Hacker Newsnew | past | comments | ask | show | jobs | submit | aditgupta's commentslogin

Jim nailed the core problem. I've been building exactly this "missing layer" for past few months. The challenge isn't just connecting product decisions to code. It's that product context lives in a format that's optimized for human communication, not machine consumption. When engineers feed this to LLMs, they spend massive effort "re-contextualizing" what stakeholders already decided. I built TypMo (https://typmo.com) around two structured formats that serve as this context layer: PTL (Product Thinking Language)- Structures product decisions (personas, objectives, constraints, requirements) in a format both humans can read/edit and LLMs can parse precisely. Think YAML for product thinking. and Interface Structure Language (ISL) - Defines wireframes and component hierarchies in structured syntax that compiles into visual mockups and production-ready prompts. LLMs don't need more context, they need structured context. The workflow Jim describes (stakeholder meeting → manager aggregates → engineer re-contextualizes for LLM) becomes: stakeholder meeting → PTL compilation → IA generation → production prompts.

LEt's see where it goes!


Nice and I'm thinking along similar lines but not DSLs.

My intuition off reading what you wrote is... Nobody is gonna want to write PTLs and ISLs.


Exactly right, and that's the core point. Users don't write PTL or ISLs. Let's say you have customer interactions (fetched from Zoom) or product/research notes. The AI structures that into PTL automatically. You see clean, editable notes and visual wireframes + high-fidelity prototypes and prompts. The structured formats exist in the background for token efficiency and interoperability.


Ah I see. When I say write.. I also mean review and revise.


... and understand. Whenever I vibe code something and there is an issue with it, I either have to step in myself and have to learn what was generated or hand it back to the ai to fix it.

The generation of these meta documents will also be fixed by ai, not humans.


I wanted a minimal tool to easily track, organize, and reflect on my reading—so I built one: https://bookstates.app. I'm familiar with StoryGraph, but aimed for something even simpler. (and with a sprinkle of AI)


Governance is always a bigger challenge than creation.


It's the difference between raising your hands in the air, vs keeping them in the air indefinitely.


I made a simple iOS app for designers called as UX Assist. It has been around for 6 years now. I’m just happy that it’s out there and helping some folks now and then. [UX Assist](https://apps.apple.com/in/app/ux-assist/id1491649877)


First, not everybody will have interest in Math. And that's fine. Secondly, you need to be creative for teaching math, especially at a junior level. I think Lockhart gives a very good insight on this matter - http://www.maa.org/devlin/lockhartslament.pdf


This was more for basic understanding. There can be a follow-up article with a more rigorous approach :) The same was done here - http://functionspace.org/articles/17/Solving----sum----1----...


Guys, there's lot more here - http://functionspace.org/discussion/new :)


I will certainly not call it "cool"


It'd be far cooler than the conspiracy theories and assumption of malice that mediate poor and rich relations right now.


This looks cool! We made something similar - http://blog.functionspace.org/news/2013/5/8/a-beautiful-read...



Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: