Notes on AI Tools and Workflows

When people talk about AI, the conversation often revolves around models, benchmarks and which tool is currently the smartest.

In practice, I have found that the biggest gains come from workflows rather than models.

Most people are looking for a tool that will build a product for them. What they should be looking for is a collection of tools that reduce friction at every stage of the product development process.

Today, it is entirely possible to design, prototype and launch real products using mostly free AI tools. The trick is understanding what each tool is good at and combining them effectively.

AI Is Not a Magic Button

One of the biggest misconceptions about AI is that it can replace thinking.

In reality, the best results come when AI acts as a collaborator rather than an authority.

I often think of AI tools as extremely fast interns.

They can generate ideas, create drafts, write code and explore possibilities. However, they still require direction, judgement and verification.

AI works best when it accelerates thinking rather than replaces it.

The people getting the most value from AI are rarely the ones asking it to do everything. They are the ones building systems that combine human judgement with machine speed.

Building Products With Free AI Tools

One of the most interesting developments over the past year has been the quality of free AI tools.

A few years ago, building software required strong technical expertise from day one. Today, AI significantly lowers the barrier to entry.

The combination I find particularly powerful includes:

• ChatGPT for ideation, planning and architecture.
• Google Stitch for interface design and wireframing.
• Lovable for rapid product development.
• Cursor for refining, debugging and extending code.

Individually, each tool is useful.

Combined, they create a surprisingly capable product development workflow.

My Workflow

I generally start with ChatGPT.

Rather than asking it to generate code immediately, I use it to think through product ideas, user journeys, edge cases and technical architecture.

At this stage, the goal is clarity.

What problem are we solving?

Who is the user?

What does a successful experience look like?

Once the product direction is clear, I move to Google Stitch.

Stitch is particularly useful for turning ideas into visual concepts quickly. Instead of spending hours creating wireframes manually, I can rapidly explore layouts, navigation structures and user flows.

The designs are not final products, but they create a solid foundation for the next stage.

After that, I move into Lovable.

This is where the product starts becoming real. Lovable can take designs and product requirements and turn them into working applications remarkably quickly.

Finally, Cursor becomes useful for refining the generated code, fixing issues, improving architecture and adding custom functionality.

Each tool contributes a different piece of the process.

Building Matiks Mathiss

One project where I followed this workflow was Matiks Mathiss, a reflex-based mathematics game inspired by classic arcade experiences.

The project started as a simple idea for a reflex-based mathematics game inspired by classic arcade experiences.

The first version of the concept was created using Google Stitch. This allowed me to experiment with layouts, game screens and user flows without writing code.

Once the overall experience felt right, I used ChatGPT to refine the architecture, think through gameplay mechanics and identify implementation details.

After that, Lovable was used to generate the actual application.

The generated foundation was surprisingly strong and allowed me to move from concept to working prototype much faster than a traditional development process.

From there, the focus shifted towards refinement, testing and iteration.

The final result was a playable product built through a combination of design tools, AI-assisted planning and AI-generated code.

Workflows Matter More Than Prompts

Many discussions about AI focus on prompts.

While prompting is important, I think workflows matter far more.

A good workflow creates repeatability.

Instead of asking a model to perform a task once, you create a process that consistently produces useful results.

The biggest productivity gains often come from identifying repetitive work and building systems around it.

For example:

• Research workflows.
• Product planning workflows.
• Analytics workflows.
• Content creation workflows.
• Coding workflows.

Once AI becomes part of a repeatable process, its value compounds over time.

Keep Humans in the Loop

Despite all the progress, AI still makes mistakes.

Hallucinations exist.

Generated code can contain flaws.

Design suggestions are not always optimal.

This is why I always keep a human review step in the process.

AI is excellent at generating options.

Humans remain responsible for selecting the right option.

The best results come from combining machine speed with human judgement.

Final Thoughts

The future of product building is not about replacing people with AI.

It is about giving individuals leverage.

Today, a single person can move from idea to prototype dramatically faster than before by combining tools such as ChatGPT, Google Stitch, Lovable and Cursor.

The real opportunity is not using AI as a magic button.

It is learning how to combine multiple tools into workflows that consistently turn ideas into working products.

The tools will continue to improve. The people who benefit the most will be those who learn how to build effective systems around them.