suveen_ellawela

6 ms·

1. 1.▲
   I built a better UI to explore YCombinator X25 startups(data.suveenellawela.com)

   1 points by suveen_ellawela 1y ago|0 comments
2. 2.▲
   by suveen_ellawela 2y ago

   Two companies just setup the self-hosted version too. I do not have any case studies unfortunately. However, the README file guides you through the self-hosting process.
3. 3.▲
   by suveen_ellawela 2y ago

   Thanks!!
4. 4.▲
   Show HN: Open-source self-hosted AI voice interviewer platform for Hiring(github.com)

   10 points by suveen_ellawela 2y ago|5 comments
5. 5.▲
   by suveen_ellawela 2y ago

   haha, i've been in similar situations, but this one's something else.
6. 6.▲
   by suveen_ellawela 2y ago

   agree. i think getting to have a look at o1's reasoning was pretty fun.
7. 7.▲
   by suveen_ellawela 2y ago

   definietly an interesting approach. I started writing down gameplays when i play with friends. then eventually stopped to enjoy the moment.
8. 8.▲
   by suveen_ellawela 2y ago

   yep, agree. One big part of the experiment was to see how well it does the reasoning by asking it to output the reasoning.
9. 9.▲
   by suveen_ellawela 2y ago

   goated.
10. 10.▲
    by suveen_ellawela 2y ago

    yea, didn't mean to take the fun out of the original game. the idea for this came when we asked chatgpt how to connect the words 'carrot' and 'ray'. maybe you can give a try too!
11. 11.▲
    by suveen_ellawela 2y ago

    my friends were bad clue givers. i just had to switch to ai.
12. 12.▲
    by suveen_ellawela 2y ago

    yep, agree. One big part of the experiment was to see how well it does the reasoning by asking it to output the reasoning.
13. 13.▲
    by suveen_ellawela 2y ago

    haha!
14. 14.▲
    by suveen_ellawela 2y ago

    cool stuff!
15. 15.▲
    by suveen_ellawela 2y ago

    i thought of adding a feature where you can get your own spy master. you can give it all your personal info and the clues would be customized. the botteleneck is the other human spymaster has to help with updating the game state cus I(guess
16. 16.▲
    by suveen_ellawela 2y ago

    totally agree!
17. 17.▲
    by suveen_ellawela 2y ago

    2 weeks!
18. 18.▲
    by suveen_ellawela 2y ago

    that is a really interesting point. if it is true, this shows direct usage of a single training data point ( cus there are no other resources talking about this fact)
19. 19.▲
    by suveen_ellawela 2y ago

    I wanted to play around with the temperature, but unfortunately o1 only supports '1' as the value.
20. 20.▲
    by suveen_ellawela 2y ago

    thanks for the comment. I actually tried explicitly mentioning in the prompt that 'Your guesser follows the same reasoning process'. But this did not make any clear improvements. Maybe I should've done more prompt engineering
21. 21.▲
    by suveen_ellawela 2y ago

    That's a nice experiment! I think codenames could definietly be an evaluation method for LLMs.
22. 22.▲
    OpenAI's o1 Playing Codenames(suveenellawela.com)

    234 points by suveen_ellawela 2y ago|129 comments
23. 23.▲
    by suveen_ellawela 2y ago

    thanks for the suggestion! will do that!
24. 24.▲
    I got OpenAI o1 to play the boardgame Codenames and it's super good(github.com)

    3 points by suveen_ellawela 2y ago|2 comments