Skip to main content

AI Assisted Development

The code samples and applications we provide are specifically designed with AI assistance in mind:

  • We have incorporated targeted training data to enhance API comprehension
  • We made sure that the primary AI tools are fully compatible with our codebase
  • Type hints are used throughout application code to minimize ambiguity and boost AI interpretation
  • Additional context, including data types and examples, is provided to the models wherever necessary
  • Notebooks include results to further enrich AI understanding of API responses
  • The programming languages we use - Python and SQL - are well-supported by AI tools

Below you can find a brief overview of each tool and its capabilities.

Recommended Use:

These AI tools significantly reduce the barrier to entry for working with our products, allowing you to focus more on strategy development and less on implementation details.

FundPro AI Assistants​

With the release of GPT-5.2, ChatGPT became one of the most powerful general-purpose models available. We use it to power two purpose-built assistants:

Both AI Assistants should be used with "Extended thinking" enabled.

We recommend using these assistants for strategy design and product/API questions, then switching to a coding agent (Codex or Claude Code) for implementation work.

OpenAI Codex​

OpenAI Codex (often shown as ChatGPT Codex) is a dedicated coding agent that can navigate a repository, apply multi-file changes, and iterate with tests and linters. It is on par with (and in many cases better than) Anthropic's Opus, while typically providing more generous daily and weekly usage limits.

We recommend Codex for agentic development work such as:

  • Implementing new features across multiple files
  • Refactors, migrations, and dependency updates
  • Running tests and iterating until the change is correct

While Merlin has been developed prior to OpenAI's Codex release, it is fully compatible with Codex and can be used for agentic development tasks.

New projects should leverage either ChatGPT Codex or Claude Code, building on top of Merlin or the MesoLive SDK:

Claude Code​

Claude Code offers powerful capabilities for understanding and extending codebases using a command line interface.

Merlin has been developed with first-class Claude Code support, allowing it to:

  • Analyze and understand the Merlin codebase structure
  • Make changes to the code base and validate the change using tests and static analyzers
  • Generate new apps using the provided APIs

As Claude Code has both the context and the ability to execute code, it is the most powerful tool for code generation and modification tasks.

Just clone the Merlin repository and run claude to start vibe coding.

Gemini​

Gemini (built into Google Colab) provides a convenient way to access MesoSim, Q-API, and Merlin features, particularly for smaller code segments that fit within notebooks.

With zero setup required, you can start using Gemini immediately through Google Colab to:

  • Generate code for data analysis and visualization
  • Get real-time assistance while working with the APIs
  • Modify existing notebooks to meet your requirements
  • Learn through interactive examples and explanations

The provided notebooks can be used with Gemini to quickly explore capabilities and build custom solutions.