PSX War Room

What the company appears to be: PSX War Room is a self-reported hackathon project that proposes an AI-powered system for analyzing Pakistan Stock Exchange (PSX) stocks using four specialized AI agents — Quantitative, Macro, Geo-Sentiment, and Fundamentals — each offering a directional signal with confidence and basis. A fifth agent, the Moderator, synthesizes these views into a single verdict without making price predictions or buy/sell calls.

What changed: The project is described as originating from a hackathon submission, with no evidence of prior development, funding, or commercial traction beyond its initial prototype.

Single most important open question: Is there any evidence that the system has moved beyond a proof-of-concept prototype and into actual use by investors or students?

The description states that PSX War Room is a system that runs four AI agents in parallel on a single PSX ticker, each analyzing it from a different angle:

• Quantitative (technical analysis) — price and volume
• Macro (economic policy, inflation, currency, IMF program)
• Geo-Sentiment (political and regional risk)
• **Fundamentals (valuation ratios, debt, earnings trends)

A fifth agent, the Moderator, synthesizes their outputs into one verdict without asserting a price target or making buy/sell recommendations.

The system enforces constraints on output format: each agent must return only:

• SIGNAL: BULLISH | CONFIDENCE: 0.4 | BASIS: ...

These outputs are parsed and validated to prevent LLMs from producing uncontrolled responses such as price targets.

Not evidenced: No evidence of actual data feeds, live integration with PSX, or real-world deployment beyond the prototype.

The authors state that PSX War Room was built to address a gap in beginner-friendly tools for investors in Pakistan. It is positioned not as an automated investment advisor but as a tool to help users reason about the market instead of being handed answers by algorithms.

They claim it avoids common pitfalls like overconfidence or misleading predictions, and emphasizes transparency in uncertainty.

Inference: The positioning implies a focus on education and decision support rather than direct financial advice or trading automation.

Not evidenced: No evidence that this is a product in use, nor any indication of how it would be monetized or scaled beyond the prototype.

The authors describe their target as first-time investors and students, particularly those from STEM backgrounds who are new to investing in the Pakistan Stock Exchange (PSX).

They state that existing tools either assume prior knowledge or give overly confident advice — which they aim to avoid.

Not evidenced: No evidence of customer acquisition, user feedback, or actual adoption by target users.

The description does not contain any information about pricing, monetization, or business model. It is presented as a hackathon project with no indication of commercial intent or revenue streams.

Not evidenced: No evidence of a business model, pricing structure, or customer contracts.

• Built using Python and OpenAI Agents SDK.
• Uses an agents-as-tools architecture.
• Demo runs on Streamlit.
• The system enforces strict output formatting via prompt engineering, banned-language validators, and format parsers.
• Designed to handle uncertainty gracefully — e.g., confidence caps based on data quality, exclusion of failed agents with reasons shown in UI.
• Refactored for model agnosticism (e.g., supports switching between OpenAI and Gemini).
• Currently runs on mock fixtures; no live data integration is mentioned.

Inference: The architecture shows some sophistication in managing LLM outputs and system robustness, but it remains a prototype.

Not evidenced: No evidence of production-grade infrastructure, scalability, or deployment beyond the demo.

The project is described as a hackathon submission, built by two first-year students. It currently covers only three tickers as a prototype and has no live data feed.

There is no evidence of:

• Revenue
• Customers
• User engagement
• Product adoption
• Iteration beyond the initial prototype
• Any form of commercial traction

Not evidenced: No evidence of any traction or maturity beyond the prototype stage.

The description does not mention competitors or market context. It is unclear whether similar tools exist in the PSX or broader financial AI space, nor how PSX War Room would differentiate itself if it were to scale.

Not evidenced: No competitive analysis or awareness of existing solutions.

• Prototype-only: The system is described as a hackathon prototype with no live data integration.
• No commercial traction: No evidence of users, customers, or revenue.
• Unverified claims: The product’s ability to deliver accurate or useful signals is unproven without real-world testing.
• Limited scope: Currently covers only three tickers and lacks full KSE-100 coverage.
• Dependency on external APIs: The system currently runs on mock data, and live integration is described as a future step.

Inference: The project may be more of an educational or exploratory effort than a commercial product.

1. What is the current status of live data integration with PSX?
2. How does the system handle real-world data quality issues, and what happens when data sources are unavailable?
3. Has there been any user testing or feedback from students or first-time investors?
4. Are there plans to monetize this product, and if so, how?
5. What is the timeline for expanding coverage beyond the current three tickers?
6. How does the system validate that its agents are actually providing useful insights rather than just generating plausible-sounding but inaccurate outputs?