When even the most seniors don't know where to go in the AI era

When even the most seniors don’t know where to go in the AI era

In this conversation with Christiano Milfont from the ProdOps Channel, we talked a bit informally about some aspects of development with the advent of Artificial Intelligence.

Our agenda was as follows:

1) Market and AI Consolidation

We talked a bit about how the project-based model has its days numbered, as it no longer makes sense in terms of a timeline to have year-long projects when the cost of construction and experimentation is falling by orders of magnitude.

2) Technical Trends for 2025/2026

My bet will still be on Agents, given that there is a huge amount of low-relevance tasks involving consistency and repetition where agents will shine.

At least for me, I am certain that the next frontier will not just be about productivity, but agents will initially come for augmentation, and later to reduce human labor activity.

A point of contention we had was regarding MCP (Model Context Protocol), which I believe is a good interoperability standard between LLMs and applications/data in legacy contexts, but we still have doubts if this will be the standard.

One aspect we discussed is that Data Engineering will still be the cornerstone of AI, given that there is no AI without data.

Another point of contention was regarding Vibe Coding, where I think we are at the ideal sweet spot of abstractions between knowing programming and machine language.

3) Impact on Career and Professional Profile For Juniors, the scenario is hostile, and a large part of the low-relevance tasks that can be automated, AI can and will take over.

I think the best way to enter and stay in the field will be “selling shovels” in the gold rush, i.e., going into activities like data engineering, infrastructure, and SRE (Site Reliability Engineering) instead of going into Data Science or Machine Learning Research.

For seniors, the gain in leverage has changed the game, but those who refuse to understand and use it daily will be left behind and will become intentionally obsolete.

I think AI has somewhat leveled the playing field and, in an analogy, even Batman can fly now, and the differentiator is how to use this new set of powers daily, given that some previously exclusive skills have been fully diluted.

4) Learning Strategies

Milfont proposed an interesting strategy that I occasionally use, which is to use certification as a minimum knowledge roadmap model. In his case, he used data certifications from the big clouds as a study map.