Advisory AI: Get Brutally Honest Feedback on Your Ideas

The Moment You Realize You Need Real Feedback

Picture this: You've been working on your startup idea for months. You've talked to friends, bounced it off your team, maybe even pitched it to a few potential customers. Everyone says it sounds "interesting" or "has potential." But deep down, you know something's missing.

What you really need is someone to tell you the brutal truth. Someone like Steve Jobs who would cut through the fluff and ask, "What would you remove if you only had one screen?" Or Charlie Munger who would invert your entire approach and show you exactly how it fails. Or Naval Ravikant who would demand to know where your leverage is.

The problem? Hiring real advisors costs $50K+ per year. Your friends are too polite. Your team is too invested. And most feedback you get is either too nice, too vague, or too shallow to actually help.

What If You Could Assemble Your Dream Advisory Team?

Imagine having access to advisors that include the sharpest minds in business and philosophy. Not just their quotes or books, but their actual thinking patterns applied to your specific situation.

What if Steve Jobs could tell you exactly what to cut from your product? What if Naval could map out your leverage opportunities? What if Charlie Munger could show you the mental models you're missing and the ways your idea will definitely fail?

That's exactly what Advisory AI delivers - five AI agents, each embodying the thinking patterns of legendary advisors, providing structured critique that's actionable, honest, and transformative.

Meet Your AI Advisory Team

Each agent brings a unique lens to your ideas, asking the questions that real advisors would ask and providing the kind of feedback that turns good ideas into great ones.

Jobs Lens: Ruthless Simplicity
Cuts through feature bloat to find the essential user need. Asks what you'd remove if you only had one screen. Scores you on simplicity, user focus, and execution clarity. The brutal truth about whether you're building a product or a feature factory.

Naval Lens: Leverage & Wealth Creation
Maps your leverage opportunities across code, media, capital, and people. Identifies network effects and scalability potential. Asks the hard questions about whether you're building a time-for-money business or something that compounds.

Munger Lens: Mental Models & Inversion
Applies relevant mental models to your situation and inverts your approach to find failure modes. Identifies cognitive biases in your planning. Shows you the second and third-order consequences you haven't considered.

Indian Philosophy Lens: Dharma & Long-term Thinking
Evaluates the dharmic purpose of your endeavor and its impact on all stakeholders. Asks about consequences seven generations from now. Challenges you to align with higher purpose beyond profit.

Ruthless Capitalist Lens: Moats & Pricing Power
Analyzes competitive dynamics and defensibility. Asks what prevents competitors from copying you in six months. Evaluates your pricing power and path to market dominance. No mercy for weak business models.

Synthesis Agent: Unified Action Plan
Reconciles disagreements between advisors and creates one actionable plan. Identifies consensus recommendations and productive tensions. Gives you a clear path forward that addresses multiple concerns.

What You Actually Get

This isn't just another AI tool that gives you generic advice. Each agent provides structured, actionable feedback that you can immediately use to improve your idea.

Every critique includes a brutal one-liner that cuts to the core issue, key questions that need answers, assumptions analysis labeled as fact/assumption/guess, risk assessment with specific failure modes, one bold counterintuitive move to consider, numerical scorecards on relevant dimensions, and a concrete 30-day experiment to validate or invalidate your approach.

The output is beautifully formatted in Markdown by default, with optional JSON for automation. You can run individual lenses or get the full advisory treatment with synthesis. Interactive mode lets agents ask clarifying questions before providing feedback.

# Get brutally honest feedback from world-class advisors
# Zero-config setup with local AI models

# Install and setup
git clone https://github.com/anupsahoo/advisory-ai
cd advisory-ai
make setup

# Verify everything works
advisory doctor

# Get full advisory critique
advisory critique --input my-idea.md --output critique.md

# Single lens analysis
advisory critique --lens jobs --input my-idea.md
advisory critique --lens naval --input my-idea.md
advisory critique --lens munger --input my-idea.md

# JSON output for automation
advisory critique --input my-idea.md --json --output results.json

# Interactive mode with clarifying questions
advisory critique --input my-idea.md --interactive

Built for Independence and Privacy

The system runs completely offline with Ollama by default. No API keys, no external services, no data leaving your machine. Your ideas stay private while you get world-class feedback.

Optional OpenAI support is available via environment variable if you prefer cloud models. But the default experience requires zero configuration - just install Ollama, clone the repo, and start getting feedback.

The architecture follows the same independence philosophy as our other systems. Each agent is self-contained, the CLI is built with Typer for reliability, and everything is designed to work without external dependencies.

Real Example: AI Fitness App Critique

Let me show you what this looks like in practice. Here's how the advisory team would critique an AI-powered fitness app idea:

Jobs Lens: "You're building a feature factory, not solving a real problem. Cut the social features, gamification, and multiple workout types. Focus on one thing that users will do every single day. What would you remove if you only had one screen?"

Naval Lens: "Where's your leverage? This looks like a time-for-money business disguised as software. How does this scale without your direct involvement? What network effects make this more valuable over time?"

Munger Lens: "You're assuming users will behave rationally and stick to workout plans. Apply the mental model of incentive alignment - what really motivates behavior change? Invert this: how does a fitness app definitely fail?"

Indian Philosophy Lens: "What's the dharmic purpose here? Are you serving genuine health transformation or feeding into vanity and comparison culture? How does this impact all stakeholders, not just paying customers?"

Ruthless Capitalist Lens: "You have no moat. Competitors will copy your features and undercut on price. Where's your pricing power? What prevents Nike or Apple from crushing you with better distribution?"

Synthesis: "The advisory team agrees: focus on one transformative behavior, find the leverage angle that creates network effects, address the motivation problem with better incentive design, ensure you're serving genuine health outcomes, and build defensible advantages before competitors notice."

Technical Architecture That Just Works

The system is built with production-grade Python using Pydantic for data validation, Typer for the CLI, and a clean agent architecture that's easy to extend.

Each agent follows a consistent interface but implements unique prompting strategies. The synthesis agent reconciles disagreements and creates unified action plans. Blog reference matching connects critiques to relevant articles from a curated knowledge base.

Error handling is robust, the CLI provides helpful feedback, and the `advisory doctor` command verifies that everything is working correctly. The codebase includes comprehensive tests and follows best practices for maintainability.

Getting Started in Minutes

The quickstart experience is designed to be friction-free. Install Ollama, clone the repository, run `make setup`, and you're ready to get world-class feedback on your ideas.

The system includes example idea files so you can see how it works immediately. The CLI is intuitive with helpful error messages and clear documentation. Everything is designed for developers who want powerful tools without complexity.

You can start with single-lens analysis to understand each perspective, then graduate to full advisory critiques with synthesis. The JSON output option makes it easy to integrate with other tools or build automation around the feedback process.

Why This Approach Works

Traditional feedback is either too nice (friends), too invested (team), or too expensive (real advisors). AI-generated feedback is often too generic or lacks the specific mental models that make great advisors valuable.

This system bridges that gap by encoding the specific thinking patterns of legendary advisors into AI agents. You get the brutal honesty of Jobs, the leverage thinking of Naval, the mental models of Munger, the long-term perspective of Indian philosophy, and the competitive analysis of ruthless capitalism.

The structured output format ensures feedback is actionable rather than just interesting. The synthesis agent prevents analysis paralysis by creating unified recommendations. The local-first approach ensures your ideas stay private during the vulnerable early stages.

Beyond Basic Feedback

Once you have the system running, you can customize it for your specific needs. Add new agent lenses by implementing the agent interface. Modify prompts to adjust agent behavior. Create custom scoring rubrics for your industry.

The blog reference system can be populated with your own articles and knowledge base. The JSON output enables integration with project management tools, automated reporting, or custom dashboards.

For teams, you can create standardized critique processes where every new idea gets advisory feedback before moving to development. The deterministic output makes it easy to track how ideas evolve based on advisory recommendations.

The Independence Advantage

Building your own AI advisory system means you control the entire experience. No monthly fees, no usage limits, no terms of service changes. Your feedback system evolves with your needs rather than being constrained by external platforms.

The agents can be fine-tuned for your specific industry or business model. The prompts can be adjusted based on what you learn about effective feedback. The scoring rubrics can be customized for your decision-making process.

This independence compounds over time. As you use the system more, you develop better prompts, more relevant examples, and deeper insights into what makes feedback actionable. You're building a business asset, not renting access to someone else's platform.

Real Impact on Decision Making

The goal isn't just to get feedback - it's to make better decisions faster. By systematically applying different thinking frameworks to your ideas, you catch blind spots before they become expensive mistakes.

The structured format makes it easy to compare different approaches or track how ideas evolve. The synthesis agent helps you prioritize which feedback to act on first. The 30-day experiment format turns insights into concrete next steps.

Teams report that having consistent, high-quality feedback early in the ideation process leads to better outcomes and fewer pivots later. The brutal honesty helps kill bad ideas quickly, freeing up resources for better opportunities.

Getting Your Copy

Advisory AI is open source and available on GitHub. The setup process is designed to be friction-free - you can be getting world-class feedback on your ideas within minutes of cloning the repository.

The system includes comprehensive documentation, example files, and test suites. The codebase follows best practices and is designed for easy extension and customization.

Whether you're validating a startup idea, evaluating a product feature, or making a strategic business decision, having access to multiple expert perspectives can transform your decision-making process.

Stop settling for polite feedback or expensive consultants. Get the brutal honesty and strategic insight you need to turn good ideas into great ones.

Next Steps

• Download Advisory AI
• Explore 35+ More Thinking Agents
• Build Your Own Independent Systems