I analysed 17 years of fast food and coffee spending using OpenAI Codex (opens in new tab)

(chrisflemming.com)

2 pointscpf_au1mo ago1 comments

1 comments

I wish there was more detail on verifying things like "code parsed PDFs to CSV". LLMs have trouble with arithmetic, so some verification should take place, but isn't mentioned.

j / k navigate · click thread line to collapse

1 comments

bediger40001mo ago

I wish there was more detail on verifying things like "code parsed PDFs to CSV". LLMs have trouble with arithmetic, so some verification should take place, but isn't mentioned.

j / k navigate · click thread line to collapse