I Started A Coding Habit & You Should Too...

Key Concepts

• Non-Goal Oriented Coding/Studying: A deliberate practice of coding or studying purely for enjoyment and exploration, without a specific project goal or deadline.
• Vibe Coding: A term used to describe coding driven by exploration and enjoyment rather than strict problem-solving.
• Multimodal Embedding: Combining embedding spaces of different data types (e.g., images and text) into a unified representation.
• Vector Store: A database used to store and efficiently search vector embeddings.
• Proactive Portfolio Building: Developing a consistent coding history on platforms like GitHub to demonstrate continuous learning and skill development.

Benefits of a Non-Goal Oriented Coding Habit

The speaker advocates for a daily or near-daily coding/studying habit focused on exploration rather than achieving specific outcomes. He distinguishes this from his regular work (startup, freelancing, video preparation) which is goal-oriented. The core benefit lies in the enjoyment of coding itself, leading to deeper understanding and unexpected opportunities.

Specifically, he highlights:

• Enhanced Skill Development: Even without a direct application, consistent practice improves programming, engineering, and business skills. “You will become a better programmer. You will become a better engineer, uh a better business person, depending on what you're doing, of course, uh you will become better at what you're doing just by playing around with stuff.”
• Improved Knowledge Retention: Revisiting previously learned material, even from a theoretical perspective, reinforces understanding. He’s currently re-working through “Hands-On Machine Learning with Scikit-Learn and TensorFlow,” skipping TensorFlow specifics in favor of PyTorch, but finding value in revisiting core concepts.
• Increased Competence in Research: Regularly studying foundational topics like linear algebra improves the ability to understand and apply research papers. “Having good linear algebra skills, it’s just nice when you read papers.”
• Idea Generation: Exploratory projects often spark new video ideas and potential business ventures as a byproduct.
• Portfolio Building: Consistent activity on platforms like GitHub demonstrates proactive learning and skill development to potential employers or clients.

Current Projects & Approach

The speaker details two ongoing personal projects illustrating his approach:

1. Latent Assets (Image/Text Embedding Tool):

• Problem: Difficulty managing and reusing assets (logos, images) when creating thumbnails for videos in Canva. Canva duplicates assets with each upload.
• Solution: A web application (FastAPI backend, React frontend) that uses a multimodal embedding model (Gina CLIP v2) to embed images and text into a shared vector space.
• Functionality: Users can upload images and add text descriptions. The model intelligently identifies similar assets, even without exact tag matches. A search function allows quick retrieval of assets, and clicking on an asset copies it to the clipboard. Manual tagging is also supported for ambiguous cases.
• Development Process: Initial exploration of the inference API, vector store interaction, and the CIFAR-10 dataset were done manually. The code was then generated using Claude Code (or a similar tool) based on the manually created logic.
• Technical Details: Utilizes FastAPI for the backend, React for the frontend, and Gina CLIP v2 for multimodal embeddings. The vector store is used for efficient similarity searches.

2. Mail Deer Stat (Email Inbox Visualization):

• Problem: A full email inbox with a 1GB limit, making it difficult to identify and delete unnecessary emails.
• Solution: A Flask application that visualizes emails as squares, with size representing email size. Allows grouping by sender to identify space-consuming accounts.
• Functionality: Displays emails as interactive squares, allowing users to preview content and identify large or irrelevant emails for deletion.
• Inspiration: Inspired by QR Stat (a disk usage visualization tool for KDE).
• Technical Details: Built with Flask, utilizing a similar approach to Latent Assets – manual exploration followed by code generation with an AI tool.

Inspiration & Role Models

The speaker cites Peter Steinbergger (creator of OpenCLaw) as a key inspiration. He wasn’t impressed by OpenCLaw’s success, but by Steinbergger’s consistent activity on GitHub, demonstrating a habit of building projects to “scratch his own itch.” “I wanted to follow this example and just start building stuff without having a goal in mind like making a video about it or selling it to someone or making a product out of it.”

Methodology & Recommendations

The speaker recommends a simple methodology:

1. Dedicate Time: Reserve 30-60 minutes daily for non-goal-oriented coding/studying.
2. Follow Your Interests: Work on projects that genuinely interest you.
3. Embrace Exploration: Don’t worry about usefulness or impressiveness. A “Hello World” program in a new language is a valid starting point.
4. Leverage AI Tools: Use tools like Claude Code to accelerate development after establishing the core logic manually.
5. Focus on Consistency: Prioritize consistent activity over ambitious projects. A steady stream of small projects builds a stronger portfolio than infrequent, large-scale efforts.

Notable Quotes

• “It shouldn’t be an obligation. It shouldn’t feel forced.” – Describing the ideal mindset for this habit.
• “You’re not doing it because you need to do something by the end of next week.” – Highlighting the key difference between goal-oriented and exploratory coding.
• “This is actually made for an audience of one. This is made for me.” – Emphasizing that the projects are driven by personal need and enjoyment.

Conclusion

The speaker advocates for a deliberate practice of coding and studying for the sheer joy of it. This approach, characterized by exploration, consistency, and a lack of immediate goals, fosters skill development, generates new ideas, and builds a compelling portfolio. The key takeaway is to carve out dedicated time for personal projects driven by curiosity, leveraging AI tools to accelerate development while maintaining a focus on learning and enjoyment.