Show HN: I challenged 10 AI giants using one open-source PDF (with full results)
3 WFGY 2 6/23/2025, 2:02:00 PM zenodo.org ↗
Hey HN,
This started as a personal experiment: one person, one framework, ten AI models.
I built a semantic reasoning engine (WFGY: All Principles Return to One) and tested how well each model could handle abstract logic, conceptual shifts, and consistent inference—all using the same PDF.
The results are posted above. No fancy wrappers, no login walls—just raw data, an illustrated battle poster, and the full experiment.
Yes, it's a bit weird. But it's real. And honestly? I just hope someone out there sees the effort and the courage it took to do this solo.
Happy to answer questions. Would love your feedback, criticism, or even memes. Thanks for taking a look
Like—going up against 10 big models at once, making it look like some kung fu battle, and then just dropping all the data out in the open? That’s kinda nuts (in a good way).
So, which model surprised you the most? Did any of them totally flip your prompt in a way you didn’t see coming?
But Gemini pulled something even crazier — it rewrote my prompt into a corporate mission statement I didn’t know whether to laugh or cry.
Each of them has their own “personality,” which is what made this challenge so wild. And yeah, dropping the data open-source was part courage, part madness, part… strategy
Still curious which one you think held up the best?