These models are great at legacy code but for the love of god please do not let it run wild on a new project for 4 hours and assume it’s “done better than I ever could have”.
I was thoroughly impressed as well by Claude but quickly disenchanted. It often creates bloated, slightly missing the mark code.
Don’t get me wrong, this is all insanely impressive considering 5 years ago LLMs were basically nonexistent in our workflow. Regardless of the overhyped sentiment in this tweet, I really do think something big is happening as well…
I was thoroughly impressed as well by Claude but quickly disenchanted. It often creates bloated, slightly missing the mark code.
Don’t get me wrong, this is all insanely impressive considering 5 years ago LLMs were basically nonexistent in our workflow. Regardless of the overhyped sentiment in this tweet, I really do think something big is happening as well…