Hacker News Reader

Comment by cpard | Hacker News Reader

There’s an emerging pattern in agentic systems and this project is a great example.

A deterministic layer like a compiler or generator of code with some kind of IR that the LLM generates and feeds it with.

I feel we will be seeing this more and more in the near future.

View on HN

When I first saw Claude generating PPT decks by writing Python code instead of making the XML directly, it was sort of an "aha moment" for me. This seems to be the path for many things. It also feels slightly limiting, and like a hack LONG term, but 100% correct approach for a while.

Have you seen marp? https://marp.app/

Another somewhat similar concept is how you can write Quarto markdown notebooks and have them output a RevealJS slideshow

I've been using slidev recently. How does this compare?

This is awesome

Do we have more details / documentation on this capability. Can we replicate the same using a generic agent skill with custom apps / harness.

I think I saw some library named pptxjs? in the thinking traces of Claude. Or was it this: https://www.npmjs.com/package/pptxgenjs

Anthropic released their PPTX skill[0] a while ago, the relies a bit on python and pptxgenjs, depending on the usecase:

> Read/analyze content....python -m markitdown presentation.pptx > Edit or create from template....Read editing.md > Create from scratch...Read pptxgenjs.md

[0] https://github.com/anthropics/skills/blob/main/skills/pptx/S...

Thanks -- will study and adopt this. Appreciate the pointer.

can you save that python and regenrate that exact powerpoint later ? maybe check it into repo isntead of powerpoint to genrate variations.

A well designed intermediary enables both validation and control over the output independent of the AI. This changes the interaction model between human and AI from delegation to collaboration.

also user interaction afterwards -- if can be frustrating if the only way the user can interact with the chart is to chat with the agent again (simple spec allows easy UI interaction!)

I'm all in on this idea. Every piece of agentic coding I have done in the last month has been via an intermediate representation. Iteration is done in the IR layer mainly. It's remarkable how close you can get to a deterministic coding output using this methodology.

Yes, this has been the pattern for agentic systems since the beginning: permissive generation, that retries over and over until it gets the right size shape through the hole, and the input validates.

Programming is more alive than ever

Reply on news.ycombinator.com