Using AI to Build Practice

So, these are a few examples on how one can use AI to do what I like to call building practice.

Building Practice

Building practice is about systematically tackling a desired skill like 3d printing, playing the piano and so on, through the process of building a system that involves a closed looop environment for improving particular aspects of that skill.

Fetching data from documents

Document-based learning can be enhanced by uploading technical PDFs and instructing the model to answer questions specifically from that content, with references to exact pages. This creates a focused learning loop for understanding specific software functions or technical concepts. The model can also help locate relevant information online through a search action that can potentially save a lot of time on manual searches across multiple URLs.

Breakdown pattern

One fundamental pattern is what I call the breakdown pattern where you feed the model your ultimate goal and ask it to break it down into smaller, more manageable tasks or projects.

This creates a natural progression from simple to complex challenges, allowing you to build the necessary skills incrementally while maintaining an optimal balance of difficulty at each step.

Suggestion pattern

To complement this approach, the suggestion pattern helps maintain momentum in your learning journey. When you're unsure what to tackle next, you can describe your current skill level and completed projects to the model, which can then suggest appropriate next steps. This eliminates friction in decision-making and keeps you moving forward in your skill development.

Real-time assistance

Real-time assistance when learning a new software tool for example you can use something like Google AI Studio's real-time screen-sharing feature with Gemini to navigate complex software interfaces like AutoDesk Fusion 360.

While not perfect, the model can provide helpful guidance and suggestions for navigating complicated user interfaces. There are some cool experimental features being built on top of this like this Gemini Cursor.

There are also other interesting approaches related to this concept and the idea of computer use like the new tool from Microsoft: OmniParser 2.

More gemini-based apps can be found here: Awesome-Gemini-Apps.

Brainstorming

Brainstorming with AI becomes more effective when you carefully craft your prompts to receive specific, targeted interactive feedback. Rather than accepting vague advice from a generic conversation, you can structure the interaction to get precise, actionable information in your preferred format. For example, when practicing writing you might want the model to be more tuned to a certain style of writing like technical, or screenwriting and so on.

Planning learning goals

AI can also assist with planning by helping organize tasks around existing commitments and limitations. This extends beyond skill development to general productivity, allowing you to optimize your learning schedule and maintain steady progress.

You can prompt the model to breakdown a specific goal with a certain deadline into a set of tasks that should be executed within some pre-defined schedule (that could also be given by the model).

Maintaining flow

The overarching principle in using AI for skill development is maintaining flow state. It's crucial to avoid treating AI as a gimmick and instead craft specific, ultra-specialized interactions that provide exactly what you need. The goal isn't to replace your cognitive effort but to facilitate it.

By developing small-scale closed learning loops with clear feedback mechanisms, you create an environment where AI enhances rather than substitutes for your learning process.

This focused approach ensures that your interaction with AI directly supports your skill development while keeping you actively engaged in the learning process.