Journal Week 02 — Adapting AI to my day-to-day

Brief

Hello, continuing this journal, today I will keep talking a little about my onboarding process and include how using AI through a slightly more indirect approach sped it up considerably.

How much onboarding? Yes.

Yep, honestly, onboarding has been quite lengthy. Between learning about the company and its policies, as well as the team, product, processes, and operations, it has become a process that is NOT complicated, but… time-consuming.

I have had previous experience using CLI LLMs such as Codex, Claude, and OpenCode, whether for personal or professional use. In this case, I was able to speed up my learning process a little and reduce my learning bottleneck.

Using AI for notes

A small disclaimer: the strategy I followed may not be as effective for everyone; it varies according to preferences, but I think it can provide a good example of a simple use of AI.

To begin with, I took advantage of the fact that, for accessibility reasons, the video onboarding—which was the slowest part in my case—had subtitles. With a relatively quick script, which I also used AI to generate, I was able to convert them into text.

flowchart LR
    A[Onboarding videos] --> B[Script]
    B --> C[Plain text]

From there, I used AI to convert the plain text into Markdown with Mermaid diagrams, which I could digest more easily.

flowchart LR
    A[courses.txt] --> B[LLM]
    B --> C[courses.md]

From there, I saved that knowledge in folders, creating reliable documentation that I could read faster than a video. As a result of its effectiveness, I was able to complete the company onboarding more quickly (two days instead of one week…) and pass the subsequent quizzes.

For my application onboarding, I also followed a similar process, but added a skill I wanted to try, known as Graphify, by passing my folders through it.

flowchart LR
    A[courses.md] --> B[LLM: Graphify]
    B --> C[Knowledge graph]

This was especially useful for several reasons.

I was able to see related components in the docs and map this information in a more accessible way.
AI became a digestible wiki that I could compare with a real knowledge base.
This graph was especially useful because it gave the LLM fast, effective context that helped me understand more about the application.

Learning and progress

Beyond using AI effectively, I think practicing with these tools and using them day to day shows us a little more about the future of the industry.

Now I have reliable documentation that I have genuinely learned to use and apply in a way that is more tailored to me.

Next steps

I have begun to take on more responsibilities and have earned more trust from the team. I think I still need to improve my pace of work to learn more and contribute more effectively.
I want to streamline my workflow further. I think the resources the company provides can be used even better and can make work more enjoyable.

Reflections

This week, I have become a little more aware of AI’s impact on large technology companies. I have understood a little more about the fear people often have around AI replacing people, and I think there are valid reasons for it. As I said, I was able to lift part of my workload with it, but I still had to be there.

We are engineers, and historically we have had to adapt—and we always have—whether to new frameworks, tools that promise to take away our work, and so on. AI is another tool that is taking on the role of the new convenience for development, so… in retrospect, if we have always adapted, why should we stop adapting now? Why should AI be the limit of productive growth?

Anyone is free to say that they can now build an app with AI on their own. But clearly, if you tell them that both of you should make one with AI and see which is better, they will recognize that experience, technological background, and knowing what to ask for will make it far from an even competition. So, in my view, AI will be a +b in a development-learning line y, where engineering and the study of it are still the ax.