Inspiration
We both love playing Go (it is a board game like chess) as noticed that you can thinking on your next move
What it does
It takes a reasoning problem which ChatGPT can't solve quickly but a model like Gemini Thinking or R1 can solve and solves it quickly! It splits the problem into steps: step1->step2->...->stepN. Then for step 1 in parallel:
- run (slow) thinking for step 1
- make a model guess the result of step 1 (many guesses! and fast! no thinking!)
Not waiting for step 1 to complete, take guessed result of step 1 and launch step 2:
- slow thinking
- guesses
Do this for all the steps!
Once step 1 is finished the step 2 and all the rest will also be finished! Now we chaise the correct guesses and arrive to the answer!
How we built it
Lots of prompting in Google AI Studio and Together chat UI (for R1) Python and openAI API.
Challenges we ran into
- No lunch! W&B rushed the clocks by 1h today, did they use thoughts rusher???? hhh
- Hard to make AI work for reliably for all problems, had to debug a bunch
Accomplishments that we're proud of
That it works!
What we learned
You can use unrelated real life inspiration for creating an app!
What's next for Thought Rusher
Hope AI will get a million times cheaper! Then the Thoughts Rusher will be 1 million times faster.
Log in or sign up for Devpost to join the conversation.