29d (edited) • 📚 Resources & Tools
Save on LLM tokens! Most of you are paying for whitespace.
Seriously. Every time you paste a chunk of HTML, a Notion export, a Google Doc, or a scraped webpage into ChatGPT or Claude, you're spending tokens on:
- <div class="flex flex-col gap-4"> garbage
- Inline styles nobody asked for
- 47 nested spans wrapping the word "hello"
- Tracking pixels and footer junk
- The same nav menu repeated on every page you scraped
The model doesn't need any of it. It just needs the content.
So I built Prompt2Markdown. Paste the messy thing in, get clean markdown out, drop that into your prompt instead. I've seen inputs shrink 60 to 80% on real documents. That's the difference between hitting a context limit and not.
The hot take: prompt engineering discourse is obsessed with clever wording, and almost nobody talks about the fact that half your prompt is literally invisible formatting the LLM has to read anyway.
Clean your inputs before you tune your instructions. Try it on your next big paste. You'll be annoyed at how much you've been wasting.
Free for now:
2
0 comments
Save on LLM tokens! Most of you are paying for whitespace.
powered by
skool.com/design-and-flo-7559
Learn to build your website and how to leverage AI for branding, content creation, and website automation.
Suggested communities
Powered by