Improving LLM chatbot UX by utilizing a stateful Markdown stream processor

shopify.engineering

4 pointsatesgoral2 comments

Or you can specify that LLMs respond in valid HTML and save yourself the processing step.

Ignoring the additional token count cost of HTML, how do you envision rendering HTML that is spit out by an LLM directly (without any processing) to the user?

Reply on news.ycombinator.com