my last HN post and pretty much every alternative scored much better than what I used. And that post did really poorly, so guess I should test more ideas first! The web-ui you built around the ML prediction algorithm is really nice. Simple and clean.
Kimmo Ihanus
Maker
@alejandro_cantarero1 thanks a lot!
Vladimír Seman
thank you, kimmo, bookmarked. do you plan to increase the dataset for the model in the future (2018-2020)?
Kimmo Ihanus
Maker
@vladojsem thanks dude! Yup that is the plan. I'll work to improve the algorithm and also increase the data.
Christine Renee
I know it's optimized for Hacker News, but is there any reason this wouldn't work for any headline writing? 🤔
Kimmo Ihanus
Maker
@christine_renee yes i've used it previously for email headlines and such. All you need is the data. Hacker News is a good use case as there is a lot data available, and the title has a character limit. But as you said, I can feed in any headline type of data. It could be google ad texts or anything.
Robert Thelen
Very cool technology. Looking forward to perhaps expanding it to ProductHunt :). Would love a tool to let me know if my upcoming launch title and post is catchy. Congrats on the launch!
Todd Venegas
This looks great
Kimmo Ihanus
Maker
Hacker-AI updates! I've added Product Hunt as a channel! So now besides Hacker News, you can pretest your Product Hunt taglines and description texts before submitting your products here ;) The prediction models are now updated with new data regularly, and the Hacker News algorithm has an update that improved its accuracy (from approx. 60% -> 70%) in guessing the winning title.
Stefano Guglielmi
No longer available