The Mission: Create The Best Code Mentor

The Vision: Everyone Can Become A Great Developer!

My name is Franck, and this is the story of why and how CodeCleaner AI came to be.

How It All Started:

I have a degree in business, not computer science.

Despite that I’ve been working as a CTO for different companies for the last 20 years. I had to learn how to code the hard way, combing websites, documentation, and Stack Overflow to find answers to my questions.

Almost every day, I still have to learn about new languages, new frameworks, new technologies that we need to evaluate, assess or use to build the products and solutions we use internally and for our customers.

Working in a regulated industry I know how critical it is to make sure that the code and application that we create are robust, secure, and production ready.

How We Build Software:

Building software is an iterative and collaborative process. It involves different people with different skills and expertise working together to create a product.

• The Product Manager who holds the vision.
• The Architect who decides on the technology stack and designs the system to make sure that it’ll be scalable and maintainable over time.
• The Front-end and Back-end Developers who build the user experience and server-side logic.
• The DevOps Engineer who streamlines and automates deployment and operations.
• The Quality Assurance Engineer who test the application.
• Code reviewers who make sure that the code is robust and production ready.

Generative AI can help all these people, but it is not a replacement for them all, especially if you already have an existing codebase that you need to maintain, refactor, or improve.

The Rise Of Generative AI:

• In November 2022, OpenAI released ChatGPT. It was a game changer: questions that used to take hours of painful searches, reading documentation, and trying code found on Stack Overflow could now be answered in minutes if you knew how to write the right prompt.
• That was not good enough; copy/pasting pre-written prompts was inefficient.
• In November 2023 Open AI released GPTs. That was a useful improvement. I published several GPTs back then (see if you can pick where our logo is coming from):
  • AI Agent blog - Code Explainer GPT
  • AI Agent blog - Code Optimizer GPT
  • AI Agent blog - Code Coach GPT.

Competition Heats Up - The LLM Landscape Evolves:

Gen AI is evolving at break-neck speed.

Anthropic Claude became the best LLM for coding tasks in 2023. Open AI and Google Gemini responded with new, more powerful models.

If we wanted the best results, we needed a solution that could seamlessly switch between different LLMs and models to get the best results, for the best possible cost.

Context Is Key - LLMs Are Becoming A Commodity:

This evolving landscape reinforced a core belief: LLMs are becoming a commodity, but context is what matters most.

In a LinkedIn article I wrote in June 2024, I explained why providing the right, rich context to the AI is the most critical factor in achieving high-quality, relevant results. Without it, even the most powerful LLM will fall short.

The End Of Programming? Not So Fast…

Dr. Matt Welsh talks to Computer Science Students at Harvard - Oct 2023, and predicts:

• The end of programming as we know it.
• The arrival of “The Natural Language Computer”, that will allow anyone to build software using “Natural Language” (like English) as a programming language.

We Still Need Humans In The Loop (HITL):

AI will not replace you, someone who knows how to use AI will!

Generative AI is an incredibly powerful tool, but to unlock its full potential and ensure widespread, responsible adoption, we need to build trust.

We must acknowledge that AI can make mistakes – sometimes subtle ones – that can be hard to detect if we blindly rely on “Black Box AIs” that are coding FOR you, instead of WITH you.

The Airplane Analogy:

Like modern airplanes that have incredibly sophisticated GPS, autopilot systems, and countless other automations still have, and need, a human pilot, you still need programmers to build production ready software.

But now, they can build that with the help of AI!

We Need Explainable AI:

In an article I wrote back for Antler global back in August 2020, I discussed why we need explainable AI.

For AI to be truly useful and trustworthy, we need to understand how it arrives at its conclusions.

Explainable AI (XAI) helps demystify the “black box,” allowing users to understand the reasoning behind AI’s suggestions.

The Egg Theory of AI Agents:

The “Egg Theory” is from the 1950s when instant cake mixes initially sold poorly because they were too easy.

Sales only took off when manufacturers changed the recipe to require adding a fresh egg.

That simple act of cracking an egg made people feel like they were contributing, giving them a sense of ownership and control, even if the bulk of the work was done by the mix.

This applies directly to AI agents. As Rex Woodbury discussed in The Egg theory of AI Agents, even if an AI could do everything autonomously (and it’s not there yet), users often prefer to be involved, to make choices, and to feel in control.

It’s about AI as a “copilot,” augmenting human capabilities rather than completely replacing them.

This is crucial for adoption and trust.

The Genesis Of CodeCleaner AI:

I realized that what I (like other developers) really needed wasn’t just another code generator.

We needed a tool that:

• Prioritizes Education + Execution: Helps you learn while you build, not just churns out code.
• Puts You in Control (HITL): Acts as your assistant, providing suggestions and explanations, but always leaving the final decision to you. It should never modify your code directly.
• Leverages Context Intelligently: Understands your project, your existing code, and your specific goals to provide truly relevant guidance.
• Is LLM-Agnostic: Gives you the flexibility to use the best AI model for the task at hand.
• Promotes Explainability: Helps you understand why a suggestion is made, fostering deeper learning and trust.
• Respects Your Data: Operates with a clear commitment that none of your code is used to train any LLMs or AI models. Your intellectual property is yours alone.
• Addresses Real Developer Pain Points:
  • For Beginners: Provides the mentorship needed to bridge the gap between theory and practice, helping understand how to code and why bugs occur.
  • For Experienced Developers: Acts as an intelligent pair-programming buddy to debug faster, identify areas for improvement, and understand existing codebases more efficiently.
  • For Expert Reviewers: Assists in reviewing and validating code more quickly, identifying potential weaknesses and ensuring quality.

This is why we created CodeCleaner AI.

It’s designed to be the AI-powered mentor I wished I’d had, and the intelligent partner every developer deserves.

What’s Next?

We’re building the first AI platform that teaches while it codes — accelerating software development and helping every developer level up."

Our commitment is to empower you, the developer. To provide a tool that not only makes you more productive but also a better, more knowledgeable engineer.

We’re focused on building a platform that you can trust, that respects your autonomy, and that genuinely helps you grow.