Inspiration
Trading tools often tell people what to do: buy, sell, long, short.
But many bad decisions happen because traders confuse a reaction with confirmation, enter after a move has already happened, or act under FOMO and pressure.
AIX PROOF FIRST was created around a different question:
What is actually visible and proven in the chart right now?
The goal is not to predict the market or replace the user’s judgment. The goal is to slow the decision down, separate evidence from assumptions, and help the user understand what would confirm or invalidate an idea.
What it does
AIX PROOF FIRST is a Human-First market decision coach.
The user uploads a chart screenshot and asks a simple question such as:
- What do we actually see?
- Is this only a reaction or already proof?
- What is still missing?
- What would confirm or invalidate the idea?
- Is a sweep or open stop pool visible?
- How high is the chase risk?
The application analyzes the screenshot and returns a structured explanation containing:
- Visible chart evidence
- Missing proof
- Confirmation conditions
- Invalidation conditions
- Momentum assessment
- Chase-risk assessment
- Proof Meter
- FOMO Lock
Instead of producing a blind buy or sell signal, AIX PROOF FIRST may deliberately answer:
Wait. The reaction is visible, but confirmation is still missing.
That is the core of the product: proof before opinion.
How we built it
The application was built with Python, Flask, HTML, CSS and JavaScript.
It uses the OpenAI Responses API to understand uploaded chart screenshots and transform visual market information into a structured, readable explanation.
The frontend includes:
- German and English language modes
- Simple and expert modes
- Chart screenshot upload
- Predefined natural-language questions
- Structured analysis cards
- Proof Meter and FOMO Lock
- A responsive black-and-gold AIX interface
The application is deployed on Railway and the source code is maintained in a private GitHub repository.
Codex was used as a development and security partner. It reviewed the application, hardened the backend, created automated tests and documented remaining production risks.
Security improvements include:
- Strict JSON and request-size limits
- Base64 and image-signature validation
- MIME-type validation
- Decoded image-size limits
- Secure response headers
- Same-origin browser protection
- Per-IP and global rate limiting
- Prompt-injection boundaries
- Generic client-facing errors
- No permanent screenshot storage
The automated test suite currently passes all 11 tests.
Challenges we faced
One major challenge was preventing the AI from sounding more certain than the chart evidence allows.
Charts contain many indicators, labels, zones and user annotations. The system therefore has to distinguish between:
- What is directly visible
- What is only an interpretation
- What evidence is still missing
- What would invalidate the interpretation
Another challenge was security. Uploaded screenshots and optional user notes are untrusted input. We had to validate the files carefully, limit requests, reduce prompt-injection risk and avoid exposing internal errors or secrets.
We also wanted the interface to remain understandable. The result had to be useful for experienced traders without overwhelming beginners.
Accomplishments that we are proud of
We built and deployed a working end-to-end product:
Upload chart → ask a question → analyze visual evidence → receive structured confirmation and risk guidance.
We are especially proud that the product does not reward impulsive behavior.
When the evidence is incomplete, it says so clearly.
AIX PROOF FIRST is designed to protect the user from chasing a move, treating every wick as proof, or following an AI-generated signal without understanding the risk.
What we learned
We learned that an AI trading assistant does not need to predict prices to provide value.
A powerful assistant can instead improve the quality of the user’s decision process by identifying uncertainty, requesting confirmation and making invalidation visible.
We also learned how valuable Codex can be beyond writing features. It helped turn a functioning prototype into a more secure and testable controlled-beta application.
What's next
The next development stage includes:
- User accounts and authentication
- Usage quotas and transparent pricing
- Persistent rate limiting through Redis or an API gateway
- Stronger Content Security Policy
- More structured chart-pattern education
- Expanded controlled beta testing
- Additional Human-First decision tools
The long-term vision is to build an assistant that does not create pressure or sell certainty.
No blind signal. No chase. Proof first. Human First.
Built With
- api
- codex
- css3
- flask
- gunicorn
- html5
- javascript
- openai
- python
- railway
- responses
Log in or sign up for Devpost to join the conversation.