Public Market Comps Tool

Motivation:

• There's all this talk about SaaS being dead because even the least tech-savvy can build anything with AI!
• Coding-proficient people (like me) need to learn how to integrate AI effectively -- and ideally be able to build live, working products quickly
• Somehow, even with all of the AI and software in the world, I can't seem to find tools that work exactly like I want them to!

Product:

• One important step of VC is public market comps, which helps give a good estimate of what the company could be worth in X years (e.g. using a simple EV/revenue multiple)
• We have an AlphaSense license, but it (a) somehow isn't able to pull this financial data well (it struggled to get US-based public market caps) and (b) AlphaSense really, really, really wants to lock you into their product (which I hate!) by making all links point to an alphasense.com URL

Instead of trying to build this from scratch (again), I threw it to Google Antigravity. My prompt:

i'd like your help creating a new project that finds public market comps for an input company, industry, or sub-industry. i think this should include a step for market scanning/filtering, and a step for pulling metrics like rev, EV/rev, market cap, company name, country/exchange, brief description. this should be built modularly. also use OpenRouter for LLM calls.

also build a simple streamlit frontend for this

A couple notes on this process:

• Being "good at AI" is slowly meaning "being able to clearly state everything you want, exactly how you want it." My prompt above probably gets a solid B-
• Being conversant in software development, however, is a huge plus! Knowing what APIs or frameworks are good (and why) is a very human, consider-the-tradeoffs skill, at least for now. For example, I like OpenRouter because it ensures the whole app is LLM-agnostic. I also like Streamlit because it is in Python and so simple (meaning: I don't have to go learn another language and framework). 
• This is also helpful to know how this "agent" works. Right now, it's a two-step process: (1) "scan" the market with an LLM prompt, then (2) pull the market data from a Yahoo Finance API (yfinance). 
• Should this be insufficient, a few ideas that I could use: (a) upgrade the single LLM call to aggregate the results across multiple, (b) use a market scanning API in addition (or instead of) the LLM call, (c) replace yfinance with a Bloomberg API, etc. etc. etc.
• I had to upgrade to Google AI Pro ($20/month) for Antigravity to finish the app (it ran out of credits!). How many credits the Pro upgrade comes with still remains a mystery. 

I created the app in about 1.5 hours, while watching the US/Slovakia men's hockey game. (Antigravity did most of the work.) It is imperfect and not to my exacting expectations, but (a) I recognize this is my fault in not being crystal clear, (b) it works well enough, and (c) it is completed. Easier to edit than write anew. 

Anyways, here's the link: https://market-c-fsjognovanewadxgz7o72g.streamlit.app/. If you happen to use this, don't incinerate my money by using too many tokens. (Fortunately, it looks like right now, about 100 runs of this would cost less than 10 cents.)