A benchmark for AGI

I have crazy ideas.

I kept trying...

• make good models
• use models in a good way
• try prompt engineering
• everything

But none of it made me feel like it was AGI.

A lot of work was poured into this...

It's too expensive.

Research.
Coding.
Training.
Fine-tuning.
Prompting.
All of it.

Not in mixed ways.
Not on free trials.

It's never going to work.

The reason for this dataset is one:

Let others make AGI.

It's still not perfect.

I don't say it's accurate.

But it is the farthest I could go...

see you soon.