I did not realize when I started Cakelisp how freeing it felt. All of the sudden, I got to decide what made sense to me, not what made sense to previous language designers.
The evaluation uses a pairwise comparison methodology with Gemini 3 as the judge model. The judge evaluates responses across four dimensions: fluency, language/script correctness, usefulness, and verbosity. The evaluation dataset and corresponding prompts are available here.
。新收录的资料是该领域的重要参考
I was inspired by Naughty Dog's use of Game Oriented Assembly Lisp, GOOL, and Racket/Scheme (on their modern titles). I've also taken several ideas from Jonathan Blow's talks on Jai.
The only part that was difficult to figure out initially was press emails. I get these from random companies all the time, and not all of them are in my contacts list. For this problem, the solution I came up with was similar to the one I used for junk emails above. I had the AI create a filter to sort any email that includes common press email words and phrases like “announce,” “launch,” and “embargo.” This filter required manual tuning over the course of a couple of weeks as more phrases and words were identified, but it has mostly worked well.