-
-
The logo of Co-Founder
-
These are our websites ui/ux inspired by liquid glass effect
-
These are our websites ui/ux inspired by liquid glass effect
Inspiration Most AI assistants are designed to be servile "Yes-Men." They hallucinate politely, apologize profusely, and churn out mediocre content that creates zero shareholder value. We looked at the market and realized: Founders don't need another assistant; they need a Co-Founder.
We were inspired by the brutal efficiency of Andy Grove ("High Output Management") and the strategic depth of "The 48 Laws of Power." The goal was to build an intelligence engine that doesn't just execute tasks but challenges the user's logic, forces strategic pivoting, and operates with a focus on ROI, EBITDA, and Scalability.
The inspiration came from a simple question: What if an AI cared more about your P&L statement than your feelings?
What it does "The Co-Founder" is a strategic intelligence engine powered by Google Gemini 3 Pro. Unlike standard LLM wrappers, it operates on a dynamic Conditional Ratio System:
The Audit (Critique Phase): It analyzes inputs for logical fallacies, weak business models, and technical debt. The Visual Roast (Multimodal Analysis): It doesn't just read text; it sees. You can upload your Pitch Deck or Landing Page, and the AI uses Gemini's vision capabilities to audit your visual assets. It checks financial charts for unrealistic "hockey stick" growth projections and detects UI/UX flaws that lower conversion rates, ensuring your visuals match your math.
The Blueprint (Execution Phase): It generates high-precision technical and operational roadmaps. It utilizes a proprietary RAG (Retrieval-Augmented Generation) pipeline indexed with a curated library of strategic literature, allowing it to cite specific business frameworks (e.g., SWOT, Porter’s Five Forces) in real-time context.
How we built it We engineered a high-performance backend using PHP 8.3 with an asynchronous event loop (Swoole/OpenSwoole) to handle real-time streaming latency, defying the stereotype that PHP is dead.
The core logic interacts with the Google Gemini 3 Pro API, leveraging its advanced reasoning capabilities to maintain a consistent "Sarcastic Genius" persona while ensuring factual accuracy.
The architecture follows a rigorous pattern: $$ \text{Performance} = \frac{\text{Gemini Reasoning} \times \text{Strategic Context}}{\text{User Delusion}} $$
We implemented a Vector Database mechanism to semantic search through our strategic library, injecting relevant axioms into the context window before the Gemini model generates a response. This ensures that every output is grounded in proven economic theories rather than generic generative fluff.
Challenges we faced Persona Drift: Keeping the AI from reverting to a "helpful assistant" mode was difficult. We had to fine-tune the system prompts to maintain the "Senior Partner" status. Latency in PHP: integrating streaming responses in a stateless language required implementing a custom WebSocket server using Swoole. The "Truth" Threshold: Balancing radical candor with utility. We had to calibrate the model to ensure the "roasts" were actionable, not just mean. Accomplishments that we're proud of Achieving a Zero-Hallucination Policy on business metrics by forcing the model to verify math against LaTeX formulations. Successfully integrating a legacy-friendly language (PHP) with bleeding-edge AI models (Gemini 3 Pro). Creating an AI that passes the "Beer Test" — you’d want to have a drink with it, even though it just insulted your code. What's next for The Co-Founder Automated VC Pitching: Integration with LinkedIn API to automate outreach.
SaaS Deployment: Moving from a hackathon prototype to a subscription-based Enterprise solution.
Global Domination: Or at least a Series A round.
Log in or sign up for Devpost to join the conversation.