I include the prompts for each iteration but hide the parts where I painstaikingly re-prompted the LLMs until they got it right. WebVG can do some basic things right and I am now adding Slidus functionalities to it, it’ll be so good once it’s done.
Sometimes it’s crazy obvious stuff like “see how Elements do it”, which is no wonder why people have AI fatigue. Do we even consider that cheating? I mean it already got it right so just reuse stuff that works.
Slidus was a big endeavor and I really hope that WebVG can learn and replicate it.