Avatar

Francisco Moschetti

Uruguay.
Web developer alergic to leetcode.

Back to blog

My strategy for the AI future.

My bet is to double down and become more technical.

Yes, I don’t give a fuck if GPT 5.65 can one shot a OS kernel.

2 reasons why.

It’s what I enjoy.

There is so much to learn in this space man, we limit ourselves because the job market force us to become specialized in one space, one language, one framework.

There is 3 things I want to do until my last day and I consider constants of my life: lifting, coding, and gaming.

I don’t plan on changin those for anything.

AI as the user.

I’m, willing to bet that the main outcome from wide adoption of AI, is that the tooling we humans build will be the building blocks for these agents to do stuff.

There is going to be 3 types of tooling that will come out from this:

Orchestrators

The chat we use today will evolve and support more ways of interacting with LLMS: task boards, canvas, etc.

Here you will define workflows and the AI units that will perform said workflow.

This space will become commoditized and fase a ton of competition, High UX and fast itreations speeds will be needed to win here.

The only thing that can change this is if model providers close API access, that wont happen.

In 2 years you will have over 10 options of AI studios to work with llms and make agents.

Infrastructure providers

Anyone facilitating the task of putting something in production will have a good decade. PaaS, IaC, anything you can think of in this space.

The strat to win here is to have a wide service offering with code driven workflows that are optimized for LLMs.

Blocks (dev tools).

Here we have fancy tools that integrate at the code level to facilitate a development task like getstream.com or novu.co. And Anything that can be considered a block or a step of a more complicated workflow: A branding API that fetches logo and colors from a website, a email inbox provider via api. To win here you need APIs optimized for LLMs.

Conclusion

All those services are aimed to be used by LLMs, because now people will want to delegate login in to an app to do something.

And all the good oportunites left will be those that are unreachable by basic prompting and people that only have framework specific knowledge, a broad set of skills will let you see more of these oportunities.

One thing worth mentioning is that those changes are only for B2B, B2C will have a boom as entertainment use cases for this tech are in its infancy.