This episode reveals how to build a powerful, no-code AI research agent for crypto for free by strategically combining the unique strengths of specialized large language models.
Introduction to the AI Stack for Crypto Research
The speaker demystifies what many perceive as a complex AI stack, framing it as a simple "horses for different courses" approach. He explains that the key is understanding which AI model—such as Grok, GPT-4, Claude, or Gemini—is best suited for a specific crypto research task. This episode provides a practical guide to these workflows and concludes with a step-by-step tutorial on creating a free, no-code crypto research assistant.
The "Cheat Sheet": Which AI Model to Use and When
The core confusion for many users is knowing when to use a specific AI model. The speaker presents a clear breakdown of each model's primary strengths, offering a "cheat sheet" for crypto researchers to optimize their workflows. This section sets the stage for a deeper dive into each tool's specific applications, from data handling and coding to real-time market analysis.
Claude: The Specialist for Data Handling and Coding
- Claude, developed by Anthropic, excels at data handling, coding, and seamless integrations via its Model Context Protocol (MCP). The speaker explains that MCP is a protocol developed by Anthropic that allows large language models to have two-way communication with external databases and applications, like a personal finance spreadsheet.
- Key Strength: Claude's native support for MCP makes it ideal for connecting to data sources like CoinGecko for real-time market reports.
- Primary Use Cases: Coding, complex data analysis, and creating automated reports by linking directly to crypto data APIs.
Grok: The Real-Time Crypto Twitter Analyst
- Grok holds a significant advantage due to its proprietary, real-time access to Twitter's (X) data stream. While other models are trained on historical internet data, Grok can tap into the live, user-generated content where the crypto conversation happens.
- Strategic Implication: For any research involving new protocols, airdrop speculation, or real-time market sentiment, Grok is the superior tool. It provides insights from Crypto Twitter (CT) that other models cannot access.
- The speaker notes, "Everyone knows crypto marketing and basically the conversation lives on Twitter. So if you can get access to web search... and then also live information from the six million tweets... that's where Grok comes in."
ChatGPT: The Versatile Everyday Assistant
- ChatGPT is positioned as the perfect "everyday model" due to its speed, responsiveness, and strong memory function. Its ability to recall previous context makes conversations more personalized and efficient.
- Key Features: The speaker highlights its strengths in image generation (using the DALL-E 3 model integrated within it), content formatting, and general day-to-day tasks.
- Practical Use: Its user-friendly interface makes it excellent for reformatting messy outputs from other models into clean, digestible text.
Gemini: The Unsung Hero for Deep Research and Integration
- The speaker calls Gemini the "unsung hero," arguing that Google creates the best models. Gemini's power lies in its deep integration with the Google Suite (Docs, Sheets, Drive) and its massive context window, which can process up to one million tokens of input.
- Actionable Insight: For researchers working with large documents, spreadsheets, or whitepapers, Gemini is the best choice. It can analyze vast amounts of data and seamlessly export formatted reports directly into Google Docs or Sheets.
- Caution: The speaker warns that the Gemini models currently integrated into Google products are often not the most up-to-date versions. For maximum performance, users should go directly to the Gemini web interface.
Practical Workflow: Building an Overnight Crypto Report with Claude and CoinGecko
- This section provides a step-by-step guide on creating a daily, automated crypto market report. The speaker demonstrates how to connect Claude to the CoinGecko MCP—a rich crypto data aggregator—using a simple API URL.
- Process:
- Obtain the keyless access URL from CoinGecko's MCP page.
- Add it as a "custom connector" in Claude.
- Prompt Claude to generate an overnight report on specific tokens (e.g., Bitcoin, Ethereum, Solana), including top gainers, losers, and trending coins.
- The speaker uses Whisper Flow, a dictation-to-text Chrome extension, to speed up the prompting process, emphasizing his strategy of "how to get good at being lazy."
- The resulting report provides a high-level market snapshot in minutes, identifying trending tokens like Tree and Zora for further investigation.
Deep Dive Analysis: Using Grok for Real-Time Protocol Research
- After identifying a new, trending token (e.g., "Tree") from the Claude report, the workflow moves to Grok for deeper, real-time analysis. Since the token is new, Grok's access to live Twitter data is critical for gathering sentiment and breaking news.
- Prompt Strategy: The speaker asks Grok for a TL;DR on the project, its backers, who is talking about it, and the general sentiment.
- Outcome: Grok quickly identifies that "Tree" is a new DeFi protocol whose airdrop just went live, pulling in commentary from respected figures on Twitter. This provides an immediate, context-rich starting point for due diligence that other models would miss.
Refining and Verifying AI Outputs: A Multi-Model Approach
- The speaker demonstrates a crucial step: verifying AI-generated information. After using Grok to analyze a complex deal involving Ethena, he moves the output to ChatGPT for better formatting. However, when he asks ChatGPT to analyze an image of Ethena's token unlock schedule, it hallucinates, incorrectly stating that 85% of the supply is already unlocked.
- Critical Insight: "You can't just take one model's word for it. There is like extremely high hallucination rate and you have to be able to verify it across models."
- Verification Tools:
- Manual: Take the incorrect information back to Grok and ask it to verify the claim and "show its workings."
- Automated: Use a tool like Clock (from the Mirror team), which uses a consensus mechanism across different models to validate outputs and reduce hallucinations.
Advanced Automation and Final Thoughts
- For more technical users, the speaker recommends N8n, a platform for building automated workflows connecting various AI stacks and APIs. Users can find pre-built crypto and trading workflows or create their own.
- Example Workflows: Automating crypto funding fee tracking via the Binance API or creating AI agent-driven interviews.
- The speaker mentions his team has effectively gained "five or six employees" by creating automated back-end workflows, highlighting the massive efficiency gains possible.
Conclusion
This episode demonstrates that a multi-model AI workflow is essential for comprehensive crypto research. By leveraging each model's unique strengths—Claude for data, Grok for real-time sentiment, and Gemini for deep analysis—investors can build a powerful, free research agent to gain a significant analytical edge.
